diff --git a/CLAUDE.md b/CLAUDE.md
index 7dba5283..54243962 100644
--- a/CLAUDE.md
+++ b/CLAUDE.md
@@ -84,6 +84,7 @@ The frontend (`frontend-web/`) is a React 19 + Vite SPA:
 - **LLM streaming**: `@langchain/langgraph-sdk` streams messages from the backend via SSE
 - **UI**: Radix UI primitives + TailwindCSS v4 + shadcn-style components in `src/components/ui/`
 - **Shims**: `src/shims/` provides `next/link` and `next/navigation` polyfills (original code was Next.js)
+- **Login URL params** (`LoginPage.tsx`, frontend-only — the backend never sees these): besides the auth params (`ticket`/`userId`/`env`/`theme`/`yUserId`/`token`/`access_token`/`authToken`/`password`), the page supports a **task deep-link** via `?taskId=&goPath=`. After login it calls the consumer API `GET {VITE_AIPI_CONSUMER_URL}/taskAnalyseSearch/cop-task-detail?taskId=` (header `Authorization: Bearer admin`, auto-injected by `utils/request.ts`), stashes the full `data` (`id/overview/taskDirection/taskName/taskContent`) into `sessionStorage["login.copTaskDetail"]`, then routes by `goPath`: base codes `3qfx`/`rwfx`/`xdfx` → that business's configured **single agent** Q&A (resolved via `/api/business-mapping`, opening message in `sessionStorage.pendingAgentMessage`); the `-A` suffix (`3qfx-A`/…) → **multi-agent roundtable** with `roundtablePrefill.businessCode` so the page auto-selects the configured 业务链条. goPath→businessCode: `3qfx`→`3Q`, `rwfx`→`6BF`, `xdfx`→`7BF`. **Both targets render embedded under the 任务管理 layout** — nested routes under `task-management/<business-path>/{roundtable,agent/:agent_id/chats/:thread_id}` in `StrategyRoutes.tsx` (`<business-path>` = `plan-task-sentiment` / `plan-task/plan-task-fx` / `plan-task/plan-task-action`, matching the COH中任务 menu paths in `sidebar-menu.ts`) so the global sidebar keeps 任务管理 → COH中任务 → the matching business highlighted. The agent chat is rendered via `TaskEmbeddedAgentChat`, which passes `AgentChatPage`'s `chatBasePath` prop so the post-send URL rewrite + new-chat button stay in the task-management context (instead of the standalone `/agents/...` route). Any failure (no consumer URL / fetch error / unmapped goPath / no configured agent) silently falls back to the default home. Helpers: `core/auth/task-deeplink.ts`.
 
 ### Configuration
 
diff --git a/frontend-web/.env.development b/frontend-web/.env.development
index d5dfe0f9..22563825 100644
--- a/frontend-web/.env.development
+++ b/frontend-web/.env.development
@@ -4,6 +4,9 @@
 # 任务看板数据源：true=真实接口（默认）；改成 false 可让任务分析/三情分析两页走本地假数据。
 VITE_USE_REAL_DATA=true
 
+# 任务详情（登录深链 cop-task-detail）数据源：默认真实接口；改成 true 走本地假数据（consumer 外网不可达时调试用）。
+VITE_MOCK_COP_TASK_DETAIL=false
+
 # Gateway / DeerFlow API
 VITE_BACKEND_BASE_URL=http://127.0.0.1:8001
 VITE_LANGGRAPH_BASE_URL=http://127.0.0.1:8001/api
@@ -27,7 +30,7 @@ VITE_AIPI_B2ACTOR_XWT_PATH=/fromB/toXWT?isFrame=1
 VITE_AIPI_KNOWLEDGE_IFRAME_API_URL=https://ch1.b.uat.4.cn/knowes/api/iframeKnowledgeUrl
 VITE_AIPI_KNOWLEDGE_IFRAME_SKIN=red
 VITE_AIPI_MOPECHAT_URL=https://ch1.b.uat.4.cn/likun_b1_mope
-VITE_AIPI_BACKENP_URL=http://localhost:8003
+VITE_AIPI_BACKENP_URL=http://47.88.25.99:8000
 # 工作室/空间后端接口（侧边栏“工作室”列表、空间详情等都走这里，不是 VITE_NOTE_API_BASE_URL）
 # 工作室/空间前端地址（用于 auth/callback 等跳转）
 # VITE_AIPI_FRONTEND_URL=https://knowledgetoolserver.b.uat.4.cn
diff --git a/frontend-web/.env.example b/frontend-web/.env.example
index e254e6c8..1b3d80bb 100644
--- a/frontend-web/.env.example
+++ b/frontend-web/.env.example
@@ -33,6 +33,11 @@
 # 仅本机 .env 默认 false（假数据）。改后需重启 dev / 重新打包生效。
 # VITE_USE_REAL_DATA=true
 
+# --- 任务详情（登录深链 cop-task-detail）数据源开关 ---
+# 默认真实接口（未设/false）；改成 true 时该接口走本地假数据（consumer 外网不可达时调试用）。
+# 与 VITE_USE_REAL_DATA 独立。改后需重启 dev / 重新打包生效。
+# VITE_MOCK_COP_TASK_DETAIL=false
+
 # MaxKey 鉴权根地址（无界 onlineTicket 换 token）
 # VITE_AUTH_BASE_URL=https://auth.y.sit.4.cn
 # VITE_AIPI_CONSUMER_URL=https://ch1.b.uat.4.cn/consumer
diff --git a/frontend-web/.env.network b/frontend-web/.env.network
index 569b8422..471b6289 100644
--- a/frontend-web/.env.network
+++ b/frontend-web/.env.network
@@ -4,6 +4,9 @@
 # 任务看板数据源：true=真实接口（默认）；改成 false 可让任务分析/三情分析两页走本地假数据。
 VITE_USE_REAL_DATA=true
 
+# 任务详情（登录深链 cop-task-detail）数据源：默认真实接口；改成 true 走本地假数据（调试用）。
+VITE_MOCK_COP_TASK_DETAIL=false
+
 # Gateway / DeerFlow API
 VITE_BACKEND_BASE_URL=http://47.88.25.99:7001
 VITE_LANGGRAPH_BASE_URL=http://47.88.25.99:7001/api
diff --git a/frontend-web/.env.production b/frontend-web/.env.production
index 08bf2a70..97415742 100644
--- a/frontend-web/.env.production
+++ b/frontend-web/.env.production
@@ -4,6 +4,9 @@
 # 任务看板数据源：true=真实接口（默认）；改成 false 可让任务分析/三情分析两页走本地假数据。
 VITE_USE_REAL_DATA=true
 
+# 任务详情（登录深链 cop-task-detail）数据源：默认真实接口；改成 true 走本地假数据（调试用）。
+VITE_MOCK_COP_TASK_DETAIL=false
+
 # Gateway / DeerFlow API
 VITE_BACKEND_BASE_URL=https://ch1.b.uat.4.cn/deerflow
 VITE_LANGGRAPH_BASE_URL=https://ch1.b.uat.4.cn/deerflow/api
diff --git a/frontend-web/.env.sit b/frontend-web/.env.sit
index d4a4af7b..91ba650f 100644
--- a/frontend-web/.env.sit
+++ b/frontend-web/.env.sit
@@ -4,6 +4,9 @@
 # 任务看板数据源：true=真实接口（默认）；改成 false 可让任务分析/三情分析两页走本地假数据。
 VITE_USE_REAL_DATA=true
 
+# 任务详情（登录深链 cop-task-detail）数据源：默认真实接口；改成 true 走本地假数据（调试用）。
+VITE_MOCK_COP_TASK_DETAIL=false
+
 # Gateway / DeerFlow API
 VITE_BACKEND_BASE_URL=https://ch1.b.sit.4.cn/deerflow
 VITE_LANGGRAPH_BASE_URL=https://ch1.b.sit.4.cn/deerflow/api
diff --git a/frontend-web/src/App.tsx b/frontend-web/src/App.tsx
index 0c191a09..88ee59df 100644
--- a/frontend-web/src/App.tsx
+++ b/frontend-web/src/App.tsx
@@ -86,12 +86,22 @@ function AppProviders({ children }: { children: React.ReactNode }) {
   );
 }
 
+/**
+ * 根路由 `/` → `/login` 重定向，**保留 query string**。
+ * 外部系统的任务深链落在 `#/?taskId=...&goPath=...`，裸 `<Navigate to="/login">` 会丢掉
+ * query，导致 LoginPage 读不到参数；这里把 search 一并带过去。
+ */
+function RootRedirect() {
+  const { search } = useLocation();
+  return <Navigate to={`/login${search}`} replace />;
+}
+
 export function App() {
   return (
     <HashRouter>
       <AppProviders>
         <Routes>
-          <Route path="/" element={<Navigate to="/login" replace />} />
+          <Route path="/" element={<RootRedirect />} />
           <Route path="/login" element={<LoginPage />} />
           <Route path="/login/:username" element={<LoginPage />} />
           <Route path="/landing" element={<LandingPage />} />
diff --git a/frontend-web/src/components/wiki/KnowledgeManage.tsx b/frontend-web/src/components/wiki/KnowledgeManage.tsx
new file mode 100644
index 00000000..de42b5ea
--- /dev/null
+++ b/frontend-web/src/components/wiki/KnowledgeManage.tsx
@@ -0,0 +1,429 @@
+import { Plus, Trash2 } from "lucide-react";
+import { useEffect, useMemo, useState } from "react";
+import { toast } from "sonner";
+
+import { Button } from "@/components/ui/button";
+import {
+  Dialog,
+  DialogContent,
+  DialogDescription,
+  DialogFooter,
+  DialogHeader,
+  DialogTitle,
+} from "@/components/ui/dialog";
+import { Input } from "@/components/ui/input";
+import { ScrollArea } from "@/components/ui/scroll-area";
+import {
+  Select,
+  SelectContent,
+  SelectItem,
+  SelectTrigger,
+  SelectValue,
+} from "@/components/ui/select";
+import { Textarea } from "@/components/ui/textarea";
+import {
+  useExtractTemplateActions,
+  useExtractTemplates,
+  useFolderActions,
+  useKnowledgeFolders,
+} from "@/core/knowledge/hooks";
+import type {
+  ExtractTemplateScope,
+  KnowledgeExtractTemplate,
+} from "@/core/knowledge/types";
+import { cn } from "@/lib/utils";
+
+const NONE = "__none__";
+
+/** A dropdown of managed directories (+ an "unfiled" option). */
+export function FolderSelect({
+  value,
+  onChange,
+  disabled,
+  triggerClassName,
+  allowNone = true,
+}: {
+  value: string | null;
+  onChange: (v: string | null) => void;
+  disabled?: boolean;
+  triggerClassName?: string;
+  allowNone?: boolean;
+}) {
+  const { data: folders = [] } = useKnowledgeFolders();
+  return (
+    <Select
+      value={value ?? NONE}
+      onValueChange={(v) => onChange(v === NONE ? null : v)}
+      disabled={disabled}
+    >
+      <SelectTrigger className={cn("h-9", triggerClassName)}>
+        <SelectValue placeholder="选择目录" />
+      </SelectTrigger>
+      <SelectContent>
+        {allowNone ? <SelectItem value={NONE}>（不归类）</SelectItem> : null}
+        {folders.map((f) => (
+          <SelectItem key={f.id} value={f.path}>
+            {f.path}
+          </SelectItem>
+        ))}
+      </SelectContent>
+    </Select>
+  );
+}
+
+/** A dropdown of enabled extraction templates applicable to a scope. */
+export function TemplateSelect({
+  value,
+  onChange,
+  scope,
+  disabled,
+  triggerClassName,
+}: {
+  value: string | null;
+  onChange: (v: string | null) => void;
+  scope: ExtractTemplateScope;
+  disabled?: boolean;
+  triggerClassName?: string;
+}) {
+  const { data: templates = [] } = useExtractTemplates(scope, true);
+  return (
+    <Select
+      value={value ?? NONE}
+      onValueChange={(v) => onChange(v === NONE ? null : v)}
+      disabled={disabled}
+    >
+      <SelectTrigger className={cn("h-9", triggerClassName)}>
+        <SelectValue placeholder="默认模板" />
+      </SelectTrigger>
+      <SelectContent>
+        <SelectItem value={NONE}>默认模板</SelectItem>
+        {templates.map((t) => (
+          <SelectItem key={t.id} value={t.id}>
+            {t.name}
+            {t.is_default ? " · 默认" : ""}
+          </SelectItem>
+        ))}
+      </SelectContent>
+    </Select>
+  );
+}
+
+/** Create / rename / delete knowledge directories. */
+export function FolderManageDialog({
+  open,
+  onOpenChange,
+}: {
+  open: boolean;
+  onOpenChange: (v: boolean) => void;
+}) {
+  const { data: folders = [], isLoading } = useKnowledgeFolders();
+  const { create, rename, remove } = useFolderActions();
+  const [newPath, setNewPath] = useState("");
+
+  const handleCreate = async () => {
+    const path = newPath.trim();
+    if (!path) {
+      toast.error("请输入目录路径");
+      return;
+    }
+    try {
+      await create.mutateAsync(path);
+      toast.success("目录已创建");
+      setNewPath("");
+    } catch (error) {
+      toast.error(error instanceof Error ? error.message : "创建失败");
+    }
+  };
+
+  const handleRename = async (folderId: string, current: string) => {
+    const next = window.prompt("重命名 / 移动目录（输入新的完整路径，斜杠分隔）", current);
+    if (next === null) return;
+    try {
+      await rename.mutateAsync({ folderId, path: next });
+      toast.success("已更新");
+    } catch (error) {
+      toast.error(error instanceof Error ? error.message : "更新失败");
+    }
+  };
+
+  const handleDelete = async (folderId: string, path: string) => {
+    if (!window.confirm(`删除目录「${path}」？其中的知识将变为未归类（不会被删除）。`)) return;
+    try {
+      await remove.mutateAsync({ folderId });
+      toast.success("目录已删除");
+    } catch (error) {
+      toast.error(error instanceof Error ? error.message : "删除失败");
+    }
+  };
+
+  return (
+    <Dialog open={open} onOpenChange={onOpenChange}>
+      <DialogContent className="max-w-lg">
+        <DialogHeader>
+          <DialogTitle>目录管理</DialogTitle>
+          <DialogDescription>
+            自建知识目录,用斜杠分隔多级(如「投研/行业」)。删除目录只会让其中的知识变为未归类。
+          </DialogDescription>
+        </DialogHeader>
+        <div className="flex items-center gap-2">
+          <Input
+            placeholder="新目录路径,如 投研/行业"
+            value={newPath}
+            onChange={(e) => setNewPath(e.target.value)}
+            onKeyDown={(e) => {
+              if (e.key === "Enter") void handleCreate();
+            }}
+          />
+          <Button onClick={() => void handleCreate()} disabled={create.isPending}>
+            <Plus className="size-4" /> 新建
+          </Button>
+        </div>
+        <ScrollArea className="max-h-[50vh]">
+          {isLoading ? (
+            <div className="text-muted-foreground p-3 text-sm">加载中...</div>
+          ) : folders.length === 0 ? (
+            <div className="text-muted-foreground rounded-md border border-dashed p-6 text-center text-sm">
+              还没有目录,先在上方新建一个。
+            </div>
+          ) : (
+            <ul className="flex flex-col gap-1 p-1">
+              {folders.map((f) => (
+                <li
+                  key={f.id}
+                  className="flex items-center justify-between gap-2 rounded-md border px-3 py-2 text-sm dark:border-slate-700"
+                >
+                  <span className="truncate font-mono text-[13px]">{f.path}</span>
+                  <div className="flex shrink-0 items-center gap-1">
+                    <Button size="sm" variant="ghost" onClick={() => void handleRename(f.id, f.path)}>
+                      重命名
+                    </Button>
+                    <Button
+                      size="sm"
+                      variant="ghost"
+                      className="text-red-600 hover:text-red-700"
+                      onClick={() => void handleDelete(f.id, f.path)}
+                    >
+                      <Trash2 className="size-4" />
+                    </Button>
+                  </div>
+                </li>
+              ))}
+            </ul>
+          )}
+        </ScrollArea>
+        <DialogFooter>
+          <Button variant="outline" onClick={() => onOpenChange(false)}>
+            关闭
+          </Button>
+        </DialogFooter>
+      </DialogContent>
+    </Dialog>
+  );
+}
+
+const SCOPE_LABELS: Record<ExtractTemplateScope, string> = {
+  thread: "对话",
+  document: "文档",
+  both: "通用",
+};
+
+const EMPTY_DRAFT = {
+  name: "",
+  scope: "both" as ExtractTemplateScope,
+  description: "",
+  system_prompt: "",
+  is_default: false,
+  enabled: true,
+};
+
+/** Create / edit / delete the editable extraction-prompt templates. */
+export function ExtractTemplateManageDialog({
+  open,
+  onOpenChange,
+}: {
+  open: boolean;
+  onOpenChange: (v: boolean) => void;
+}) {
+  const { data: templates = [], isLoading } = useExtractTemplates();
+  const { create, update, remove } = useExtractTemplateActions();
+  const [selectedId, setSelectedId] = useState<string | null>(null);
+  const [draft, setDraft] = useState({ ...EMPTY_DRAFT });
+
+  const selected = useMemo(
+    () => templates.find((t) => t.id === selectedId) ?? null,
+    [templates, selectedId],
+  );
+
+  // Load the selected template into the editable draft (or reset for "new").
+  useEffect(() => {
+    if (selected) {
+      setDraft({
+        name: selected.name,
+        scope: selected.scope,
+        description: selected.description ?? "",
+        system_prompt: selected.system_prompt,
+        is_default: selected.is_default,
+        enabled: selected.enabled,
+      });
+    } else {
+      setDraft({ ...EMPTY_DRAFT });
+    }
+  }, [selected]);
+
+  const startNew = () => setSelectedId(null);
+
+  const handleSave = async () => {
+    if (!draft.name.trim() || !draft.system_prompt.trim()) {
+      toast.error("请填写名称与提示词正文");
+      return;
+    }
+    try {
+      if (selectedId) {
+        await update.mutateAsync({ templateId: selectedId, data: draft });
+        toast.success("已保存");
+      } else {
+        const created = await create.mutateAsync(draft);
+        setSelectedId(created.id);
+        toast.success("模板已创建");
+      }
+    } catch (error) {
+      toast.error(error instanceof Error ? error.message : "保存失败");
+    }
+  };
+
+  const handleDelete = async (t: KnowledgeExtractTemplate) => {
+    if (!window.confirm(`删除模板「${t.name}」？`)) return;
+    try {
+      await remove.mutateAsync(t.id);
+      if (selectedId === t.id) setSelectedId(null);
+      toast.success("已删除");
+    } catch (error) {
+      toast.error(error instanceof Error ? error.message : "删除失败");
+    }
+  };
+
+  return (
+    <Dialog open={open} onOpenChange={onOpenChange}>
+      <DialogContent className="w-[95vw] max-w-[95vw] sm:max-w-[1800px]">
+        <DialogHeader>
+          <DialogTitle>抽取提示词模板</DialogTitle>
+          <DialogDescription>
+            编辑驱动「知识沉淀」的提炼提示词,可新建多套抽取逻辑。沉淀对话/导入文件时可选用对应模板;未选则用该类型的默认模板。
+          </DialogDescription>
+        </DialogHeader>
+        <div className="grid grid-cols-[240px_minmax(0,1fr)] gap-4">
+          {/* template list */}
+          <div className="flex min-h-0 flex-col gap-2">
+            <Button size="sm" variant={selectedId === null ? "default" : "outline"} onClick={startNew}>
+              <Plus className="size-4" /> 新建模板
+            </Button>
+            <ScrollArea className="max-h-[50vh] rounded-md border dark:border-slate-700">
+              {isLoading ? (
+                <div className="text-muted-foreground p-3 text-sm">加载中...</div>
+              ) : (
+                <ul className="flex flex-col">
+                  {templates.map((t) => (
+                    <li key={t.id}>
+                      <button
+                        type="button"
+                        onClick={() => setSelectedId(t.id)}
+                        className={cn(
+                          "flex w-full items-center justify-between gap-2 border-b px-3 py-2 text-left text-sm last:border-b-0 dark:border-slate-800",
+                          selectedId === t.id ? "bg-slate-100 dark:bg-slate-800" : "hover:bg-slate-50 dark:hover:bg-slate-800/50",
+                        )}
+                      >
+                        <span className="min-w-0">
+                          <span className="block truncate">{t.name}</span>
+                          <span className="text-muted-foreground text-[11px]">
+                            {SCOPE_LABELS[t.scope]}
+                            {t.is_default ? " · 默认" : ""}
+                            {t.enabled ? "" : " · 已停用"}
+                          </span>
+                        </span>
+                      </button>
+                    </li>
+                  ))}
+                </ul>
+              )}
+            </ScrollArea>
+          </div>
+
+          {/* editor */}
+          <div className="flex flex-col gap-3">
+            <div className="flex items-center gap-2">
+              <Input
+                placeholder="模板名称"
+                value={draft.name}
+                onChange={(e) => setDraft((d) => ({ ...d, name: e.target.value }))}
+              />
+              <Select
+                value={draft.scope}
+                onValueChange={(v) => setDraft((d) => ({ ...d, scope: v as ExtractTemplateScope }))}
+              >
+                <SelectTrigger className="w-28">
+                  <SelectValue />
+                </SelectTrigger>
+                <SelectContent>
+                  <SelectItem value="both">通用</SelectItem>
+                  <SelectItem value="thread">对话</SelectItem>
+                  <SelectItem value="document">文档</SelectItem>
+                </SelectContent>
+              </Select>
+            </div>
+            <Input
+              placeholder="简介(可选)"
+              value={draft.description}
+              onChange={(e) => setDraft((d) => ({ ...d, description: e.target.value }))}
+            />
+            <Textarea
+              rows={16}
+              placeholder="提示词正文(system prompt)。这里直接控制提炼粒度——想保留更多信息,就在此放宽要求、要求结构化更详细的输出。"
+              value={draft.system_prompt}
+              onChange={(e) => setDraft((d) => ({ ...d, system_prompt: e.target.value }))}
+              className="h-[50vh] max-h-[50vh] resize-none overflow-y-auto font-mono text-xs"
+            />
+            <div className="flex flex-wrap items-center gap-4 text-sm">
+              <label className="flex items-center gap-1.5">
+                <input
+                  type="checkbox"
+                  checked={draft.is_default}
+                  onChange={(e) => setDraft((d) => ({ ...d, is_default: e.target.checked }))}
+                  className="size-4"
+                />
+                设为该类型默认
+              </label>
+              <label className="flex items-center gap-1.5">
+                <input
+                  type="checkbox"
+                  checked={draft.enabled}
+                  onChange={(e) => setDraft((d) => ({ ...d, enabled: e.target.checked }))}
+                  className="size-4"
+                />
+                启用
+              </label>
+              <div className="ml-auto flex items-center gap-2">
+                {selected ? (
+                  <Button
+                    variant="ghost"
+                    className="text-red-600 hover:text-red-700"
+                    onClick={() => void handleDelete(selected)}
+                  >
+                    <Trash2 className="size-4" /> 删除
+                  </Button>
+                ) : null}
+                <Button onClick={() => void handleSave()} disabled={create.isPending || update.isPending}>
+                  {selectedId ? "保存" : "创建"}
+                </Button>
+              </div>
+            </div>
+          </div>
+        </div>
+        <DialogFooter>
+          <Button variant="outline" onClick={() => onOpenChange(false)}>
+            关闭
+          </Button>
+        </DialogFooter>
+      </DialogContent>
+    </Dialog>
+  );
+}
diff --git a/frontend-web/src/components/wiki/WikiContent.tsx b/frontend-web/src/components/wiki/WikiContent.tsx
index 354f190b..e7d2a93f 100644
--- a/frontend-web/src/components/wiki/WikiContent.tsx
+++ b/frontend-web/src/components/wiki/WikiContent.tsx
@@ -22,6 +22,7 @@ import type {
 
 import { Breadcrumb } from "./Breadcrumb";
 import { DocumentToolbar } from "./DocumentToolbar";
+import { FolderSelect } from "./KnowledgeManage";
 import { FloatingToc } from "./FloatingToc";
 import { MarkdownViewer } from "./MarkdownViewer";
 import { RelatedDocs } from "./RelatedDocs";
@@ -35,6 +36,7 @@ import {
   downloadMarkdown,
   extractToc,
   formatDateTime,
+  localizeHeadings,
   noteBreadcrumbSegments,
   splitRelatedSection,
 } from "./wikiUtils";
@@ -102,10 +104,14 @@ export function WikiContent({
 
   // Pull the backend-injected `## Related` wikilinks out of the body so they
   // render as a structured module; editing still works on the full source.
-  const { body, related } = useMemo(
+  // After extraction, localize the English section headings (Context / Finding
+  // / Reasoning / …) to Chinese for display — `extractToc` and the viewer share
+  // this transformed text so heading anchors stay consistent.
+  const { body: rawBody, related } = useMemo(
     () => splitRelatedSection(note.content_md),
     [note.content_md],
   );
+  const body = useMemo(() => localizeHeadings(rawBody), [rawBody]);
   const toc = useMemo(() => extractToc(body), [body]);
   // The floating outline only shows h2/h3 — deeper levels are noise.
   const outlineItems = useMemo(
@@ -177,6 +183,15 @@ export function WikiContent({
     }
   };
 
+  const handleMoveFolder = async (folder: string | null) => {
+    try {
+      await update.mutateAsync({ noteId: note.id, data: { folder } });
+      toast.success(folder ? `已移动到「${folder}」` : "已移出目录");
+    } catch (error) {
+      toast.error(error instanceof Error ? error.message : "移动失败");
+    }
+  };
+
   const handleArchive = async () => {
     try {
       await remove.mutateAsync({ noteId: note.id });
@@ -277,6 +292,15 @@ export function WikiContent({
                   {note.created_at ? (
                     <MetaTag>创建：{formatDateTime(note.created_at).slice(0, 10)}</MetaTag>
                   ) : null}
+                  <span className="inline-flex items-center gap-1 text-xs text-slate-500 dark:text-slate-400">
+                    目录
+                    <FolderSelect
+                      value={note.folder ?? null}
+                      onChange={(v) => void handleMoveFolder(v)}
+                      disabled={update.isPending}
+                      triggerClassName="h-6 w-[10rem] text-xs"
+                    />
+                  </span>
                   {note.tags.map((tag) => (
                     <button
                       key={tag}
diff --git a/frontend-web/src/components/wiki/WikiHeader.tsx b/frontend-web/src/components/wiki/WikiHeader.tsx
index b278ad49..9c5e6429 100644
--- a/frontend-web/src/components/wiki/WikiHeader.tsx
+++ b/frontend-web/src/components/wiki/WikiHeader.tsx
@@ -1,9 +1,11 @@
 import {
   Copy,
+  FolderTree,
   Menu,
   Network,
   Plus,
   RefreshCw,
+  SlidersHorizontal,
   Upload,
 } from "lucide-react";
 
@@ -48,6 +50,8 @@ export function WikiHeader({
   onDedup,
   onGraph,
   onCreate,
+  onManageFolders,
+  onManageTemplates,
 }: {
   onToggleSidebar: () => void;
   onImportFile: () => void;
@@ -55,6 +59,8 @@ export function WikiHeader({
   onDedup: () => void;
   onGraph: () => void;
   onCreate: () => void;
+  onManageFolders: () => void;
+  onManageTemplates: () => void;
 }) {
   return (
     <header className="flex h-14 shrink-0 items-center justify-between border-b border-slate-100 bg-white px-4 sm:px-6 dark:border-slate-800 dark:bg-slate-900">
@@ -83,6 +89,16 @@ export function WikiHeader({
           label="批量导入"
           onClick={onBatchImport}
         />
+        <HeaderButton
+          icon={<FolderTree className="size-4" />}
+          label="目录管理"
+          onClick={onManageFolders}
+        />
+        <HeaderButton
+          icon={<SlidersHorizontal className="size-4" />}
+          label="抽取模板"
+          onClick={onManageTemplates}
+        />
         <HeaderButton
           icon={<Copy className="size-4" />}
           label="查重"
diff --git a/frontend-web/src/components/wiki/wikiTypes.ts b/frontend-web/src/components/wiki/wikiTypes.ts
index 3d81b7e6..66412e16 100644
--- a/frontend-web/src/components/wiki/wikiTypes.ts
+++ b/frontend-web/src/components/wiki/wikiTypes.ts
@@ -14,4 +14,6 @@ export interface TreeNode {
   type: "folder" | "file";
   children?: TreeNode[];
   item?: KnowledgeNote;
+  /** Full slash-joined directory path (folder nodes only). */
+  folderPath?: string;
 }
diff --git a/frontend-web/src/components/wiki/wikiUtils.ts b/frontend-web/src/components/wiki/wikiUtils.ts
index 9500fa93..5625ffcc 100644
--- a/frontend-web/src/components/wiki/wikiUtils.ts
+++ b/frontend-web/src/components/wiki/wikiUtils.ts
@@ -19,6 +19,73 @@ export const STATUS_LABELS: Record<string, string> = {
 
 const FALLBACK_FOLDER = "未分类";
 
+/**
+ * Display labels for known English directory segments (the wiki-capture
+ * categories that historically drove the vault path). Keyed by the lowercased
+ * segment. User-created folders are arbitrary (usually Chinese) and simply don't
+ * match, so they pass through unchanged.
+ */
+const FOLDER_LABELS: Record<string, string> = {
+  synthesis: "分析综合",
+  concepts: "概念框架",
+  concept: "概念框架",
+  references: "外部来源",
+  reference: "外部来源",
+  decision: "决策记录",
+  journal: "会话纪要",
+  session: "会话纪要",
+};
+
+/** Map a known English directory segment to its Chinese display label (else unchanged). */
+export function localizeFolderName(name: string): string {
+  return FOLDER_LABELS[name.trim().toLowerCase()] ?? name;
+}
+
+/**
+ * Canonical English wiki-capture section headings → Chinese display labels.
+ * Keyed by the lowercased, symbol-stripped heading text.
+ */
+const HEADING_LABELS: Record<string, string> = {
+  context: "背景",
+  "finding / decision": "结论 / 决策",
+  "finding/decision": "结论 / 决策",
+  finding: "结论",
+  decision: "决策",
+  reasoning: "推理依据",
+  implications: "影响 / 后续",
+  sources: "来源",
+  related: "相关知识",
+  merged: "已合并",
+};
+
+/**
+ * Replace the canonical English wiki-capture section headings (`## Context`,
+ * `## Finding / Decision`, …) with Chinese labels at render time. Operates on
+ * heading lines only (fenced code blocks are skipped) and matches the heading
+ * text case-insensitively, so **all historical notes are localized with no data
+ * migration**. Heading levels are preserved; because `extractToc` and the
+ * markdown viewer both receive this same transformed text, the slug/anchor ids
+ * stay consistent.
+ */
+export function localizeHeadings(markdown: string): string {
+  let inFence = false;
+  return markdown
+    .split("\n")
+    .map((line) => {
+      if (/^\s*(```|~~~)/.test(line)) {
+        inFence = !inFence;
+        return line;
+      }
+      if (inFence) return line;
+      const match = /^(#{1,4})\s+(.+?)\s*$/.exec(line);
+      if (!match) return line;
+      const key = match[2].replace(/[#`*_~]/g, "").trim().toLowerCase();
+      const zh = HEADING_LABELS[key];
+      return zh ? `${match[1]} ${zh}` : line;
+    })
+    .join("\n");
+}
+
 /** Stable slug for a heading: keeps CJK characters, collapses whitespace to dashes. */
 export function slugify(text: string): string {
   return text
@@ -119,16 +186,17 @@ export function splitRelatedSection(markdown: string): {
   return { body, related };
 }
 
-/** Folder path segments for a note: vault_path directories, else its source-type group. */
+/**
+ * Folder path segments for a note. The user-assigned ``folder`` (a slash-joined
+ * path) is authoritative; unfiled notes fall back to a source-type group so the
+ * tree is never empty. (The old behavior keyed off ``vault_path`` — the fixed
+ * ``projects/<project>/<category>`` scaffold — which the editable directory now
+ * replaces.)
+ */
 export function noteFolderSegments(note: KnowledgeNote): string[] {
-  const rawPath = note.vault_path?.replace(/\\/g, "/").replace(/^\/+/, "");
-  if (rawPath && rawPath.includes("/")) {
-    let segments = rawPath.split("/").slice(0, -1).filter(Boolean);
-    // Vault notes live under projects/<project>/<category>/ — the first two
-    // levels are constant scaffolding, so hide them from the directory tree.
-    if (segments[0] === "projects" || segments[0] === "project") {
-      segments = segments.slice(2);
-    }
+  const folder = (note.folder || "").replace(/^\/+|\/+$/g, "").trim();
+  if (folder) {
+    const segments = folder.split("/").map((s) => s.trim()).filter(Boolean);
     if (segments.length > 0) return segments;
   }
   return [SOURCE_TYPE_LABELS[note.source_type] ?? FALLBACK_FOLDER];
@@ -136,7 +204,7 @@ export function noteFolderSegments(note: KnowledgeNote): string[] {
 
 /** Breadcrumb segments for a note (folders only; the title is appended by the caller). */
 export function noteBreadcrumbSegments(note: KnowledgeNote): string[] {
-  return ["问答梳理", ...noteFolderSegments(note)];
+  return ["问答梳理", ...noteFolderSegments(note).map(localizeFolderName)];
 }
 
 function sortTree(nodes: TreeNode[]): void {
@@ -153,7 +221,10 @@ function sortTree(nodes: TreeNode[]): void {
  * Build the directory tree from the flat note list. Notes with a vault_path
  * become nested folders mirroring that path; others are grouped by source type.
  */
-export function buildKnowledgeTree(notes: KnowledgeNote[]): TreeNode[] {
+export function buildKnowledgeTree(
+  notes: KnowledgeNote[],
+  folderPaths: string[] = [],
+): TreeNode[] {
   const root: TreeNode[] = [];
   const folderMap = new Map<string, TreeNode>();
 
@@ -164,7 +235,9 @@ export function buildKnowledgeTree(notes: KnowledgeNote[]): TreeNode[] {
       path = path ? `${path}/${segment}` : segment;
       let node = folderMap.get(path);
       if (!node) {
-        node = { id: `folder:${path}`, name: segment, type: "folder", children: [] };
+        // Display name is localized (known English category → 中文); the id /
+        // folderPath keep the raw segment so filtering/matching is unaffected.
+        node = { id: `folder:${path}`, name: localizeFolderName(segment), type: "folder", children: [], folderPath: path };
         folderMap.set(path, node);
         children.push(node);
       }
@@ -173,6 +246,13 @@ export function buildKnowledgeTree(notes: KnowledgeNote[]): TreeNode[] {
     return children;
   };
 
+  // Pre-create managed (possibly empty) directories so a freshly-made folder
+  // shows up in the tree before any note is filed into it.
+  for (const path of folderPaths) {
+    const segments = path.split("/").map((s) => s.trim()).filter(Boolean);
+    if (segments.length) ensureFolder(segments);
+  }
+
   for (const note of notes) {
     ensureFolder(noteFolderSegments(note)).push({
       id: note.id,
diff --git a/frontend-web/src/components/workspace/admin/position-management-page.tsx b/frontend-web/src/components/workspace/admin/position-management-page.tsx
new file mode 100644
index 00000000..a747510c
--- /dev/null
+++ b/frontend-web/src/components/workspace/admin/position-management-page.tsx
@@ -0,0 +1,479 @@
+"use client";
+
+import {
+  CheckIcon,
+  Loader2,
+  PencilIcon,
+  PlusIcon,
+  RefreshCwIcon,
+  Trash2Icon,
+  UserPlusIcon,
+  UsersIcon,
+} from "lucide-react";
+import { useEffect, useMemo, useState } from "react";
+import { toast } from "sonner";
+
+import { Button } from "@/components/ui/button";
+import {
+  Dialog,
+  DialogContent,
+  DialogFooter,
+  DialogHeader,
+  DialogTitle,
+} from "@/components/ui/dialog";
+import { Input } from "@/components/ui/input";
+import { Textarea } from "@/components/ui/textarea";
+import { resolveTagColor } from "@/components/workspace/tags/tag-colors";
+import {
+  POSITION_RESOURCE_HINT,
+  POSITION_RESOURCE_LABELS,
+  POSITION_RESOURCE_TYPES,
+  resourceTypeToTagType,
+  useAddMembers,
+  useCreatePosition,
+  useDeletePosition,
+  usePositionMembers,
+  usePositionTagBindings,
+  usePositionUsers,
+  usePositions,
+  useRemoveMembers,
+  useResyncPosition,
+  useSetTagBindings,
+  useUpdatePosition,
+  type Position,
+  type PositionResourceType,
+  type TagBindings,
+} from "@/core/positions";
+import { useTags } from "@/core/tags";
+import { cn } from "@/lib/utils";
+
+function errMsg(e: unknown): string {
+  return String(e instanceof Error ? e.message : e);
+}
+
+/** Inline multi-select chip row for the tags of one resource type. */
+function TagMultiSelect({
+  resourceType,
+  value,
+  onChange,
+}: {
+  resourceType: PositionResourceType;
+  value: string[];
+  onChange: (next: string[]) => void;
+}) {
+  const { data: tags, isLoading } = useTags(resourceTypeToTagType(resourceType));
+  const toggle = (id: string) =>
+    onChange(value.includes(id) ? value.filter((x) => x !== id) : [...value, id]);
+
+  if (isLoading) {
+    return <Loader2 className="text-muted-foreground size-4 animate-spin" />;
+  }
+  if (!tags || tags.length === 0) {
+    return (
+      <span className="text-muted-foreground text-xs">
+        暂无{POSITION_RESOURCE_LABELS[resourceType]}标签，请先到「标签管理」创建。
+      </span>
+    );
+  }
+  return (
+    <div className="flex flex-wrap items-center gap-1.5">
+      {tags.map((tag) => {
+        const checked = value.includes(tag.id);
+        const color = resolveTagColor(tag.color);
+        return (
+          <button
+            key={tag.id}
+            type="button"
+            onClick={() => toggle(tag.id)}
+            className={cn(
+              "inline-flex items-center gap-1 rounded-full border px-3 py-1 text-xs font-medium transition-all",
+              color.chip,
+              checked ? "ring-foreground/40 ring-2 ring-offset-1" : "opacity-45 hover:opacity-90",
+            )}
+          >
+            {checked && <CheckIcon className="size-3" />}
+            {tag.name}
+          </button>
+        );
+      })}
+    </div>
+  );
+}
+
+function PositionFormDialog({
+  open,
+  onOpenChange,
+  editing,
+}: {
+  open: boolean;
+  onOpenChange: (open: boolean) => void;
+  editing: Position | null;
+}) {
+  const createPosition = useCreatePosition();
+  const updatePosition = useUpdatePosition();
+  const [name, setName] = useState("");
+  const [description, setDescription] = useState("");
+  const [isDefault, setIsDefault] = useState(false);
+
+  const [syncedFor, setSyncedFor] = useState<string | null>(null);
+  const syncKey = `${open}:${editing?.id ?? "new"}`;
+  if (open && syncedFor !== syncKey) {
+    setName(editing?.name ?? "");
+    setDescription(editing?.description ?? "");
+    setIsDefault(editing?.is_default ?? false);
+    setSyncedFor(syncKey);
+  }
+  if (!open && syncedFor !== null) setSyncedFor(null);
+
+  const pending = createPosition.isPending || updatePosition.isPending;
+  const handleSave = async () => {
+    const trimmed = name.trim();
+    if (!trimmed) return;
+    try {
+      if (editing) {
+        await updatePosition.mutateAsync({ id: editing.id, input: { name: trimmed, description, is_default: isDefault } });
+      } else {
+        await createPosition.mutateAsync({ name: trimmed, description, is_default: isDefault });
+      }
+      onOpenChange(false);
+    } catch (e) {
+      toast.error(errMsg(e));
+    }
+  };
+
+  return (
+    <Dialog open={open} onOpenChange={onOpenChange}>
+      <DialogContent>
+        <DialogHeader>
+          <DialogTitle>{editing ? "编辑岗位" : "新增岗位"}</DialogTitle>
+        </DialogHeader>
+        <div className="flex flex-col gap-4">
+          <div className="flex flex-col gap-1.5">
+            <label className="text-sm font-medium">岗位名称</label>
+            <Input value={name} onChange={(e) => setName(e.target.value)} placeholder="如：理赔专员" maxLength={191} />
+          </div>
+          <div className="flex flex-col gap-1.5">
+            <label className="text-sm font-medium">描述</label>
+            <Textarea value={description} onChange={(e) => setDescription(e.target.value)} rows={2} maxLength={1024} />
+          </div>
+          <label className="flex cursor-pointer items-center gap-2 text-sm">
+            <input type="checkbox" checked={isDefault} onChange={(e) => setIsDefault(e.target.checked)} />
+            设为默认岗位（未分配岗位的用户继承此岗位的配置）
+          </label>
+        </div>
+        <DialogFooter>
+          <Button variant="outline" onClick={() => onOpenChange(false)}>取消</Button>
+          <Button onClick={handleSave} disabled={pending || !name.trim()}>
+            {pending && <Loader2 className="size-4 animate-spin" />}保存
+          </Button>
+        </DialogFooter>
+      </DialogContent>
+    </Dialog>
+  );
+}
+
+function AddMembersDialog({
+  open,
+  onOpenChange,
+  positionId,
+}: {
+  open: boolean;
+  onOpenChange: (open: boolean) => void;
+  positionId: string;
+}) {
+  const { data: users, isLoading } = usePositionUsers();
+  const addMembers = useAddMembers();
+  const [selected, setSelected] = useState<Set<string>>(new Set());
+  const [search, setSearch] = useState("");
+
+  useEffect(() => {
+    if (open) setSelected(new Set());
+  }, [open]);
+
+  const filtered = useMemo(() => {
+    const list = users ?? [];
+    const q = search.trim().toLowerCase();
+    return q ? list.filter((u) => u.email.toLowerCase().includes(q)) : list;
+  }, [users, search]);
+
+  const toggle = (id: string) => {
+    setSelected((prev) => {
+      const next = new Set(prev);
+      if (next.has(id)) next.delete(id);
+      else next.add(id);
+      return next;
+    });
+  };
+
+  const handleSave = async () => {
+    if (selected.size === 0) return;
+    try {
+      await addMembers.mutateAsync({ positionId, userIds: [...selected] });
+      toast.success(`已为 ${selected.size} 名用户配置岗位`);
+      onOpenChange(false);
+    } catch (e) {
+      toast.error(errMsg(e));
+    }
+  };
+
+  return (
+    <Dialog open={open} onOpenChange={onOpenChange}>
+      <DialogContent className="max-w-lg">
+        <DialogHeader>
+          <DialogTitle>批量添加成员</DialogTitle>
+        </DialogHeader>
+        <Input value={search} onChange={(e) => setSearch(e.target.value)} placeholder="搜索邮箱…" />
+        <div className="max-h-72 overflow-y-auto rounded-md border">
+          {isLoading && (
+            <div className="text-muted-foreground flex items-center justify-center gap-2 py-6">
+              <Loader2 className="size-4 animate-spin" />加载用户…
+            </div>
+          )}
+          {!isLoading && filtered.length === 0 && (
+            <div className="text-muted-foreground py-6 text-center text-sm">无匹配用户</div>
+          )}
+          {filtered.map((u) => {
+            const checked = selected.has(u.id);
+            return (
+              <button
+                key={u.id}
+                type="button"
+                onClick={() => toggle(u.id)}
+                className={cn(
+                  "flex w-full items-center justify-between gap-2 border-b px-3 py-2 text-left text-sm last:border-b-0 hover:bg-muted/50",
+                  checked && "bg-muted/40",
+                )}
+              >
+                <span className="truncate">
+                  {u.email}
+                  {u.system_role === "admin" && <span className="text-muted-foreground ml-2 text-xs">(管理员)</span>}
+                </span>
+                <span className="flex items-center gap-2">
+                  {u.position_id && u.position_id !== positionId && (
+                    <span className="text-muted-foreground text-xs">已属其他岗位</span>
+                  )}
+                  <span className={cn("flex size-4 items-center justify-center rounded border", checked ? "bg-foreground text-background border-foreground" : "border-muted-foreground/40")}>
+                    {checked && <CheckIcon className="size-3" />}
+                  </span>
+                </span>
+              </button>
+            );
+          })}
+        </div>
+        <DialogFooter>
+          <Button variant="outline" onClick={() => onOpenChange(false)}>取消</Button>
+          <Button onClick={handleSave} disabled={addMembers.isPending || selected.size === 0}>
+            {addMembers.isPending && <Loader2 className="size-4 animate-spin" />}添加 {selected.size > 0 ? `(${selected.size})` : ""}
+          </Button>
+        </DialogFooter>
+      </DialogContent>
+    </Dialog>
+  );
+}
+
+function PositionDetail({ position }: { position: Position }) {
+  const { data: serverBindings } = usePositionTagBindings(position.id);
+  const { data: members } = usePositionMembers(position.id);
+  const setBindings = useSetTagBindings();
+  const removeMembers = useRemoveMembers();
+  const resync = useResyncPosition();
+  const [draft, setDraft] = useState<TagBindings>({});
+  const [addOpen, setAddOpen] = useState(false);
+
+  // Reload the draft whenever the selected position or its server bindings change.
+  const [loadedFor, setLoadedFor] = useState<string | null>(null);
+  const loadKey = `${position.id}:${serverBindings ? "ready" : "loading"}`;
+  if (serverBindings && loadedFor !== loadKey) {
+    setDraft({ ...serverBindings });
+    setLoadedFor(loadKey);
+  }
+
+  const setType = (rt: PositionResourceType, ids: string[]) =>
+    setDraft((prev) => ({ ...prev, [rt]: ids }));
+
+  const handleSaveBindings = async () => {
+    try {
+      await setBindings.mutateAsync({ positionId: position.id, bindings: draft });
+      toast.success("标签配置已保存，已同步岗位成员");
+    } catch (e) {
+      toast.error(errMsg(e));
+    }
+  };
+
+  const handleResync = async () => {
+    try {
+      const n = await resync.mutateAsync(position.id);
+      toast.success(`已重新同步 ${n} 名成员`);
+    } catch (e) {
+      toast.error(errMsg(e));
+    }
+  };
+
+  const handleRemove = async (userId: string) => {
+    try {
+      await removeMembers.mutateAsync({ positionId: position.id, userIds: [userId] });
+    } catch (e) {
+      toast.error(errMsg(e));
+    }
+  };
+
+  return (
+    <div className="flex flex-col gap-6">
+      <div className="rounded-lg border p-4">
+        <div className="mb-3 flex items-center justify-between">
+          <h3 className="text-sm font-semibold">默认资源标签配置</h3>
+          <Button size="sm" onClick={handleSaveBindings} disabled={setBindings.isPending}>
+            {setBindings.isPending && <Loader2 className="size-4 animate-spin" />}保存标签配置
+          </Button>
+        </div>
+        <div className="flex flex-col divide-y">
+          {POSITION_RESOURCE_TYPES.map((rt) => (
+            <div key={rt} className="flex flex-col gap-2 py-3 first:pt-0 last:pb-0">
+              <div className="flex items-baseline gap-2">
+                <span className="text-sm font-medium">{POSITION_RESOURCE_LABELS[rt]}</span>
+                <span className="text-muted-foreground text-xs">{POSITION_RESOURCE_HINT[rt]}</span>
+              </div>
+              <TagMultiSelect resourceType={rt} value={draft[rt] ?? []} onChange={(ids) => setType(rt, ids)} />
+            </div>
+          ))}
+        </div>
+      </div>
+
+      <div className="rounded-lg border p-4">
+        <div className="mb-3 flex items-center justify-between">
+          <h3 className="flex items-center gap-1.5 text-sm font-semibold">
+            <UsersIcon className="size-4" />岗位成员 ({members?.length ?? 0})
+          </h3>
+          <div className="flex gap-2">
+            <Button size="sm" variant="outline" onClick={handleResync} disabled={resync.isPending}>
+              {resync.isPending ? <Loader2 className="size-4 animate-spin" /> : <RefreshCwIcon className="size-4" />}重新同步
+            </Button>
+            <Button size="sm" onClick={() => setAddOpen(true)}>
+              <UserPlusIcon className="size-4" />添加成员
+            </Button>
+          </div>
+        </div>
+        {members && members.length > 0 ? (
+          <div className="flex flex-col divide-y">
+            {members.map((m) => (
+              <div key={m.id} className="flex items-center justify-between py-2 text-sm">
+                <span className="truncate">{m.email}</span>
+                <Button variant="ghost" size="icon" onClick={() => handleRemove(m.id)} aria-label="移除">
+                  <Trash2Icon className="text-destructive size-4" />
+                </Button>
+              </div>
+            ))}
+          </div>
+        ) : (
+          <div className="text-muted-foreground py-6 text-center text-sm">暂无成员</div>
+        )}
+      </div>
+
+      <AddMembersDialog open={addOpen} onOpenChange={setAddOpen} positionId={position.id} />
+    </div>
+  );
+}
+
+export function PositionManagementPage() {
+  const { data: positions, isLoading, error } = usePositions();
+  const deletePosition = useDeletePosition();
+  const [selectedId, setSelectedId] = useState<string | null>(null);
+  const [dialogOpen, setDialogOpen] = useState(false);
+  const [editing, setEditing] = useState<Position | null>(null);
+
+  const selected = useMemo(
+    () => positions?.find((p) => p.id === selectedId) ?? null,
+    [positions, selectedId],
+  );
+  // Auto-select the first position once loaded.
+  useEffect(() => {
+    if (!selectedId && positions && positions.length > 0) setSelectedId(positions[0].id);
+  }, [positions, selectedId]);
+
+  const handleDelete = async (p: Position) => {
+    if (!window.confirm(`确认删除岗位「${p.name}」？其成员将回退到默认岗位。`)) return;
+    try {
+      await deletePosition.mutateAsync(p.id);
+      if (selectedId === p.id) setSelectedId(null);
+    } catch (e) {
+      toast.error(errMsg(e));
+    }
+  };
+
+  return (
+    <div className="flex flex-col gap-4 p-6">
+      <div className="flex items-center justify-between">
+        <div>
+          <h1 className="text-lg font-semibold">岗位管理</h1>
+          <p className="text-muted-foreground text-sm">
+            为岗位配置默认智能体/技能/定时任务/推荐问题的标签，并批量分配给用户。
+          </p>
+        </div>
+        <Button size="sm" onClick={() => { setEditing(null); setDialogOpen(true); }}>
+          <PlusIcon className="size-4" />新增岗位
+        </Button>
+      </div>
+
+      {isLoading && (
+        <div className="text-muted-foreground flex items-center justify-center gap-2 py-12">
+          <Loader2 className="size-4 animate-spin" />加载中…
+        </div>
+      )}
+      {error && <div className="text-destructive text-sm">{errMsg(error)}</div>}
+
+      {!isLoading && !error && (
+        <div className="grid grid-cols-1 gap-6 lg:grid-cols-[280px_1fr]">
+          <div className="flex flex-col gap-2">
+            {positions && positions.length === 0 && (
+              <div className="text-muted-foreground rounded-lg border border-dashed py-8 text-center text-sm">
+                还没有岗位，点击「新增岗位」创建。
+              </div>
+            )}
+            {positions?.map((p) => (
+              <div
+                key={p.id}
+                onClick={() => setSelectedId(p.id)}
+                className={cn(
+                  "group cursor-pointer rounded-lg border p-3 transition-colors",
+                  selectedId === p.id ? "border-foreground bg-muted/40" : "hover:bg-muted/30",
+                )}
+              >
+                <div className="flex items-center justify-between">
+                  <div className="flex items-center gap-2">
+                    <span className="text-sm font-medium">{p.name}</span>
+                    {p.is_default && (
+                      <span className="rounded bg-blue-100 px-1.5 py-0.5 text-[10px] text-blue-700 dark:bg-blue-950 dark:text-blue-300">默认</span>
+                    )}
+                  </div>
+                  <div className="flex opacity-0 transition-opacity group-hover:opacity-100">
+                    <Button variant="ghost" size="icon" onClick={(e) => { e.stopPropagation(); setEditing(p); setDialogOpen(true); }} aria-label="编辑">
+                      <PencilIcon className="size-3.5" />
+                    </Button>
+                    <Button variant="ghost" size="icon" onClick={(e) => { e.stopPropagation(); void handleDelete(p); }} aria-label="删除">
+                      <Trash2Icon className="text-destructive size-3.5" />
+                    </Button>
+                  </div>
+                </div>
+                <div className="text-muted-foreground mt-1 flex items-center gap-2 text-xs">
+                  <UsersIcon className="size-3" />{p.member_count} 名成员
+                </div>
+              </div>
+            ))}
+          </div>
+
+          <div>
+            {selected ? (
+              <PositionDetail key={selected.id} position={selected} />
+            ) : (
+              <div className="text-muted-foreground rounded-lg border border-dashed py-16 text-center text-sm">
+                选择左侧一个岗位以配置标签与成员
+              </div>
+            )}
+          </div>
+        </div>
+      )}
+
+      <PositionFormDialog open={dialogOpen} onOpenChange={setDialogOpen} editing={editing} />
+    </div>
+  );
+}
diff --git a/frontend-web/src/components/workspace/admin/tag-management-page.tsx b/frontend-web/src/components/workspace/admin/tag-management-page.tsx
index d0484985..57bc6a2a 100644
--- a/frontend-web/src/components/workspace/admin/tag-management-page.tsx
+++ b/frontend-web/src/components/workspace/admin/tag-management-page.tsx
@@ -27,6 +27,7 @@ import {
   type Tag,
   type TagType,
 } from "@/core/tags";
+import { TAG_TYPES, TAG_TYPE_LABELS } from "@/core/tags/types";
 import { cn } from "@/lib/utils";
 
 function TagFormDialog({
@@ -191,11 +192,11 @@ export function TagManagementPage() {
             onValueChange={(v) => setActiveType(v as TagType)}
           >
             <TabsList variant="line">
-              <TabsTrigger value="agent">{t.adminTags.typeAgent}</TabsTrigger>
-              <TabsTrigger value="skill">{t.adminTags.typeSkill}</TabsTrigger>
-              <TabsTrigger value="scheduled_task">
-                {t.adminTags.typeScheduledTask}
-              </TabsTrigger>
+              {TAG_TYPES.map((tt) => (
+                <TabsTrigger key={tt} value={tt}>
+                  {TAG_TYPE_LABELS[tt]}
+                </TabsTrigger>
+              ))}
             </TabsList>
           </Tabs>
           <Button size="sm" onClick={openCreate}>
diff --git a/frontend-web/src/components/workspace/agents/agent-card.tsx b/frontend-web/src/components/workspace/agents/agent-card.tsx
index 21c710f4..cfd7e8c2 100644
--- a/frontend-web/src/components/workspace/agents/agent-card.tsx
+++ b/frontend-web/src/components/workspace/agents/agent-card.tsx
@@ -2,6 +2,7 @@
 
 import {
   BotIcon,
+  ClockIcon,
   CopyIcon,
   EyeOffIcon,
   GlobeIcon,
@@ -78,6 +79,7 @@ import type { Agent, AgentQuestion, UpdateAgentRequest } from "@/core/agents";
 import { getStoredAuth } from "@/core/auth";
 import { useI18n } from "@/core/i18n/hooks";
 import { useSkills } from "@/core/skills/hooks";
+import { formatDateTime } from "@/core/utils/datetime";
 import { desensitize } from "@/lib/desensitize";
 import { cn } from "@/lib/utils";
 
@@ -605,6 +607,17 @@ export function AgentCard({ agent, layout = "card", onPublisherClick }: AgentCar
       </button>
     ) : null;
 
+  const createdAtText = formatDateTime(agent.created_at);
+  const createdAtNode = createdAtText ? (
+    <div
+      className="text-muted-foreground flex max-w-full items-center gap-1 text-xs"
+      title={`创建时间：${createdAtText}`}
+    >
+      <ClockIcon className="size-3.5 shrink-0" />
+      <span className="truncate">创建时间：{createdAtText}</span>
+    </div>
+  ) : null;
+
   // 技能英文名 → 中文名映射,用于把智能体的技能徽章显示为中文(agent.skills 只存英文名)。
   const skillNameZhMap = useMemo(() => {
     const map = new Map<string, string>();
@@ -820,6 +833,7 @@ export function AgentCard({ agent, layout = "card", onPublisherClick }: AgentCar
           </CardDescription>
         )}
         {publisherNode && <div className="mt-2">{publisherNode}</div>}
+        {createdAtNode && <div className="mt-1">{createdAtNode}</div>}
       </CardHeader>
 
       {(agent.tags?.length ?? 0) > 0 && (
@@ -866,6 +880,7 @@ export function AgentCard({ agent, layout = "card", onPublisherClick }: AgentCar
           </p>
         )}
         {publisherNode && <div className="mt-0.5">{publisherNode}</div>}
+        {createdAtNode && <div className="mt-0.5">{createdAtNode}</div>}
       </div>
       <div className="flex shrink-0 items-center gap-1">
         {showActionRow && actionButtons}
diff --git a/frontend-web/src/components/workspace/agents/agent-detail-panel.tsx b/frontend-web/src/components/workspace/agents/agent-detail-panel.tsx
index cedeb6c8..f9a3292e 100644
--- a/frontend-web/src/components/workspace/agents/agent-detail-panel.tsx
+++ b/frontend-web/src/components/workspace/agents/agent-detail-panel.tsx
@@ -187,23 +187,30 @@ export function AgentDetailPanel({
   const isSaving = updateAgent.isPending || setFeatured.isPending;
 
   // 对话生成文件（md/txt 等 artifact）并在右侧展开预览时，自动收起智能体
-  // 信息面板给预览让位；新文件落地时同样收起。编辑中不打断用户，用户随时
+  // 信息面板给预览让位；切换到新文件时同样收起。编辑中不打断用户，用户随时
   // 可手动重新展开。
+  //
+  // 关键：判定用 open + selectedArtifact，而非 artifacts.length。流式写文件时
+  // message-group 一开始就调用 select()+setOpen(true) 打开预览并选中该文件，但
+  // 该文件要等生成结束、artifact 入库后才会进 thread.values.artifacts。若按
+  // 数量判定就只能等生成结束才收起——正是之前「生成完才收起」的 bug 根因。
   const artifactsCtx = useArtifactsOptional();
   const artifactCount = artifactsCtx?.artifacts.length ?? 0;
-  const artifactPanelVisible = !!artifactsCtx?.open && artifactCount > 0;
+  const selectedArtifact = artifactsCtx?.selectedArtifact ?? null;
+  const artifactPanelVisible =
+    !!artifactsCtx?.open && (!!selectedArtifact || artifactCount > 0);
   const prevArtifactPanelVisible = useRef(false);
-  const prevArtifactCount = useRef(0);
+  const prevSelectedArtifact = useRef<string | null>(null);
   useEffect(() => {
     const opened = artifactPanelVisible && !prevArtifactPanelVisible.current;
-    const grew =
-      artifactPanelVisible && artifactCount > prevArtifactCount.current;
+    const switched =
+      artifactPanelVisible && selectedArtifact !== prevSelectedArtifact.current;
     prevArtifactPanelVisible.current = artifactPanelVisible;
-    prevArtifactCount.current = artifactCount;
-    if ((opened || grew) && !editing) {
+    prevSelectedArtifact.current = selectedArtifact;
+    if ((opened || switched) && !editing) {
       setCollapsed(true);
     }
-  }, [artifactPanelVisible, artifactCount, editing]);
+  }, [artifactPanelVisible, selectedArtifact, editing]);
 
   const displayName = desensitize(agent?.name ?? agentName);
   const questions = useMemo(
diff --git a/frontend-web/src/components/workspace/agents/agent-module-settings-dialog.tsx b/frontend-web/src/components/workspace/agents/agent-module-settings-dialog.tsx
new file mode 100644
index 00000000..b1e71d7e
--- /dev/null
+++ b/frontend-web/src/components/workspace/agents/agent-module-settings-dialog.tsx
@@ -0,0 +1,174 @@
+"use client";
+
+import { Bot, Loader2 } from "lucide-react";
+import { useEffect, useMemo, useRef, useState } from "react";
+import { toast } from "sonner";
+import { Select } from "tdesign-react";
+
+import { Button } from "@/components/ui/button";
+import {
+  Dialog,
+  DialogContent,
+  DialogDescription,
+  DialogFooter,
+  DialogHeader,
+  DialogTitle,
+} from "@/components/ui/dialog";
+import { ScrollArea } from "@/components/ui/scroll-area";
+import { useAgents } from "@/core/agents";
+import type { Agent } from "@/core/agents";
+import {
+  useAgentModuleConfig,
+  useUpdateAgentModuleConfig,
+} from "@/core/system-settings/hooks";
+import { useTags } from "@/core/tags";
+import { desensitize } from "@/lib/desensitize";
+
+/**
+ * 模块（zzJC / zzXD）类型配置弹窗（需求 0616 #3/#4，admin）。
+ *
+ * 单选一个「类型」（agent 标签，TDesign 下拉框）→ 下方实时预览该类型下的智能体（名称+描述）→
+ * 确认后写入后端（``/api/system-settings/agent-modules/{key}``，单选即存单元素列表），
+ * 该模块左栏据此只展示选中类型的智能体。
+ */
+export function AgentModuleSettingsDialog({
+  open,
+  onOpenChange,
+  moduleKey,
+  moduleTitle,
+}: {
+  open: boolean;
+  onOpenChange: (open: boolean) => void;
+  moduleKey: string;
+  moduleTitle: string;
+}) {
+  const { data: tags = [] } = useTags("agent");
+  const { agents } = useAgents();
+  const { data: config } = useAgentModuleConfig(moduleKey);
+  const { mutate: save, isPending } = useUpdateAgentModuleConfig(moduleKey);
+
+  const [selectedTagId, setSelectedTagId] = useState<string | undefined>();
+  // Radix Dialog(modal) 把 body 设为 pointer-events:none，TDesign 浮层默认挂到 body 上
+  // 会点不动 —— 用 attach 把下拉浮层挂到弹窗内部容器（pointer-events 正常）。
+  const selectWrapRef = useRef<HTMLDivElement>(null);
+
+  // 每次打开同步当前已保存配置（取首个，单选语义）。
+  useEffect(() => {
+    if (open) setSelectedTagId(config?.tag_ids?.[0]);
+  }, [open, config?.tag_ids]);
+
+  const tagOptions = useMemo(
+    () => tags.map((t) => ({ label: t.name, value: t.id })),
+    [tags],
+  );
+
+  // 选中类型下的智能体（含名称 + 描述）。
+  const previewAgents = useMemo<Agent[]>(() => {
+    if (!selectedTagId) return [];
+    return agents.filter((agent) =>
+      (agent.tags ?? []).some((t) => t.id === selectedTagId),
+    );
+  }, [agents, selectedTagId]);
+
+  const handleConfirm = () => {
+    save(selectedTagId ? [selectedTagId] : [], {
+      onSuccess: () => {
+        toast.success(`已更新「${moduleTitle}」展示的类型`);
+        onOpenChange(false);
+      },
+      onError: (err) =>
+        toast.error(err instanceof Error ? err.message : "保存失败，请重试"),
+    });
+  };
+
+  return (
+    <Dialog open={open} onOpenChange={onOpenChange}>
+      <DialogContent className="flex h-[78vh] max-h-[88vh] min-h-[620px] flex-col gap-0 p-0 sm:max-w-[760px]">
+        <DialogHeader className="border-b border-border/60 px-6 py-4">
+          <DialogTitle>{moduleTitle} · 类型配置</DialogTitle>
+          <DialogDescription>
+            选择该模块左栏展示的智能体类型（标签）；下方预览该类型下的智能体。
+          </DialogDescription>
+        </DialogHeader>
+
+        <div className="flex min-h-0 flex-1 flex-col gap-5 overflow-hidden px-6 py-5">
+          <div className="shrink-0">
+            <div className="text-muted-foreground/80 mb-2 text-sm font-semibold">
+              选择类型
+            </div>
+            {tags.length === 0 ? (
+              <p className="text-muted-foreground py-1 text-sm">
+                暂无智能体标签，请先到「标签管理」创建。
+              </p>
+            ) : (
+              <div ref={selectWrapRef} className="relative">
+                <Select
+                  className="w-full"
+                  value={selectedTagId}
+                  options={tagOptions}
+                  filterable
+                  clearable
+                  placeholder="搜索并选择一个类型"
+                  popupProps={{
+                    attach: () => selectWrapRef.current ?? document.body,
+                  }}
+                  onChange={(v) => setSelectedTagId((v as string) || undefined)}
+                />
+              </div>
+            )}
+          </div>
+
+          <div className="flex min-h-0 flex-1 flex-col">
+            <div className="text-muted-foreground/80 mb-2 shrink-0 text-sm font-semibold">
+              该类型下的智能体（{previewAgents.length}）
+            </div>
+            {!selectedTagId ? (
+              <p className="text-muted-foreground py-1 text-sm">
+                未选择类型时，左栏展示全部可见智能体。
+              </p>
+            ) : previewAgents.length === 0 ? (
+              <p className="text-muted-foreground py-1 text-sm">
+                该类型下暂无智能体。
+              </p>
+            ) : (
+              <ScrollArea className="min-h-0 flex-1 pr-3">
+                <div className="flex flex-col gap-2">
+                  {previewAgents.map((agent) => (
+                    <div
+                      key={agent.id}
+                      className="flex items-start gap-3 rounded-xl border border-border/60 bg-card p-3"
+                    >
+                      <div className="bg-primary/10 text-primary mt-0.5 flex size-8 shrink-0 items-center justify-center rounded-lg">
+                        <Bot className="size-4" />
+                      </div>
+                      <div className="min-w-0 flex-1">
+                        <div className="truncate text-sm font-semibold text-foreground">
+                          {desensitize(agent.name)}
+                        </div>
+                        <p className="text-muted-foreground mt-0.5 line-clamp-2 text-xs leading-relaxed">
+                          {agent.description?.trim()
+                            ? desensitize(agent.description)
+                            : "暂无描述"}
+                        </p>
+                      </div>
+                    </div>
+                  ))}
+                </div>
+              </ScrollArea>
+            )}
+          </div>
+        </div>
+
+        <DialogFooter className="border-t border-border/60 px-6 py-4">
+          <Button variant="outline" onClick={() => onOpenChange(false)}>
+            取消
+          </Button>
+          <Button onClick={handleConfirm} disabled={isPending}>
+            {isPending && <Loader2 className="size-4 animate-spin" />}
+            确认
+          </Button>
+        </DialogFooter>
+      </DialogContent>
+    </Dialog>
+  );
+}
diff --git a/frontend-web/src/components/workspace/agents/agent-module-sidebar.tsx b/frontend-web/src/components/workspace/agents/agent-module-sidebar.tsx
new file mode 100644
index 00000000..2266b48d
--- /dev/null
+++ b/frontend-web/src/components/workspace/agents/agent-module-sidebar.tsx
@@ -0,0 +1,278 @@
+"use client";
+
+import { Bot, ChevronDown, MessageSquare, Settings } from "lucide-react";
+import Link from "next/link";
+import { usePathname } from "next/navigation";
+import { useMemo, useState } from "react";
+
+import { useAgents } from "@/core/agents";
+import type { Agent } from "@/core/agents";
+import { useAgentModuleConfig } from "@/core/system-settings/hooks";
+import { useRouteBase } from "@/core/route-base";
+import { useThreads } from "@/core/threads/hooks";
+import type { AgentThread } from "@/core/threads/types";
+import {
+  pathOfThread,
+  sortThreadsPinnedFirst,
+  titleOfThread,
+} from "@/core/threads/utils";
+import type { Tag } from "@/core/tags";
+import { desensitize } from "@/lib/desensitize";
+import { cn } from "@/lib/utils";
+
+/** 单个「按类型聚合」模块配置（需求 0616 #3 zzJC / #4 zzXD）。 */
+export interface AgentModuleConfig {
+  /** 模块标识（路由 /page/strategy/qa/{key} 的 key）。 */
+  key: string;
+  /** 模块标题。 */
+  title: string;
+  /** 未配置类型时的默认落地智能体 id。 */
+  fallbackAgentId: string;
+}
+
+/**
+ * 已知「按类型聚合」模块 —— 按 **moduleKey** 索引（路由 /page/strategy/qa/{key}）。
+ * zzJC / zzXD 布局完全一致，仅默认落地智能体不同。
+ */
+export const AGENT_MODULES: Record<string, AgentModuleConfig> = {
+  zzjc: { key: "zzjc", title: "zzJC支持", fallbackAgentId: "6bf" },
+  zzxd: {
+    key: "zzxd",
+    title: "zzXD应用",
+    fallbackAgentId: "10d69686718349688e58a8790cb61024",
+  },
+};
+
+const UNTAGGED_KEY = "__untagged__";
+
+interface AgentGroup {
+  /** 分组键（tag id 或未分类常量）。 */
+  key: string;
+  /** 分组显示名。 */
+  label: string;
+  agents: Agent[];
+}
+
+/**
+ * 按 agent 标签把可见智能体分组（多标签智能体会出现在多个组）。
+ *
+ * - ``allowedTagIds`` 非空（admin 已配置该模块的类型）：仅保留这些标签、按其配置顺序，
+ *   空组（无智能体）不展示；不含「未分类」。
+ * - ``allowedTagIds`` 为空（未配置）：回退展示全部标签 + 「未分类」。
+ */
+function groupAgentsByTag(agents: Agent[], allowedTagIds: string[]): AgentGroup[] {
+  if (allowedTagIds.length > 0) {
+    const allowed = new Set(allowedTagIds);
+    const byTag = new Map<string, { tag: Tag; agents: Agent[] }>();
+    for (const agent of agents) {
+      for (const tag of agent.tags ?? []) {
+        if (!allowed.has(tag.id)) continue;
+        const entry = byTag.get(tag.id) ?? { tag, agents: [] };
+        entry.agents.push(agent);
+        byTag.set(tag.id, entry);
+      }
+    }
+    const groups: AgentGroup[] = [];
+    for (const tagId of allowedTagIds) {
+      const entry = byTag.get(tagId);
+      if (entry && entry.agents.length > 0) {
+        groups.push({ key: tagId, label: entry.tag.name, agents: entry.agents });
+      }
+    }
+    return groups;
+  }
+
+  const byTag = new Map<string, { tag: Tag; agents: Agent[] }>();
+  const untagged: Agent[] = [];
+  for (const agent of agents) {
+    const tags = agent.tags ?? [];
+    if (tags.length === 0) {
+      untagged.push(agent);
+      continue;
+    }
+    for (const tag of tags) {
+      const entry = byTag.get(tag.id) ?? { tag, agents: [] };
+      entry.agents.push(agent);
+      byTag.set(tag.id, entry);
+    }
+  }
+  const groups: AgentGroup[] = [...byTag.values()]
+    .map((e) => ({ key: e.tag.id, label: e.tag.name, agents: e.agents }))
+    .sort((a, b) => a.label.localeCompare(b.label, "zh"));
+  if (untagged.length > 0) {
+    groups.push({ key: UNTAGGED_KEY, label: "未分类", agents: untagged });
+  }
+  return groups;
+}
+
+/**
+ * 「按类型聚合智能体 + 聊天记录」左栏（需求 0616 #3 zzJC / #4 zzXD）。
+ *
+ * - 上半：可见智能体按标签（类型）分组、可折叠；点击进入该智能体对话（保持当前路由
+ *   命名空间，全量模式下为 /page/strategy）。
+ * - 下半：聊天记录 —— 仅与本模块智能体（即左栏列出的智能体）的会话，按 metadata.agent_id 过滤。
+ * - 宽度与通用问答左栏一致（``--sidebar-width``）。
+ */
+export function AgentModuleSidebar({
+  moduleKey,
+  activeAgentId,
+  onOpenSettings,
+}: {
+  moduleKey: string;
+  activeAgentId?: string;
+  /** 传入则在左栏顶部显示设置齿轮（admin）。 */
+  onOpenSettings?: () => void;
+}) {
+  const routeBase = useRouteBase();
+  const pathname = usePathname();
+  const { agents } = useAgents();
+  const { data: threads = [] } = useThreads();
+  const { data: moduleConfig } = useAgentModuleConfig(moduleKey);
+  const [collapsed, setCollapsed] = useState<Set<string>>(() => new Set());
+
+  const allowedTagIds = useMemo(
+    () => moduleConfig?.tag_ids ?? [],
+    [moduleConfig?.tag_ids],
+  );
+  const groups = useMemo(
+    () => groupAgentsByTag(agents, allowedTagIds),
+    [agents, allowedTagIds],
+  );
+  // 本模块智能体集合 = 左栏实际展示的智能体（聊天记录据此过滤，仅本模块相关会话）。
+  const moduleAgentIds = useMemo(
+    () => new Set(groups.flatMap((g) => g.agents.map((a) => a.id))),
+    [groups],
+  );
+
+  // 聊天记录：仅与本模块智能体的会话（按 metadata.agent_id 过滤），置顶优先。
+  const moduleThreads = useMemo(
+    () =>
+      sortThreadsPinnedFirst(
+        threads.filter((thread: AgentThread) => {
+          const agentId = thread.metadata?.agent_id;
+          return typeof agentId === "string" && moduleAgentIds.has(agentId);
+        }),
+      ),
+    [threads, moduleAgentIds],
+  );
+
+  const toggleGroup = (key: string) =>
+    setCollapsed((prev) => {
+      const next = new Set(prev);
+      if (next.has(key)) next.delete(key);
+      else next.add(key);
+      return next;
+    });
+
+  return (
+    <aside className="flex h-full w-(--sidebar-width) shrink-0 flex-col overflow-hidden border-r border-border/70 bg-card/60 backdrop-blur-sm">
+      {onOpenSettings && (
+        <div className="flex shrink-0 items-center justify-between border-b border-border/60 px-3 py-2">
+          <span className="text-xs font-semibold text-foreground/80">智能体</span>
+          <button
+            type="button"
+            onClick={onOpenSettings}
+            title="模块设置"
+            aria-label="模块设置"
+            className="text-muted-foreground hover:text-foreground rounded-md p-1 transition-colors hover:bg-sidebar-accent/60"
+          >
+            <Settings className="size-4" />
+          </button>
+        </div>
+      )}
+
+      {/* 上半：按类型分组的智能体 */}
+      <div className="flex min-h-0 flex-[3] flex-col overflow-y-auto px-2 py-3">
+        <h2 className="text-muted-foreground/70 px-2 mb-1.5 text-[10px] font-semibold tracking-[0.12em] uppercase select-none">
+          智能体类型
+        </h2>
+        {groups.length === 0 ? (
+          <p className="px-2 py-1.5 text-xs text-muted-foreground/70">暂无智能体</p>
+        ) : (
+          <div className="flex flex-col gap-0.5">
+            {groups.map((group) => {
+              const isCollapsed = collapsed.has(group.key);
+              return (
+                <div key={group.key}>
+                  <button
+                    type="button"
+                    onClick={() => toggleGroup(group.key)}
+                    className="flex w-full items-center justify-between gap-1 rounded-lg px-2 py-1.5 text-xs font-semibold text-foreground/80 hover:bg-sidebar-accent/60"
+                  >
+                    <span className="flex items-center gap-1.5 truncate">
+                      <span className="truncate">{group.label}</span>
+                      <span className="text-muted-foreground/60 text-[10px] font-normal tabular-nums">
+                        {group.agents.length}
+                      </span>
+                    </span>
+                    <ChevronDown
+                      className={cn(
+                        "size-3.5 shrink-0 transition-transform duration-200",
+                        isCollapsed && "-rotate-90",
+                      )}
+                    />
+                  </button>
+                  {!isCollapsed && (
+                    <div className="mt-0.5 flex flex-col gap-0.5 pl-1">
+                      {group.agents.map((agent) => {
+                        const href = `${routeBase}/agents/${agent.id}/chats/new`;
+                        const isActive = agent.id === activeAgentId;
+                        return (
+                          <Link
+                            key={`${group.key}:${agent.id}`}
+                            href={href}
+                            className={cn(
+                              "flex items-center gap-2 rounded-lg px-2.5 py-1.5 text-[13px] text-muted-foreground hover:text-foreground hover:bg-sidebar-accent/60",
+                              isActive &&
+                                "bg-primary/10 text-primary hover:bg-primary/15 hover:text-primary",
+                            )}
+                          >
+                            <Bot className="size-3.5 shrink-0 text-muted-foreground/70" />
+                            <span className="truncate">{desensitize(agent.name)}</span>
+                          </Link>
+                        );
+                      })}
+                    </div>
+                  )}
+                </div>
+              );
+            })}
+          </div>
+        )}
+      </div>
+
+      {/* 下半：聊天记录（仅本模块智能体会话） */}
+      <div className="flex min-h-0 flex-[2] flex-col overflow-y-auto border-t border-border/60 px-2 py-3">
+        <h2 className="text-muted-foreground/70 px-2 mb-1.5 text-[10px] font-semibold tracking-[0.12em] uppercase select-none">
+          聊天记录
+        </h2>
+        {moduleThreads.length === 0 ? (
+          <p className="px-2 py-1.5 text-xs text-muted-foreground/70">暂无聊天记录</p>
+        ) : (
+          <div className="flex flex-col gap-0.5">
+            {moduleThreads.map((thread) => {
+              const href = pathOfThread(thread, undefined, routeBase);
+              const isActive = href === pathname;
+              return (
+                <Link
+                  key={thread.thread_id}
+                  href={href}
+                  className={cn(
+                    "flex items-center gap-2 rounded-lg px-2.5 py-1.5 text-[13px] text-muted-foreground hover:text-foreground hover:bg-sidebar-accent/60",
+                    isActive &&
+                      "bg-primary/10 text-primary hover:bg-primary/15 hover:text-primary",
+                  )}
+                >
+                  <MessageSquare className="size-3.5 shrink-0 text-muted-foreground/70" />
+                  <span className="truncate">
+                    {desensitize(titleOfThread(thread, "未命名对话"))}
+                  </span>
+                </Link>
+              );
+            })}
+          </div>
+        )}
+      </div>
+    </aside>
+  );
+}
diff --git a/frontend-web/src/components/workspace/agents/agent-selector.tsx b/frontend-web/src/components/workspace/agents/agent-selector.tsx
index a605ae7f..cf4d1f1a 100644
--- a/frontend-web/src/components/workspace/agents/agent-selector.tsx
+++ b/frontend-web/src/components/workspace/agents/agent-selector.tsx
@@ -20,9 +20,10 @@ import {
 } from "@/components/ui/dialog";
 import { PromptInputButton } from "@/components/ai-elements/prompt-input";
 import { AgentManageDialog } from "@/components/workspace/agents/agent-manage-dialog";
-import { useAgentSelector } from "@/core/agents";
+import { useAgentSelector, useAgentsPaginated } from "@/core/agents";
 import type { Agent } from "@/core/agents";
 import { getStoredAuth } from "@/core/auth";
+import { desensitize } from "@/lib/desensitize";
 import { cn } from "@/lib/utils";
 
 interface AgentSelectorProps {
@@ -51,12 +52,21 @@ export function AgentSelector({
   const [open, setOpen] = useState(false);
   const [manageOpen, setManageOpen] = useState(false);
   const { agents, isLoading } = useAgentSelector();
+  // 弹窗列表只展示管理弹窗里配置的 agent（agents）；但「已选中 agent」的名称解析要更宽——
+  // 侧边栏「常用智能体」可以选中未配置进管理弹窗的 agent（如用户自己的智能体），
+  // 这些 agent 也要能在胶囊上显示名称，故额外用「我的 / 广场」列表兜底解析（缓存与侧边栏共享）。
+  const { agents: mineAgents } = useAgentsPaginated({ scope: "mine", page: 1, pageSize: 20 });
+  const { agents: squareAgents } = useAgentsPaginated({ scope: "square", page: 1, pageSize: 20 });
   const isAdmin = useMemo(() => getStoredAuth()?.system_role === "admin", []);
 
   const selectedAgent = useMemo<Agent | null>(() => {
     if (!selectedAgentId) return null;
-    return agents.find((a) => a.id === selectedAgentId) ?? null;
-  }, [agents, selectedAgentId]);
+    return (
+      [...agents, ...mineAgents, ...squareAgents].find(
+        (a) => a.id === selectedAgentId,
+      ) ?? null
+    );
+  }, [agents, mineAgents, squareAgents, selectedAgentId]);
 
   const handleSelect = (agent: Agent) => {
     onAgentChange(agent.id);
@@ -68,7 +78,7 @@ export function AgentSelector({
     onAgentChange(null);
   };
 
-  const triggerLabel = selectedAgent?.name ?? "选择 Agent";
+  const triggerLabel = selectedAgent ? desensitize(selectedAgent.name) : "选择 Agent";
 
   return (
     <>
@@ -125,7 +135,7 @@ export function AgentSelector({
                   {agents.map((agent) => (
                     <CommandItem
                       key={agent.id}
-                      value={`${agent.name} ${agent.description ?? ""}`}
+                      value={`${desensitize(agent.name)} ${desensitize(agent.description ?? "")}`}
                       onSelect={() => handleSelect(agent)}
                       className={cn(
                         "flex w-full min-w-0 items-start gap-2 overflow-hidden",
@@ -134,10 +144,10 @@ export function AgentSelector({
                     >
                       <BotIcon className="text-muted-foreground mt-0.5 size-4 shrink-0" />
                       <div className="flex min-w-0 flex-1 flex-col">
-                        <span className="truncate text-sm">{agent.name}</span>
+                        <span className="truncate text-sm">{desensitize(agent.name)}</span>
                         {agent.description && (
                           <span className="line-clamp-2 whitespace-normal break-words text-[10px] leading-snug text-muted-foreground">
-                            {agent.description}
+                            {desensitize(agent.description)}
                           </span>
                         )}
                       </div>
diff --git a/frontend-web/src/components/workspace/artifacts/artifact-file-detail.tsx b/frontend-web/src/components/workspace/artifacts/artifact-file-detail.tsx
index 5188d8d5..37bdd9cf 100644
--- a/frontend-web/src/components/workspace/artifacts/artifact-file-detail.tsx
+++ b/frontend-web/src/components/workspace/artifacts/artifact-file-detail.tsx
@@ -122,6 +122,7 @@ export function ArtifactFileDetail({
   threadId,
   onClose,
   allowVirtualMarkdownEdit = false,
+  disableDesensitize = false,
 }: {
   className?: string;
   filepath: string;
@@ -134,6 +135,12 @@ export function ArtifactFileDetail({
    * 默认 false —— 主聊天页等其它调用方行为不变。
    */
   allowVirtualMarkdownEdit?: boolean;
+  /**
+   * 关闭对预览内容的脱敏。多智能体会商页的「消息列表沙箱」按需求**不脱敏**沙箱内的
+   * 产物内容（用户要拿原始文件），故 Step2SandboxLayout 传 true；主聊天等其它调用方
+   * 默认 false，沿用全局脱敏。
+   */
+  disableDesensitize?: boolean;
 }) {
   const { t } = useI18n();
   const { artifacts, setOpen, select } = useArtifacts();
@@ -802,7 +809,7 @@ export function ArtifactFileDetail({
             language === "html" ||
             language === "mermaid") && (
               <ArtifactFilePreview
-                content={desensitize(displayContent ?? "")}
+                content={disableDesensitize ? (displayContent ?? "") : desensitize(displayContent ?? "")}
                 language={language ?? "text"}
                 threadId={threadId}
                 publicShareCode={publicShareCode}
@@ -815,7 +822,7 @@ export function ArtifactFileDetail({
             <CodeEditor
               className="size-full resize-none rounded-none border-none bg-transparent"
               value={
-                canEditArtifact
+                canEditArtifact || disableDesensitize
                   ? (displayContent ?? "")
                   : desensitize(displayContent ?? "")
               }
diff --git a/frontend-web/src/components/workspace/general-qa-sidebar.tsx b/frontend-web/src/components/workspace/general-qa-sidebar.tsx
new file mode 100644
index 00000000..af4b7e4f
--- /dev/null
+++ b/frontend-web/src/components/workspace/general-qa-sidebar.tsx
@@ -0,0 +1,603 @@
+"use client";
+
+/**
+ * 通用问答（全量版 / full layout）新建对话左侧栏 —— 按「样式修改.png」参考图重画。
+ *
+ * 这是**全量版专属**的独立组件：因为全量版是新功能、新需求，若继续往
+ * 共享的 WorkspaceNavChatList / FrequentAgentsList / RecentChatList 里塞会越来越难
+ * 维护，故单独成一个组件，只在 `/page/workspace/chats*` 的全量布局下挂载，互不影响。
+ *
+ * 三块结构（自上而下）：
+ *   1. 「新对话」胶囊按钮 + 收起整列的圆形按钮；
+ *   2. 「常用智能体」—— 彩色图标卡（名称 + 两行简介，命中为选中态：主色描边 + 右侧主色条）；
+ *   3. 「最近的对话」—— 计数徽标 + 折叠、历史搜索框、会话列表（悬浮「更多」菜单：
+ *      重命名/置顶/收藏/分享/导出/删除）、「更多最近对话」入口，底部品牌信息。
+ */
+
+import {
+  Bot,
+  ChevronDown,
+  ChevronRight,
+  Download,
+  FileJson,
+  FileText,
+  MessageSquare,
+  MoreHorizontal,
+  Pencil,
+  Pin,
+  PinOff,
+  Plus,
+  Search,
+  Share2,
+  Star,
+  Trash2,
+} from "lucide-react";
+import Link from "next/link";
+import { useParams, usePathname, useRouter, useSearchParams } from "next/navigation";
+import { useCallback, useMemo, useState } from "react";
+import { useNavigate } from "react-router-dom";
+import { toast } from "sonner";
+
+import { useTheme } from "@/contexts/ThemeContext";
+import { Button } from "@/components/ui/button";
+import {
+  Dialog,
+  DialogContent,
+  DialogFooter,
+  DialogHeader,
+  DialogTitle,
+} from "@/components/ui/dialog";
+import {
+  DropdownMenu,
+  DropdownMenuContent,
+  DropdownMenuItem,
+  DropdownMenuSeparator,
+  DropdownMenuSub,
+  DropdownMenuSubContent,
+  DropdownMenuSubTrigger,
+  DropdownMenuTrigger,
+} from "@/components/ui/dropdown-menu";
+import { Input } from "@/components/ui/input";
+import { ShareConversationDialog } from "@/components/workspace/thread-share-dialogs";
+import { useAgentsPaginated, type Agent } from "@/core/agents";
+import { getAPIClient } from "@/core/api";
+import { useI18n } from "@/core/i18n/hooks";
+import {
+  exportThreadAsJSON,
+  exportThreadAsMarkdown,
+} from "@/core/threads/export";
+import {
+  useDeleteThread,
+  useRenameThread,
+  useThreads,
+  useToggleThreadFavorite,
+  useToggleThreadPin,
+} from "@/core/threads/hooks";
+import type { AgentThread, AgentThreadState } from "@/core/threads/types";
+import {
+  isThreadFavorite,
+  isThreadPinned,
+  pathOfThread,
+  sortThreadsPinnedFirst,
+  titleOfThread,
+} from "@/core/threads/utils";
+import { env } from "@/env";
+import { desensitize } from "@/lib/desensitize";
+import { isIMEComposing } from "@/lib/ime";
+import { cn } from "@/lib/utils";
+
+import { SidebarCollapseButton } from "./workspace-header";
+
+const NEW_CHAT_PATH = "/page/workspace/chats/new";
+/** 「更多」常用智能体跳转目标：智能体管理页。 */
+const MORE_AGENTS_HREF = "/page/workspace/agents";
+/** 「更多最近对话」跳转目标。 */
+const MORE_CHATS_HREF = "/page/workspace/chats";
+/** 常用智能体最多展示个数。 */
+const MAX_FREQUENT = 4;
+
+// 智能体图标统一用 Bot 机器人图标；底色按 id 稳定散列到一组主题色，复刻参考图的彩色方块。
+// 浅色模式沿用原高彩度彩色（不改）；深蓝主题改用《深蓝 UI 规范》§1.1.7/§1.1.8 的蓝-青系
+// （docs/ui规范-深蓝，克制、不刺眼），仅深蓝主题下生效，互不影响。
+const AGENT_COLORS_LIGHT = [
+  { from: "#fb7185", to: "#ef4444" },
+  { from: "#60a5fa", to: "#3b82f6" },
+  { from: "#fbbf24", to: "#f59e0b" },
+  { from: "#34d399", to: "#10b981" },
+  { from: "#a78bfa", to: "#8b5cf6" },
+] as const;
+
+const AGENT_COLORS_DARK = [
+  { from: "#4096FF", to: "#1677FF" }, // 蓝（链接蓝 → 主色）
+  { from: "#36CFC9", to: "#13C2C2" }, // 青（辅色）
+  { from: "#6388FF", to: "#0958D9" }, // 紫蓝（装饰紫 → 主色深一档）
+  { from: "#5CDBD3", to: "#08979C" }, // 浅青 → 深青
+  { from: "#69B1FF", to: "#003EB3" }, // 浅蓝 → 主色按下
+] as const;
+
+function agentColor(id: string, isDark: boolean) {
+  let hash = 0;
+  for (let i = 0; i < id.length; i++) {
+    hash = (hash * 31 + id.charCodeAt(i)) >>> 0;
+  }
+  const palette = isDark ? AGENT_COLORS_DARK : AGENT_COLORS_LIGHT;
+  return palette[hash % palette.length]!;
+}
+
+/** 折叠区块标题（带 chevron 与可选计数徽标）。 */
+function SectionLabel({
+  label,
+  count,
+  collapsed,
+  onToggle,
+}: {
+  label: string;
+  count?: number;
+  collapsed: boolean;
+  onToggle: () => void;
+}) {
+  return (
+    <button
+      type="button"
+      onClick={onToggle}
+      className="flex w-full items-center justify-between px-3 py-1 text-[11px] font-semibold tracking-[0.04em] text-muted-foreground/80 transition-colors hover:text-muted-foreground select-none"
+    >
+      <span className="flex items-center gap-1.5">
+        <span>{label}</span>
+        {count != null && (
+          <span className="rounded-full bg-muted px-1.5 py-0.5 text-[10px] font-normal text-muted-foreground tabular-nums">
+            {count}
+          </span>
+        )}
+      </span>
+      <ChevronDown
+        className={cn(
+          "size-3.5 shrink-0 transition-transform duration-200",
+          collapsed && "-rotate-90",
+        )}
+      />
+    </button>
+  );
+}
+
+export function GeneralQaSidebar() {
+  const { t } = useI18n();
+  // 智能体图标配色分主题：深蓝主题用规范蓝-青系，其它（浅色）保持原高彩度色，互不影响。
+  const { theme } = useTheme();
+  const isDark = theme === "dark";
+  const pathname = usePathname();
+  const router = useRouter();
+  const navigate = useNavigate();
+
+  // ── 区块折叠状态（持久化，键独立于旧组件，避免互相影响）──
+  const [agentsCollapsed, setAgentsCollapsed] = useState(
+    () => localStorage.getItem("gqa-sidebar:agents-collapsed") === "true",
+  );
+  const [chatsCollapsed, setChatsCollapsed] = useState(
+    () => localStorage.getItem("gqa-sidebar:chats-collapsed") === "true",
+  );
+  const toggleAgents = () =>
+    setAgentsCollapsed((v) => {
+      localStorage.setItem("gqa-sidebar:agents-collapsed", String(!v));
+      return !v;
+    });
+  const toggleChats = () =>
+    setChatsCollapsed((v) => {
+      localStorage.setItem("gqa-sidebar:chats-collapsed", String(!v));
+      return !v;
+    });
+
+  // ── 常用智能体取数：置顶 → 我的 → 广场，去重补足到 MAX_FREQUENT（沿用旧逻辑）──
+  const { agents: mine } = useAgentsPaginated({ scope: "mine", page: 1, pageSize: 20 });
+  const { agents: square } = useAgentsPaginated({ scope: "square", page: 1, pageSize: 20 });
+  const frequent = useMemo<Agent[]>(() => {
+    const out: Agent[] = [];
+    const seen = new Set<string>();
+    const add = (agent: Agent | undefined) => {
+      if (!agent || out.length >= MAX_FREQUENT || seen.has(agent.id)) return;
+      seen.add(agent.id);
+      out.push(agent);
+    };
+    [...mine, ...square].filter((a) => a.is_pinned).forEach(add);
+    mine.forEach(add);
+    square.forEach(add);
+    return out;
+  }, [mine, square]);
+
+  // 当前嵌入的智能体（URL 上的 ?agent=）：用于高亮对应卡片。
+  const searchParams = useSearchParams();
+  const activeAgentId = searchParams.get("agent");
+
+  // 点击常用智能体 = 在通用问答右侧**原地嵌入**该智能体的对话页：停留在 /chats/new，
+  // 仅给 URL 加 ?agent=<id>，由 NewChatContent 把右侧问答区换成 AgentChatPage。
+  // 不跳转独立页面、左侧栏不变。每次点击都重开一个干净会话（NewChatContent 的 key）。
+  const handleSelectAgent = useCallback(
+    (agentId: string) => {
+      navigate(`/page/workspace/chats/new?agent=${encodeURIComponent(agentId)}`);
+    },
+    [navigate],
+  );
+
+  // ── 最近的对话取数 + 搜索 ──
+  const { thread_id: threadIdFromPath, agent_id: agentIdFromPath } = useParams<{
+    thread_id: string;
+    agent_id?: string;
+  }>();
+  const { data: threads = [] } = useThreads();
+  const { mutate: deleteThread } = useDeleteThread();
+  const { mutate: renameThread } = useRenameThread();
+  const { toggle: togglePin } = useToggleThreadPin();
+  const { toggle: toggleFavorite } = useToggleThreadFavorite();
+
+  const [query, setQuery] = useState("");
+  const [renameDialogOpen, setRenameDialogOpen] = useState(false);
+  const [renameThreadId, setRenameThreadId] = useState<string | null>(null);
+  const [renameValue, setRenameValue] = useState("");
+  const [shareThreadId, setShareThreadId] = useState<string | null>(null);
+
+  const visibleThreads = useMemo(
+    () =>
+      sortThreadsPinnedFirst(
+        threads.filter(
+          (thread) =>
+            thread.metadata?.thread_type !== "scheduler" &&
+            thread.metadata?.thread_type !== "bootstrap" &&
+            thread.metadata?.thread_type !== "roundtable",
+        ),
+      ),
+    [threads],
+  );
+
+  const filteredThreads = useMemo(() => {
+    const q = query.trim().toLowerCase();
+    const list = q
+      ? visibleThreads.filter((thread) =>
+          titleOfThread(thread, t.pages.newChat).toLowerCase().includes(q),
+        )
+      : visibleThreads;
+    // 未搜索时与旧版一致：最多展示 10 条
+    return q ? list : list.slice(0, 10);
+  }, [visibleThreads, query, t.pages.newChat]);
+
+  const handleNewChatClick = (event: React.MouseEvent<HTMLAnchorElement>) => {
+    if (event.metaKey || event.ctrlKey || event.shiftKey || event.button === 1) return;
+    event.preventDefault();
+    navigate(NEW_CHAT_PATH, { replace: true, state: { resetAt: Date.now() } });
+  };
+
+  const handleDelete = useCallback(
+    (threadId: string) => {
+      deleteThread({ threadId });
+      if (threadId === threadIdFromPath) {
+        const threadIndex = threads.findIndex((th) => th.thread_id === threadId);
+        let nextThreadPath = pathOfThread("new", { agent_id: agentIdFromPath });
+        if (threadIndex > -1) {
+          if (threads[threadIndex + 1]) {
+            nextThreadPath = pathOfThread(threads[threadIndex + 1]!);
+          } else if (threads[threadIndex - 1]) {
+            nextThreadPath = pathOfThread(threads[threadIndex - 1]!);
+          }
+        }
+        void router.push(nextThreadPath);
+      }
+    },
+    [agentIdFromPath, deleteThread, router, threadIdFromPath, threads],
+  );
+
+  const handleRenameClick = useCallback((threadId: string, currentTitle: string) => {
+    setRenameThreadId(threadId);
+    setRenameValue(currentTitle);
+    setRenameDialogOpen(true);
+  }, []);
+
+  const handleRenameSubmit = useCallback(() => {
+    if (renameThreadId && renameValue.trim()) {
+      renameThread({ threadId: renameThreadId, title: renameValue.trim() });
+      setRenameDialogOpen(false);
+      setRenameThreadId(null);
+      setRenameValue("");
+    }
+  }, [renameThread, renameThreadId, renameValue]);
+
+  const handleExport = useCallback(
+    async (thread: AgentThread, format: "markdown" | "json") => {
+      try {
+        const apiClient = getAPIClient();
+        const state = await apiClient.threads.getState<AgentThreadState>(
+          thread.thread_id,
+        );
+        const messages = state.values?.messages ?? [];
+        if (messages.length === 0) {
+          toast.error(t.conversation.noMessages);
+          return;
+        }
+        if (format === "markdown") exportThreadAsMarkdown(thread, messages);
+        else exportThreadAsJSON(thread, messages);
+        toast.success(t.common.exportSuccess);
+      } catch {
+        toast.error("Failed to export conversation");
+      }
+    },
+    [t],
+  );
+
+  const isStatic = env.NEXT_PUBLIC_STATIC_WEBSITE_ONLY === "true";
+
+  return (
+    // 背景走 sidebar token（浅主题=白、深蓝主题=深蓝），不要写死 bg-white，否则全量版深蓝主题下侧栏仍是白底。
+    <div className="flex h-full min-h-0 flex-col bg-sidebar text-sidebar-foreground">
+      {/* ── 顶部：新对话胶囊 + 收起按钮 ── */}
+      <div className="flex items-center gap-2 px-3 pt-3 pb-2">
+        {/* 新对话按钮：完全自定义样式（不复用 Button variant）。borderColor / color 走
+            inline style，避免被全局 border-color（--border）覆盖；无阴影。主色描边 + 5px 圆角。
+            底色用「主色混入透明」=主色低透明度叠在当前侧栏底色上，浅/红/蓝/深蓝主题都自适应
+            （不要混 white，否则深蓝主题下会变成浅色块）。 */}
+        <Link
+          href={NEW_CHAT_PATH}
+          onClick={handleNewChatClick}
+          style={{ borderColor: "var(--primary)", color: "var(--primary)" }}
+          className={cn(
+            "flex h-10 min-w-0 flex-1 items-center justify-center gap-1.5 rounded-[5px] text-[14px] font-semibold transition-colors",
+            "border border-solid",
+            "bg-[color-mix(in_srgb,var(--primary)_14%,transparent)] hover:bg-[color-mix(in_srgb,var(--primary)_24%,transparent)]",
+          )}
+        >
+          <Plus className="size-4 stroke-[2.5]" />
+          <span>{t.sidebar.newChat}</span>
+        </Link>
+        <SidebarCollapseButton className="size-9 shrink-0 rounded-full border border-border bg-card text-muted-foreground shadow-sm hover:bg-accent" />
+      </div>
+
+      <div className="min-h-0 flex-1 overflow-y-auto px-2 pb-2">
+        {/* ── 常用智能体 ── */}
+        {frequent.length > 0 && (
+          <div className="pt-2">
+            <SectionLabel
+              label="常用智能体"
+              collapsed={agentsCollapsed}
+              onToggle={toggleAgents}
+            />
+            {!agentsCollapsed && (
+              <div className="mt-1 flex flex-col gap-1">
+                {frequent.map((agent) => {
+                  const isActive = activeAgentId === agent.id;
+                  const { from, to } = agentColor(agent.id, isDark);
+                  return (
+                    <button
+                      key={agent.id}
+                      type="button"
+                      onClick={() => handleSelectAgent(agent.id)}
+                      className={cn(
+                        "group relative flex w-full items-start gap-2.5 rounded-xl border px-2.5 py-2 text-left transition-colors",
+                        isActive
+                          ? "border-primary/40 bg-primary/5"
+                          : "border-transparent hover:border-border hover:bg-accent/50",
+                      )}
+                    >
+                      <span
+                        className="mt-0.5 flex size-9 shrink-0 items-center justify-center rounded-lg text-white shadow-sm"
+                        style={{ backgroundImage: `linear-gradient(135deg, ${from}, ${to})` }}
+                      >
+                        <Bot className="size-[18px]" />
+                      </span>
+                      <span className="min-w-0 flex-1">
+                        <span
+                          className={cn(
+                            "block truncate text-[13px] font-semibold",
+                            isActive ? "text-primary" : "text-foreground",
+                          )}
+                        >
+                          {desensitize(agent.name)}
+                        </span>
+                        <span className="mt-0.5 block truncate text-[11px] leading-snug text-muted-foreground">
+                          {agent.description ? desensitize(agent.description) : "智能体助手"}
+                        </span>
+                      </span>
+                      {isActive && (
+                        <span className="absolute top-1/2 right-1 h-6 w-1 -translate-y-1/2 rounded-full bg-primary" />
+                      )}
+                    </button>
+                  );
+                })}
+                <Link
+                  href={MORE_AGENTS_HREF}
+                  className="flex items-center justify-between rounded-lg px-3 py-1.5 text-[12px] text-muted-foreground/70 transition-colors hover:bg-accent/50 hover:text-foreground"
+                >
+                  <span>{t.common.more}</span>
+                  <ChevronRight className="size-3.5 shrink-0" />
+                </Link>
+              </div>
+            )}
+          </div>
+        )}
+
+        {/* ── 最近的对话 ── */}
+        <div className="mt-3 border-t border-sidebar-border/60 pt-3">
+          <SectionLabel
+            label={t.sidebar.recentChats}
+            count={visibleThreads.length}
+            collapsed={chatsCollapsed}
+            onToggle={toggleChats}
+          />
+          {!chatsCollapsed && (
+            <div className="mt-1.5">
+              {/* 历史会话搜索框 */}
+              <div className="relative mb-1.5 px-1">
+                <Search className="pointer-events-none absolute top-1/2 left-3.5 size-3.5 -translate-y-1/2 text-muted-foreground/60" />
+                <Input
+                  value={query}
+                  onChange={(e) => setQuery(e.target.value)}
+                  placeholder="搜索历史会话…"
+                  className="h-8 rounded-lg border-border bg-card pl-8 text-[12px]"
+                />
+              </div>
+
+              <div className="flex flex-col gap-0.5">
+                {filteredThreads.length === 0 ? (
+                  <div className="px-3 py-4 text-center text-[12px] text-muted-foreground/70">
+                    {query ? "没有匹配的会话" : t.chats.noChats}
+                  </div>
+                ) : (
+                  filteredThreads.map((thread) => {
+                    const isActive = pathOfThread(thread) === pathname;
+                    const pinned = isThreadPinned(thread);
+                    const favorited = isThreadFavorite(thread);
+                    return (
+                      <div
+                        key={thread.thread_id}
+                        className="group/item relative flex items-center"
+                      >
+                        <Link
+                          href={pathOfThread(thread)}
+                          className={cn(
+                            "relative flex h-9 min-w-0 flex-1 items-center gap-2 rounded-lg px-2.5 text-[13px] transition-colors",
+                            isActive
+                              ? "border border-primary/40 bg-primary/5 font-medium text-primary"
+                              : "text-muted-foreground hover:bg-accent/60 hover:text-foreground",
+                          )}
+                        >
+                          <MessageSquare
+                            className={cn(
+                              "size-3.5 shrink-0",
+                              isActive ? "text-primary" : "text-muted-foreground/70",
+                            )}
+                          />
+                          {pinned && (
+                            <Pin className="size-3 shrink-0 text-muted-foreground/70" />
+                          )}
+                          <span className="truncate">
+                            {desensitize(titleOfThread(thread, t.pages.newChat))}
+                          </span>
+                          {isActive && (
+                            <span className="absolute top-1/2 right-1 h-5 w-1 -translate-y-1/2 rounded-full bg-primary" />
+                          )}
+                        </Link>
+                        {!isStatic && (
+                          <DropdownMenu>
+                            <DropdownMenuTrigger asChild>
+                              <button
+                                type="button"
+                                className="absolute right-1 flex size-6 items-center justify-center rounded-md text-muted-foreground opacity-0 transition-opacity hover:bg-background group-hover/item:opacity-100 data-[state=open]:opacity-100"
+                              >
+                                <MoreHorizontal className="size-4" />
+                                <span className="sr-only">{t.common.more}</span>
+                              </button>
+                            </DropdownMenuTrigger>
+                            <DropdownMenuContent className="w-48 rounded-lg" side="right" align="start">
+                              <DropdownMenuItem
+                                onSelect={() =>
+                                  handleRenameClick(
+                                    thread.thread_id,
+                                    titleOfThread(thread, t.pages.newChat),
+                                  )
+                                }
+                              >
+                                <Pencil className="text-muted-foreground" />
+                                <span>{t.common.rename}</span>
+                              </DropdownMenuItem>
+                              <DropdownMenuItem onSelect={() => togglePin(thread.thread_id, !pinned)}>
+                                {pinned ? (
+                                  <PinOff className="text-muted-foreground" />
+                                ) : (
+                                  <Pin className="text-muted-foreground" />
+                                )}
+                                <span>{pinned ? t.common.unpin : t.common.pin}</span>
+                              </DropdownMenuItem>
+                              <DropdownMenuItem
+                                onSelect={() => toggleFavorite(thread.thread_id, !favorited)}
+                              >
+                                <Star
+                                  className={
+                                    favorited
+                                      ? "fill-amber-400 text-amber-400"
+                                      : "text-muted-foreground"
+                                  }
+                                />
+                                <span>{favorited ? t.common.unfavorite : t.common.favorite}</span>
+                              </DropdownMenuItem>
+                              <DropdownMenuItem onSelect={() => setShareThreadId(thread.thread_id)}>
+                                <Share2 className="text-muted-foreground" />
+                                <span>{t.common.share}</span>
+                              </DropdownMenuItem>
+                              <DropdownMenuSub>
+                                <DropdownMenuSubTrigger>
+                                  <Download className="text-muted-foreground" />
+                                  <span>{t.common.export}</span>
+                                </DropdownMenuSubTrigger>
+                                <DropdownMenuSubContent>
+                                  <DropdownMenuItem onSelect={() => handleExport(thread, "markdown")}>
+                                    <FileText className="text-muted-foreground" />
+                                    <span>{t.common.exportAsMarkdown}</span>
+                                  </DropdownMenuItem>
+                                  <DropdownMenuItem onSelect={() => handleExport(thread, "json")}>
+                                    <FileJson className="text-muted-foreground" />
+                                    <span>{t.common.exportAsJSON}</span>
+                                  </DropdownMenuItem>
+                                </DropdownMenuSubContent>
+                              </DropdownMenuSub>
+                              <DropdownMenuSeparator />
+                              <DropdownMenuItem onSelect={() => handleDelete(thread.thread_id)}>
+                                <Trash2 className="text-muted-foreground" />
+                                <span>{t.common.delete}</span>
+                              </DropdownMenuItem>
+                            </DropdownMenuContent>
+                          </DropdownMenu>
+                        )}
+                      </div>
+                    );
+                  })
+                )}
+                {!query && (
+                  <Link
+                    href={MORE_CHATS_HREF}
+                    className="mt-0.5 flex items-center justify-between rounded-lg px-3 py-1.5 text-[12px] text-muted-foreground/70 transition-colors hover:bg-accent/50 hover:text-foreground"
+                  >
+                    <span>更多最近对话</span>
+                    <ChevronRight className="size-3.5 shrink-0" />
+                  </Link>
+                )}
+              </div>
+            </div>
+          )}
+        </div>
+      </div>
+
+      {/* 重命名弹窗 */}
+      <Dialog open={renameDialogOpen} onOpenChange={setRenameDialogOpen}>
+        <DialogContent className="sm:max-w-[425px]">
+          <DialogHeader>
+            <DialogTitle>{t.common.rename}</DialogTitle>
+          </DialogHeader>
+          <div className="py-4">
+            <Input
+              value={renameValue}
+              onChange={(e) => setRenameValue(e.target.value)}
+              placeholder={t.common.rename}
+              onKeyDown={(e) => {
+                if (e.key === "Enter" && !isIMEComposing(e)) {
+                  e.preventDefault();
+                  handleRenameSubmit();
+                }
+              }}
+            />
+          </div>
+          <DialogFooter>
+            <Button variant="outline" onClick={() => setRenameDialogOpen(false)}>
+              {t.common.cancel}
+            </Button>
+            <Button onClick={handleRenameSubmit}>{t.common.save}</Button>
+          </DialogFooter>
+        </DialogContent>
+      </Dialog>
+
+      {/* 分享弹窗 */}
+      <ShareConversationDialog
+        threadId={shareThreadId}
+        open={shareThreadId !== null}
+        onOpenChange={(o) => {
+          if (!o) setShareThreadId(null);
+        }}
+      />
+    </div>
+  );
+}
diff --git a/frontend-web/src/components/workspace/input-box.tsx b/frontend-web/src/components/workspace/input-box.tsx
index 3e3b53b4..f4a30d63 100644
--- a/frontend-web/src/components/workspace/input-box.tsx
+++ b/frontend-web/src/components/workspace/input-box.tsx
@@ -1,1264 +1,1289 @@
-"use client";
-
-import type { ChatStatus } from "ai";
-import {
-  BrainCircuitIcon,
-  BookMarkedIcon,
-  CheckIcon,
-  FileTextIcon,
-  GraduationCapIcon,
-  LightbulbIcon,
-  PaperclipIcon,
-  PlusIcon,
-  SparklesIcon,
-  RocketIcon,
-  XIcon,
-  ZapIcon,
-} from "lucide-react";
-import { useSearchParams } from "next/navigation";
-import {
-  useCallback,
-  useEffect,
-  useMemo,
-  useRef,
-  useState,
-  type ComponentProps,
-} from "react";
-
-import {
-  PromptInput,
-  PromptInputActionMenu,
-  PromptInputActionMenuContent,
-  PromptInputActionMenuItem,
-  PromptInputActionMenuTrigger,
-  PromptInputAttachment,
-  PromptInputAttachments,
-  PromptInputBody,
-  PromptInputButton,
-  PromptInputFooter,
-  PromptInputSubmit,
-  PromptInputTextarea,
-  PromptInputTools,
-  usePromptInputAttachments,
-  usePromptInputController,
-  type PromptInputMessage,
-} from "@/components/ai-elements/prompt-input";
-import { Button } from "@/components/ui/button";
-import { ConfettiButton } from "@/components/ui/confetti-button";
-import {
-  Dialog,
-  DialogContent,
-  DialogDescription,
-  DialogFooter,
-  DialogHeader,
-  DialogTitle,
-} from "@/components/ui/dialog";
-import {
-  DropdownMenuGroup,
-  DropdownMenuLabel,
-  DropdownMenuSeparator,
-} from "@/components/ui/dropdown-menu";
-import { apiFetch } from "@/core/api";
-import { getBackendBaseURL } from "@/core/config";
-import { useI18n } from "@/core/i18n/hooks";
-import { useModels } from "@/core/models/hooks";
-import type { AgentThreadContext } from "@/core/threads";
-import { textOfMessage } from "@/core/threads/utils";
-import { desensitize } from "@/lib/desensitize";
-import { cn } from "@/lib/utils";
-
-import { AgentSelector } from "@/components/workspace/agents/agent-selector";
-
-import {
-  ModelSelector,
-  ModelSelectorContent,
-  ModelSelectorInput,
-  ModelSelectorItem,
-  ModelSelectorList,
-  ModelSelectorName,
-  ModelSelectorTrigger,
-} from "../ai-elements/model-selector";
-import { Suggestion, Suggestions } from "../ai-elements/suggestion";
-import {
-  DropdownMenu,
-  DropdownMenuContent,
-  DropdownMenuItem,
-  DropdownMenuTrigger,
-} from "../ui/dropdown-menu";
-
-import { useRecommendedQuestions } from "@/core/recommended-questions/hooks";
-import { useUserPreferences } from "@/core/user-preferences";
-import {
-  isThreadRagModeLocked,
-  lockThreadRagMode,
-  useLocalSettings,
-} from "@/core/settings";
-
-
-import { CustomPromptBar } from "./custom-prompt-menu";
-import { useThread } from "./messages/context";
-import { ModeHoverGuide } from "./mode-hover-guide";
-import { Tooltip } from "./tooltip";
-
-type InputMode = "flash" | "thinking" | "pro" | "ultra";
-
-
-function getResolvedMode(
-  mode: InputMode | undefined,
-  supportsThinking: boolean,
-): InputMode {
-  if (!supportsThinking && mode !== "flash") {
-    return "flash";
-  }
-  if (mode) {
-    return mode;
-  }
-  return supportsThinking ? "pro" : "flash";
-}
-
-export function InputBox({
-  className,
-  disabled,
-  autoFocus,
-  status = "ready",
-  context,
-  extraHeader,
-  isNewThread,
-  threadId,
-  initialValue,
-  onContextChange,
-  onFollowupsVisibilityChange,
-  onSubmit,
-  onStop,
-  promptPrefix,
-  agentSelector,
-  assistantId,
-  hideAdminRecommendedQuestions,
-  compact,
-  ...props
-}: Omit<ComponentProps<typeof PromptInput>, "onSubmit"> & {
-  assistantId?: string | null;
-  status?: ChatStatus;
-  /** 紧凑模式（正常 App）新对话落地态：去掉旋转彩边、用 CM 参考图的白卡输入框，
-   *  并在推荐问题上方加「试试这些高频业务场景」标题。仅落地态生效，默认 false。 */
-  compact?: boolean;
-  disabled?: boolean;
-  context: Omit<
-    AgentThreadContext,
-    "thread_id" | "is_plan_mode" | "thinking_enabled" | "subagent_enabled"
-  > & {
-    mode: "flash" | "thinking" | "pro" | "ultra" | undefined;
-    reasoning_effort?: "minimal" | "low" | "medium" | "high";
-    writing_mode?: boolean;
-    memory_injection_enabled?: boolean;
-  };
-  extraHeader?: React.ReactNode;
-  isNewThread?: boolean;
-  /** Suppress the global admin "recommended questions" row. The agent chat page
-   * sets this so only the agent's own common questions surface on its home. */
-  hideAdminRecommendedQuestions?: boolean;
-  threadId: string;
-  initialValue?: string;
-  onContextChange?: (
-    context: Omit<
-      AgentThreadContext,
-      "thread_id" | "is_plan_mode" | "thinking_enabled" | "subagent_enabled"
-    > & {
-      mode: "flash" | "thinking" | "pro" | "ultra" | undefined;
-      reasoning_effort?: "minimal" | "low" | "medium" | "high";
-      writing_mode?: boolean;
-      memory_injection_enabled?: boolean;
-    },
-  ) => void;
-  onFollowupsVisibilityChange?: (visible: boolean) => void;
-  onSubmit?: (message: PromptInputMessage) => void;
-  onStop?: () => void;
-  /**
-   * Optional toggle rendered next to the model selector. Used for the
-   * admin-configurable prompt-prefix switch on the chat input. Omitting the
-   * prop hides the button entirely.
-   */
-  promptPrefix?: {
-    label: string;
-    on: boolean;
-    onToggle: () => void;
-  };
-  /**
-   * Optional agent picker shown alongside the model selector. Only the new-chat
-   * homepage (``ChatPage``) passes this; ``AgentChatPage`` omits it because the
-   * thread is already bound to one agent.
-   */
-  agentSelector?: {
-    selectedAgentId: string | null;
-    onAgentChange: (agentId: string | null) => void;
-  };
-}) {
-  const { t } = useI18n();
-  const [localSettings, setLocalSettings] = useLocalSettings();
-  const searchParams = useSearchParams();
-  const [modelDialogOpen, setModelDialogOpen] = useState(false);
-  const { models } = useModels();
-  const { thread, isMock } = useThread();
-  const { textInput } = usePromptInputController();
-  const promptRootRef = useRef<HTMLDivElement | null>(null);
-
-  const [followups, setFollowups] = useState<string[]>([]);
-  const [followupsHidden, setFollowupsHidden] = useState(false);
-  const [followupsLoading, setFollowupsLoading] = useState(false);
-  const lastGeneratedForAiIdRef = useRef<string | null>(null);
-  const wasStreamingRef = useRef(false);
-
-  const [confirmOpen, setConfirmOpen] = useState(false);
-  const [pendingSuggestion, setPendingSuggestion] = useState<string | null>(
-    null,
-  );
-
-  const { questions: recommendedQuestions } = useRecommendedQuestions();
-  const userPreferences = useUserPreferences();
-  const showAdminRecommendedQuestions =
-    !hideAdminRecommendedQuestions &&
-    !userPreferences.hide_admin_prompt_prefix &&
-    recommendedQuestions.length > 0;
-
-  useEffect(() => {
-    if (models.length === 0) {
-      return;
-    }
-    const currentModel = models.find((m) => m.name === context.model_name);
-    const fallbackModel = currentModel ?? models[0]!;
-    const supportsThinking = fallbackModel.supports_thinking ?? false;
-    const nextModelName = fallbackModel.name;
-    const nextMode = getResolvedMode(context.mode, supportsThinking);
-
-    if (context.model_name === nextModelName && context.mode === nextMode) {
-      return;
-    }
-
-    onContextChange?.({
-      ...context,
-      model_name: nextModelName,
-      mode: nextMode,
-    });
-  }, [context, models, onContextChange]);
-
-  const selectedModel = useMemo(() => {
-    if (models.length === 0) {
-      return undefined;
-    }
-    return models.find((m) => m.name === context.model_name) ?? models[0];
-  }, [context.model_name, models]);
-
-  const resolvedModelName = selectedModel?.name;
-
-  const supportThinking = useMemo(
-    () => selectedModel?.supports_thinking ?? false,
-    [selectedModel],
-  );
-
-  const supportReasoningEffort = useMemo(
-    () => selectedModel?.supports_reasoning_effort ?? false,
-    [selectedModel],
-  );
-
-  const handleModelSelect = useCallback(
-    (model_name: string) => {
-      const model = models.find((m) => m.name === model_name);
-      if (!model) {
-        return;
-      }
-      onContextChange?.({
-        ...context,
-        model_name,
-        mode: getResolvedMode(context.mode, model.supports_thinking ?? false),
-        reasoning_effort: context.reasoning_effort,
-      });
-      setModelDialogOpen(false);
-    },
-    [onContextChange, context, models],
-  );
-
-  const handleModeSelect = useCallback(
-    (mode: InputMode) => {
-      onContextChange?.({
-        ...context,
-        mode: getResolvedMode(mode, supportThinking),
-        reasoning_effort:
-          mode === "ultra"
-            ? "high"
-            : mode === "pro"
-              ? "medium"
-              : mode === "thinking"
-                ? "low"
-                : "minimal",
-      });
-    },
-    [onContextChange, context, supportThinking],
-  );
-
-  const handleReasoningEffortSelect = useCallback(
-    (effort: "minimal" | "low" | "medium" | "high") => {
-      onContextChange?.({
-        ...context,
-        reasoning_effort: effort,
-      });
-    },
-    [onContextChange, context],
-  );
-
-  const handleWritingModeToggle = useCallback(() => {
-    onContextChange?.({
-      ...context,
-      writing_mode: !context.writing_mode,
-    });
-  }, [context, onContextChange]);
-
-  const memoryInjectionEnabled = context.memory_injection_enabled ?? true;
-  const handleMemoryInjectionToggle = useCallback(() => {
-    onContextChange?.({
-      ...context,
-      memory_injection_enabled: !memoryInjectionEnabled,
-    });
-  }, [context, memoryInjectionEnabled, onContextChange]);
-
-  const ragModeLocked = isThreadRagModeLocked(threadId);
-  const ragModeEnabled = localSettings.display.ragModeEnabled || ragModeLocked;
-  const handleRagModeToggle = useCallback(() => {
-    if (ragModeLocked) {
-      return;
-    }
-    setLocalSettings("display", { ragModeEnabled: !ragModeEnabled });
-  }, [ragModeEnabled, ragModeLocked, setLocalSettings]);
-
-  const handleSubmit = useCallback(
-    async (message: PromptInputMessage) => {
-      if (status === "streaming") {
-        onStop?.();
-        return;
-      }
-      if (!message.text) {
-        return;
-      }
-      setFollowups([]);
-      setFollowupsHidden(false);
-      setFollowupsLoading(false);
-      if (ragModeEnabled) {
-        lockThreadRagMode(threadId);
-      }
-
-      // Guard against submitting before the initial model auto-selection
-      // effect has flushed thread settings to storage/state.
-      if (resolvedModelName && context.model_name !== resolvedModelName) {
-        onContextChange?.({
-          ...context,
-          model_name: resolvedModelName,
-          mode: getResolvedMode(
-            context.mode,
-            selectedModel?.supports_thinking ?? false,
-          ),
-        });
-        setTimeout(() => onSubmit?.(message), 0);
-        return;
-      }
-
-      onSubmit?.(message);
-    },
-    [
-      context,
-      onContextChange,
-      onSubmit,
-      onStop,
-      resolvedModelName,
-      ragModeEnabled,
-      selectedModel?.supports_thinking,
-      status,
-      threadId,
-    ],
-  );
-
-  const requestFormSubmit = useCallback(() => {
-    const form = promptRootRef.current?.querySelector("form");
-    form?.requestSubmit();
-  }, []);
-
-  const handleFollowupClick = useCallback(
-    (suggestion: string) => {
-      if (status === "streaming") {
-        return;
-      }
-      const current = (textInput.value ?? "").trim();
-      if (current) {
-        setPendingSuggestion(suggestion);
-        setConfirmOpen(true);
-        return;
-      }
-      textInput.setInput(suggestion);
-      setFollowupsHidden(true);
-      setTimeout(() => requestFormSubmit(), 0);
-    },
-    [requestFormSubmit, status, textInput],
-  );
-
-  const confirmReplaceAndSend = useCallback(() => {
-    if (!pendingSuggestion) {
-      setConfirmOpen(false);
-      return;
-    }
-    textInput.setInput(pendingSuggestion);
-    setFollowupsHidden(true);
-    setConfirmOpen(false);
-    setPendingSuggestion(null);
-    setTimeout(() => requestFormSubmit(), 0);
-  }, [pendingSuggestion, requestFormSubmit, textInput]);
-
-  const confirmAppendAndSend = useCallback(() => {
-    if (!pendingSuggestion) {
-      setConfirmOpen(false);
-      return;
-    }
-    const current = (textInput.value ?? "").trim();
-    const next = current
-      ? `${current}\n${pendingSuggestion}`
-      : pendingSuggestion;
-    textInput.setInput(next);
-    setFollowupsHidden(true);
-    setConfirmOpen(false);
-    setPendingSuggestion(null);
-    setTimeout(() => requestFormSubmit(), 0);
-  }, [pendingSuggestion, requestFormSubmit, textInput]);
-
-  const showFollowups =
-    !disabled &&
-    !isNewThread &&
-    !followupsHidden &&
-    (followupsLoading || followups.length > 0);
-
-  const followupsVisibilityChangeRef = useRef(onFollowupsVisibilityChange);
-
-  useEffect(() => {
-    followupsVisibilityChangeRef.current = onFollowupsVisibilityChange;
-  }, [onFollowupsVisibilityChange]);
-
-  useEffect(() => {
-    followupsVisibilityChangeRef.current?.(showFollowups);
-  }, [showFollowups]);
-
-  useEffect(() => {
-    return () => followupsVisibilityChangeRef.current?.(false);
-  }, []);
-
-  useEffect(() => {
-    const streaming = status === "streaming";
-    const wasStreaming = wasStreamingRef.current;
-    wasStreamingRef.current = streaming;
-    if (!wasStreaming || streaming) {
-      return;
-    }
-
-    if (disabled || isMock) {
-      return;
-    }
-
-    const lastAi = [...thread.messages].reverse().find((m) => m.type === "ai");
-    const lastAiId = lastAi?.id ?? null;
-    if (!lastAiId || lastAiId === lastGeneratedForAiIdRef.current) {
-      return;
-    }
-    lastGeneratedForAiIdRef.current = lastAiId;
-
-    const recent = thread.messages
-      .filter((m) => m.type === "human" || m.type === "ai")
-      .map((m) => {
-        const role = m.type === "human" ? "user" : "assistant";
-        const content = textOfMessage(m) ?? "";
-        return { role, content };
-      })
-      .filter((m) => m.content.trim().length > 0)
-      .slice(-6);
-
-    if (recent.length === 0) {
-      return;
-    }
-
-    const controller = new AbortController();
-    setFollowupsHidden(false);
-    setFollowupsLoading(true);
-    setFollowups([]);
-
-    apiFetch(`${getBackendBaseURL()}/api/threads/${threadId}/suggestions`, {
-      method: "POST",
-      headers: { "Content-Type": "application/json" },
-      body: JSON.stringify({
-        messages: recent,
-        n: 3,
-        model_name: context.model_name ?? undefined,
-      }),
-      signal: controller.signal,
-    })
-      .then(async (res) => {
-        if (!res.ok) {
-          return { suggestions: [] as string[] };
-        }
-        return (await res.json()) as { suggestions?: string[] };
-      })
-      .then((data) => {
-        const suggestions = (data.suggestions ?? [])
-          .map((s) => (typeof s === "string" ? s.trim() : ""))
-          .filter((s) => s.length > 0)
-          .slice(0, 5);
-        setFollowups(suggestions);
-      })
-      .catch(() => {
-        setFollowups([]);
-      })
-      .finally(() => {
-        setFollowupsLoading(false);
-      });
-
-    return () => controller.abort();
-  }, [context.model_name, disabled, isMock, status, thread.messages, threadId]);
-
-  return (
-    <div ref={promptRootRef} className="relative flex flex-col gap-6">
-      {extraHeader && (
-        <div className="flex items-center justify-center">
-          {extraHeader}
-        </div>
-      )}
-
-      {isNewThread && searchParams.get("mode") !== "skill" && showAdminRecommendedQuestions && (
-        <div className={cn(compact && "-mt-2 flex flex-col gap-3")}>
-          {compact && (
-            <div className="text-muted-foreground flex items-center justify-center gap-1.5 text-xs font-medium">
-              <SparklesIcon className="size-3 text-[var(--primary)]" />
-              <span>试试这些高频业务场景</span>
-            </div>
-          )}
-        <div className="flex flex-wrap justify-center gap-2.5 px-4 mb-[10px]">
-          {recommendedQuestions.map((q) => (
-            <button
-              key={q.id}
-              type="button"
-              className={cn(
-                "rounded-full border border-[#e4e4e7] bg-white px-4 py-2 text-sm text-[#18181b] shadow-none transition-colors hover:bg-[#f4f4f5] cursor-pointer dark:border-slate-600 dark:bg-slate-700/50 dark:text-slate-300 dark:hover:bg-slate-700",
-                // 紧凑模式（CM 参考图，F12 精确还原）：白底灰字灰边 + 主色 hover
-                compact &&
-                  "px-4 py-2 text-xs font-medium text-gray-700 border-gray-100 shadow-xs transition-all duration-200 hover:bg-[color-mix(in_srgb,var(--primary)_3%,white)] hover:text-[var(--primary)] hover:border-[color-mix(in_srgb,var(--primary)_22%,transparent)] dark:text-slate-300 dark:border-slate-600 dark:hover:text-[var(--primary)]",
-              )}
-              onClick={() => {
-                textInput.setInput(q.content);
-                setFollowupsHidden(true);
-                setTimeout(() => requestFormSubmit(), 0);
-              }}
-            >
-              {desensitize(q.title)}
-            </button>
-          ))}
-        </div>
-        </div>
-      )}
-
-      {showFollowups && (
-        <div className="flex items-center justify-center pb-2">
-          <div className="flex items-center gap-2">
-            {followupsLoading ? (
-              <div className="text-muted-foreground bg-background/80 rounded-full border px-4 py-2 text-xs backdrop-blur-sm">
-                {t.inputBox.followupLoading}
-              </div>
-            ) : (
-              <Suggestions className="min-h-16 w-fit items-start">
-                {followups.map((s) => (
-                  <Suggestion
-                    key={s}
-                    suggestion={s}
-                    onClick={() => handleFollowupClick(s)}
-                  />
-                ))}
-                <Button
-                  aria-label={t.common.close}
-                  className="text-muted-foreground cursor-pointer rounded-full px-3 text-xs font-normal"
-                  variant="outline"
-                  size="sm"
-                  type="button"
-                  onClick={() => setFollowupsHidden(true)}
-                >
-                  <XIcon className="size-4" />
-                </Button>
-              </Suggestions>
-            )}
-          </div>
-        </div>
-      )}
-      <div className="flex flex-col gap-1.5">
-      {isNewThread && (
-        <CustomPromptBar
-          className="px-1"
-          currentAgentId={agentSelector?.selectedAgentId ?? assistantId ?? null}
-          onInsert={(content) => {
-            const current = textInput.value?.trim() ?? "";
-            textInput.setInput(current ? `${current}\n${content}` : content);
-            setTimeout(() => {
-              document
-                .querySelector<HTMLTextAreaElement>("textarea[name='message']")
-                ?.focus();
-            }, 0);
-          }}
-        />
-      )}
-      <div className={cn(isNewThread && !compact && "gemini-gradient-border gemini-shadow")}>
-      <PromptInput
-        className={cn(
-          "transition-all duration-300 ease-out",
-          // 紧凑落地态（CM 参考图）：表单作为白卡，内部 input-group 透明无边框，
-          // 避免双层描边；圆角对齐 2xl。
-          compact && isNewThread
-            ? "*:data-[slot='input-group']:rounded-2xl *:data-[slot='input-group']:border-0 *:data-[slot='input-group']:bg-transparent *:data-[slot='input-group']:shadow-none"
-            : "*:data-[slot='input-group']:rounded-3xl",
-          // 紧凑落地态（CM 参考图）：白卡 + 细边 + 轻阴影 + 主色聚焦环（无旋转彩边）
-          compact && isNewThread
-            ? "rounded-2xl border border-gray-200/70 bg-white overflow-hidden shadow-[0_4.5px_22px_rgba(0,0,0,0.025)] transition-all duration-300 focus-within:border-[color-mix(in_srgb,var(--primary)_35%,transparent)] focus-within:ring-4 focus-within:ring-[color-mix(in_srgb,var(--primary)_5%,transparent)] focus-within:shadow-[0_10px_35px_color-mix(in_srgb,var(--primary)_4%,transparent)] dark:border-slate-700 dark:bg-slate-800"
-            : // landing 态：透明，让外层 gemini-gradient-border 提供白卡 + 旋转描边
-              // 对话态：常规白卡 + 中性边框 + 轻阴影
-              isNewThread
-              ? "rounded-3xl bg-transparent border-0 shadow-none ring-0"
-              : "rounded-3xl bg-background border border-border shadow-sm transition-shadow hover:shadow-md focus-within:border-ring/40 focus-within:shadow-md focus-within:ring-1 focus-within:ring-ring/20 dark:bg-slate-800 dark:border-slate-700",
-          className,
-        )}
-        disabled={disabled}
-        globalDrop
-        multiple
-        onSubmit={handleSubmit}
-        {...props}
-      >
-        <PromptInputAttachments>
-          {(attachment) => <PromptInputAttachment data={attachment} />}
-        </PromptInputAttachments>
-        <PromptInputBody
-          className={cn(!compact && "absolute top-0 right-0 left-0 z-3")}
-        >
-          <PromptInputTextarea
-            className={cn(
-              "size-full px-2 py-1 text-[15px] leading-7 placeholder:text-muted-foreground/60",
-              isNewThread && "min-h-[110px]!",
-              compact && "min-h-[84px]! px-3 pt-3 pb-1 text-[14px]",
-            )}
-            disabled={disabled}
-            placeholder={t.inputBox.placeholder}
-            autoFocus={autoFocus}
-            defaultValue={initialValue}
-          />
-        </PromptInputBody>
-        <PromptInputFooter
-          className={cn(
-            "flex items-center border-t-0 px-2 pt-1 pb-1.5",
-            compact && "gap-2 px-3 pt-1.5 pb-3",
-          )}
-        >
-          <PromptInputTools
-            className={cn("gap-1.5", compact && "flex-1 flex-wrap gap-2")}
-          >
-            {/* TODO: Add more connectors here
-          <PromptInputActionMenu>
-            <PromptInputActionMenuTrigger className="px-2!" />
-            <PromptInputActionMenuContent>
-              <PromptInputActionAddAttachments
-                label={t.inputBox.addAttachments}
-              />
-            </PromptInputActionMenuContent>
-          </PromptInputActionMenu> */}
-            <AddAttachmentsButton
-              className={cn(
-                "rounded-full border-0 bg-transparent px-2.5! shadow-none text-muted-foreground hover:bg-muted/70 hover:text-foreground dark:text-slate-400 dark:hover:bg-slate-700/70 dark:hover:text-slate-100",
-                compact &&
-                  "gap-1.5! rounded-lg px-2.5! py-1! text-[11px] font-medium",
-              )}
-            />
-            <Tooltip
-              content={
-                context.writing_mode
-                  ? "写作模式已开启：最终输出会强制保存为 Markdown 文件"
-                  : "开启写作模式：最终输出保存为 Markdown 文件并进入画布"
-              }
-            >
-              <PromptInputButton
-                className={cn(
-                  "gap-1.5! rounded-full px-3! shadow-none",
-                  compact && "rounded-lg px-2.5! py-1! text-[11px] font-medium",
-                  context.writing_mode
-                    ? "border border-amber-300 bg-amber-50 text-amber-700 hover:bg-amber-100 dark:border-amber-700 dark:bg-amber-950/40 dark:text-amber-200"
-                    : compact
-                      ? "border border-transparent bg-muted/40 text-muted-foreground hover:bg-muted hover:text-foreground"
-                      : "border-0 bg-transparent text-muted-foreground hover:bg-muted/70 hover:text-foreground dark:text-slate-400 dark:hover:bg-slate-700/70 dark:hover:text-slate-100",
-                )}
-                onClick={handleWritingModeToggle}
-              >
-                <FileTextIcon className="size-3" />
-                <span className="text-xs font-normal">写作</span>
-              </PromptInputButton>
-            </Tooltip>
-            <Tooltip
-              content={
-                memoryInjectionEnabled
-                  ? "记忆注入已开启：对话前会注入历史记忆"
-                  : "记忆注入已关闭：本次对话不注入记忆"
-              }
-            >
-              <PromptInputButton
-                className={cn(
-                  "gap-1.5! rounded-full px-3! shadow-none",
-                  compact && "rounded-lg px-2.5! py-1! text-[11px] font-medium",
-                  memoryInjectionEnabled
-                    ? "border border-blue-300 bg-blue-50 text-blue-700 hover:bg-blue-100 dark:border-blue-700 dark:bg-blue-950/40 dark:text-blue-200"
-                    : compact
-                      ? "border border-transparent bg-muted/40 text-muted-foreground hover:bg-muted hover:text-foreground"
-                      : "border-0 bg-transparent text-muted-foreground hover:bg-muted/70 hover:text-foreground dark:text-slate-400 dark:hover:bg-slate-700/70 dark:hover:text-slate-100",
-                )}
-                onClick={handleMemoryInjectionToggle}
-              >
-                <BrainCircuitIcon className="size-3" />
-                <span className="text-xs font-normal">记忆</span>
-              </PromptInputButton>
-            </Tooltip>
-            <Tooltip
-              content={
-                ragModeEnabled
-                  ? "参考文献模式已开启：检索类结果会按编号引用展示"
-                  : "参考文献模式已关闭：点击开启"
-              }
-            >
-              <PromptInputButton
-                type="button"
-                aria-pressed={ragModeEnabled}
-                disabled={ragModeLocked || disabled}
-                className={cn(
-                  "gap-1.5! rounded-full px-3! shadow-none",
-                  compact && "rounded-lg px-2.5! py-1! text-[11px] font-medium",
-                  ragModeLocked && "cursor-not-allowed opacity-80",
-                  ragModeEnabled
-                    ? "border border-emerald-300 bg-emerald-50 text-emerald-700 hover:bg-emerald-100 dark:border-emerald-700 dark:bg-emerald-950/40 dark:text-emerald-200"
-                    : compact
-                      ? "border border-transparent bg-muted/40 text-muted-foreground hover:bg-muted hover:text-foreground"
-                      : "border-0 bg-transparent text-muted-foreground hover:bg-muted/70 hover:text-foreground dark:text-slate-400 dark:hover:bg-slate-700/70 dark:hover:text-slate-100",
-                )}
-                onClick={handleRagModeToggle}
-              >
-                {ragModeEnabled && (
-                  <CheckIcon className="size-3 text-emerald-700 dark:text-emerald-200" />
-                )}
-                <BookMarkedIcon className="size-3" />
-                <span className="text-xs font-normal">参考文献</span>
-              </PromptInputButton>
-            </Tooltip>
-            <PromptInputActionMenu>
-              <ModeHoverGuide
-                mode={
-                  context.mode === "flash" ||
-                  context.mode === "thinking" ||
-                  context.mode === "pro" ||
-                  context.mode === "ultra"
-                    ? context.mode
-                    : "flash"
-                }
-              >
-                <PromptInputActionMenuTrigger
-                  className={cn(
-                    "gap-1.5! rounded-full border-0 bg-transparent px-3! text-muted-foreground shadow-none hover:bg-muted/70 hover:text-foreground dark:text-slate-400 dark:hover:bg-slate-700/70 dark:hover:text-slate-100",
-                    compact && "rounded-lg border px-2.5! py-1! text-[11px] font-medium",
-                    compact &&
-                      (context.mode === "thinking"
-                        ? "border-indigo-200 bg-indigo-50/60 text-indigo-700 hover:bg-indigo-100 dark:border-indigo-700 dark:bg-indigo-950/40 dark:text-indigo-200"
-                        : context.mode === "pro"
-                          ? "border-violet-200 bg-violet-50/60 text-violet-700 hover:bg-violet-100 dark:border-violet-700 dark:bg-violet-950/40 dark:text-violet-200"
-                          : context.mode === "ultra"
-                            ? "border-amber-200 bg-amber-50/60 text-amber-700 hover:bg-amber-100 dark:border-amber-700 dark:bg-amber-950/40 dark:text-amber-200"
-                            : "border-transparent bg-muted/40 hover:bg-muted hover:text-foreground"),
-                  )}
-                >
-                  <div>
-                    {context.mode === "flash" && <ZapIcon className="size-3" />}
-                    {context.mode === "thinking" && (
-                      <LightbulbIcon className="size-3" />
-                    )}
-                    {context.mode === "pro" && (
-                      <GraduationCapIcon className="size-3" />
-                    )}
-                    {context.mode === "ultra" && (
-                      <RocketIcon className="size-3 text-[#dabb5e]" />
-                    )}
-                  </div>
-                  <div
-                    className={cn(
-                      "text-xs font-normal",
-                      context.mode === "ultra" ? "golden-text" : "",
-                    )}
-                  >
-                    {(context.mode === "flash" && t.inputBox.flashMode) ||
-                      (context.mode === "thinking" &&
-                        t.inputBox.reasoningMode) ||
-                      (context.mode === "pro" && t.inputBox.proMode) ||
-                      (context.mode === "ultra" && t.inputBox.ultraMode)}
-                  </div>
-                </PromptInputActionMenuTrigger>
-              </ModeHoverGuide>
-              <PromptInputActionMenuContent className="w-80">
-                <DropdownMenuGroup>
-                  <DropdownMenuLabel className="text-muted-foreground text-xs">
-                    {t.inputBox.mode}
-                  </DropdownMenuLabel>
-                  <PromptInputActionMenu>
-                    <PromptInputActionMenuItem
-                      className={cn(
-                        context.mode === "flash"
-                          ? "text-accent-foreground"
-                          : "text-muted-foreground/65",
-                      )}
-                      onSelect={() => handleModeSelect("flash")}
-                    >
-                      <div className="flex flex-col gap-2">
-                        <div className="flex items-center gap-1 font-bold">
-                          <ZapIcon
-                            className={cn(
-                              "mr-2 size-4",
-                              context.mode === "flash" &&
-                                "text-accent-foreground",
-                            )}
-                          />
-                          {t.inputBox.flashMode}
-                        </div>
-                        <div className="pl-7 text-xs">
-                          {t.inputBox.flashModeDescription}
-                        </div>
-                      </div>
-                      {context.mode === "flash" ? (
-                        <CheckIcon className="ml-auto size-4" />
-                      ) : (
-                        <div className="ml-auto size-4" />
-                      )}
-                    </PromptInputActionMenuItem>
-                    {supportThinking && (
-                      <PromptInputActionMenuItem
-                        className={cn(
-                          context.mode === "thinking"
-                            ? "text-accent-foreground"
-                            : "text-muted-foreground/65",
-                        )}
-                        onSelect={() => handleModeSelect("thinking")}
-                      >
-                        <div className="flex flex-col gap-2">
-                          <div className="flex items-center gap-1 font-bold">
-                            <LightbulbIcon
-                              className={cn(
-                                "mr-2 size-4",
-                                context.mode === "thinking" &&
-                                  "text-accent-foreground",
-                              )}
-                            />
-                            {t.inputBox.reasoningMode}
-                          </div>
-                          <div className="pl-7 text-xs">
-                            {t.inputBox.reasoningModeDescription}
-                          </div>
-                        </div>
-                        {context.mode === "thinking" ? (
-                          <CheckIcon className="ml-auto size-4" />
-                        ) : (
-                          <div className="ml-auto size-4" />
-                        )}
-                      </PromptInputActionMenuItem>
-                    )}
-                    <PromptInputActionMenuItem
-                      className={cn(
-                        context.mode === "pro"
-                          ? "text-accent-foreground"
-                          : "text-muted-foreground/65",
-                      )}
-                      onSelect={() => handleModeSelect("pro")}
-                    >
-                      <div className="flex flex-col gap-2">
-                        <div className="flex items-center gap-1 font-bold">
-                          <GraduationCapIcon
-                            className={cn(
-                              "mr-2 size-4",
-                              context.mode === "pro" &&
-                                "text-accent-foreground",
-                            )}
-                          />
-                          {t.inputBox.proMode}
-                        </div>
-                        <div className="pl-7 text-xs">
-                          {t.inputBox.proModeDescription}
-                        </div>
-                      </div>
-                      {context.mode === "pro" ? (
-                        <CheckIcon className="ml-auto size-4" />
-                      ) : (
-                        <div className="ml-auto size-4" />
-                      )}
-                    </PromptInputActionMenuItem>
-                    <PromptInputActionMenuItem
-                      className={cn(
-                        context.mode === "ultra"
-                          ? "text-accent-foreground"
-                          : "text-muted-foreground/65",
-                      )}
-                      onSelect={() => handleModeSelect("ultra")}
-                    >
-                      <div className="flex flex-col gap-2">
-                        <div className="flex items-center gap-1 font-bold">
-                          <RocketIcon
-                            className={cn(
-                              "mr-2 size-4",
-                              context.mode === "ultra" && "text-[#dabb5e]",
-                            )}
-                          />
-                          <div
-                            className={cn(
-                              context.mode === "ultra" && "golden-text",
-                            )}
-                          >
-                            {t.inputBox.ultraMode}
-                          </div>
-                        </div>
-                        <div className="pl-7 text-xs">
-                          {t.inputBox.ultraModeDescription}
-                        </div>
-                      </div>
-                      {context.mode === "ultra" ? (
-                        <CheckIcon className="ml-auto size-4" />
-                      ) : (
-                        <div className="ml-auto size-4" />
-                      )}
-                    </PromptInputActionMenuItem>
-                  </PromptInputActionMenu>
-                </DropdownMenuGroup>
-              </PromptInputActionMenuContent>
-            </PromptInputActionMenu>
-            {supportReasoningEffort && context.mode !== "flash" && (
-              <PromptInputActionMenu>
-                <PromptInputActionMenuTrigger className={cn("gap-1.5! rounded-full border-0 bg-transparent px-3! text-muted-foreground shadow-none hover:bg-muted/70 hover:text-foreground dark:text-slate-400 dark:hover:bg-slate-700/70 dark:hover:text-slate-100", compact && "rounded-lg border border-border px-2.5! py-1! text-[13px]")}>
-                  <div className="text-xs font-normal">
-                    {t.inputBox.reasoningEffort}:
-                    {context.reasoning_effort === "minimal" &&
-                      " " + t.inputBox.reasoningEffortMinimal}
-                    {context.reasoning_effort === "low" &&
-                      " " + t.inputBox.reasoningEffortLow}
-                    {context.reasoning_effort === "medium" &&
-                      " " + t.inputBox.reasoningEffortMedium}
-                    {context.reasoning_effort === "high" &&
-                      " " + t.inputBox.reasoningEffortHigh}
-                  </div>
-                </PromptInputActionMenuTrigger>
-                <PromptInputActionMenuContent className="w-70">
-                  <DropdownMenuGroup>
-                    <DropdownMenuLabel className="text-muted-foreground text-xs">
-                      {t.inputBox.reasoningEffort}
-                    </DropdownMenuLabel>
-                    <PromptInputActionMenu>
-                      <PromptInputActionMenuItem
-                        className={cn(
-                          context.reasoning_effort === "minimal"
-                            ? "text-accent-foreground"
-                            : "text-muted-foreground/65",
-                        )}
-                        onSelect={() => handleReasoningEffortSelect("minimal")}
-                      >
-                        <div className="flex flex-col gap-2">
-                          <div className="flex items-center gap-1 font-bold">
-                            {t.inputBox.reasoningEffortMinimal}
-                          </div>
-                          <div className="pl-2 text-xs">
-                            {t.inputBox.reasoningEffortMinimalDescription}
-                          </div>
-                        </div>
-                        {context.reasoning_effort === "minimal" ? (
-                          <CheckIcon className="ml-auto size-4" />
-                        ) : (
-                          <div className="ml-auto size-4" />
-                        )}
-                      </PromptInputActionMenuItem>
-                      <PromptInputActionMenuItem
-                        className={cn(
-                          context.reasoning_effort === "low"
-                            ? "text-accent-foreground"
-                            : "text-muted-foreground/65",
-                        )}
-                        onSelect={() => handleReasoningEffortSelect("low")}
-                      >
-                        <div className="flex flex-col gap-2">
-                          <div className="flex items-center gap-1 font-bold">
-                            {t.inputBox.reasoningEffortLow}
-                          </div>
-                          <div className="pl-2 text-xs">
-                            {t.inputBox.reasoningEffortLowDescription}
-                          </div>
-                        </div>
-                        {context.reasoning_effort === "low" ? (
-                          <CheckIcon className="ml-auto size-4" />
-                        ) : (
-                          <div className="ml-auto size-4" />
-                        )}
-                      </PromptInputActionMenuItem>
-                      <PromptInputActionMenuItem
-                        className={cn(
-                          context.reasoning_effort === "medium" ||
-                            !context.reasoning_effort
-                            ? "text-accent-foreground"
-                            : "text-muted-foreground/65",
-                        )}
-                        onSelect={() => handleReasoningEffortSelect("medium")}
-                      >
-                        <div className="flex flex-col gap-2">
-                          <div className="flex items-center gap-1 font-bold">
-                            {t.inputBox.reasoningEffortMedium}
-                          </div>
-                          <div className="pl-2 text-xs">
-                            {t.inputBox.reasoningEffortMediumDescription}
-                          </div>
-                        </div>
-                        {context.reasoning_effort === "medium" ||
-                        !context.reasoning_effort ? (
-                          <CheckIcon className="ml-auto size-4" />
-                        ) : (
-                          <div className="ml-auto size-4" />
-                        )}
-                      </PromptInputActionMenuItem>
-                      <PromptInputActionMenuItem
-                        className={cn(
-                          context.reasoning_effort === "high"
-                            ? "text-accent-foreground"
-                            : "text-muted-foreground/65",
-                        )}
-                        onSelect={() => handleReasoningEffortSelect("high")}
-                      >
-                        <div className="flex flex-col gap-2">
-                          <div className="flex items-center gap-1 font-bold">
-                            {t.inputBox.reasoningEffortHigh}
-                          </div>
-                          <div className="pl-2 text-xs">
-                            {t.inputBox.reasoningEffortHighDescription}
-                          </div>
-                        </div>
-                        {context.reasoning_effort === "high" ? (
-                          <CheckIcon className="ml-auto size-4" />
-                        ) : (
-                          <div className="ml-auto size-4" />
-                        )}
-                      </PromptInputActionMenuItem>
-                    </PromptInputActionMenu>
-                  </DropdownMenuGroup>
-                </PromptInputActionMenuContent>
-              </PromptInputActionMenu>
-            )}
-          </PromptInputTools>
-          <PromptInputTools className={cn("gap-1.5", compact && "gap-2")}>
-            {agentSelector && (
-              <AgentSelector
-                selectedAgentId={agentSelector.selectedAgentId}
-                onAgentChange={agentSelector.onAgentChange}
-                disabled={disabled}
-                compact={compact}
-              />
-            )}
-            {promptPrefix && (
-              <Tooltip
-                content={
-                  promptPrefix.on
-                    ? `已开启：${promptPrefix.label}`
-                    : `点击开启：${promptPrefix.label}`
-                }
-              >
-                <PromptInputButton
-                  type="button"
-                  onClick={promptPrefix.onToggle}
-                  aria-pressed={promptPrefix.on}
-                  className={cn(
-                    "rounded-full px-3 shadow-none transition-colors",
-                    promptPrefix.on
-                      ? "border border-amber-300 bg-amber-50 text-amber-700 hover:bg-amber-100 dark:border-amber-700 dark:bg-amber-950/40 dark:text-amber-200"
-                      : "border-0 bg-transparent text-muted-foreground hover:bg-muted/70 hover:text-foreground",
-                  )}
-                >
-                  <span className="max-w-[160px] truncate text-xs font-normal">
-                    {promptPrefix.label}
-                  </span>
-                  {promptPrefix.on && (
-                    <CheckIcon className="ml-1 size-3.5 text-amber-700 dark:text-amber-200" />
-                  )}
-                </PromptInputButton>
-              </Tooltip>
-            )}
-            <ModelSelector
-              open={modelDialogOpen}
-              onOpenChange={setModelDialogOpen}
-            >
-              <ModelSelectorTrigger asChild>
-                <PromptInputButton
-                  className={cn(
-                    "rounded-full border-0 bg-transparent px-3 shadow-none text-muted-foreground hover:bg-muted/70 hover:text-foreground dark:text-slate-400 dark:hover:bg-slate-700/70 dark:hover:text-slate-100",
-                    compact &&
-                      "rounded-lg border border-transparent bg-gray-50/60 px-2.5! py-1! font-mono text-gray-600 hover:bg-gray-100/80 hover:text-gray-800 dark:bg-slate-700/50 dark:text-slate-300 dark:hover:bg-slate-700",
-                  )}
-                >
-                  <div className="flex min-w-0 flex-col items-start text-left">
-                    <ModelSelectorName
-                      className={cn(
-                        "text-xs font-normal",
-                        compact && "text-[11px]",
-                      )}
-                    >
-                      {selectedModel?.display_name}
-                    </ModelSelectorName>
-                  </div>
-                </PromptInputButton>
-              </ModelSelectorTrigger>
-              <ModelSelectorContent>
-                <ModelSelectorInput placeholder={t.inputBox.searchModels} />
-                <ModelSelectorList>
-                  {models.map((m) => (
-                    <ModelSelectorItem
-                      key={m.name}
-                      value={m.name}
-                      onSelect={() => handleModelSelect(m.name)}
-                    >
-                      <div className="flex min-w-0 flex-1 flex-col">
-                        <ModelSelectorName>{m.display_name}</ModelSelectorName>
-                        <span className="text-muted-foreground truncate text-[10px]">
-                          {m.model}
-                        </span>
-                      </div>
-                      {m.name === context.model_name ? (
-                        <CheckIcon className="ml-auto size-4" />
-                      ) : (
-                        <div className="ml-auto size-4" />
-                      )}
-                    </ModelSelectorItem>
-                  ))}
-                </ModelSelectorList>
-              </ModelSelectorContent>
-            </ModelSelector>
-            <PromptInputSubmit
-              className={cn(
-                "rounded-full border-border bg-background shadow-sm hover:bg-muted dark:bg-blue-500 dark:border-blue-500 dark:text-white dark:hover:bg-blue-400 dark:shadow-none",
-                compact &&
-                  "size-7 rounded-lg border-0 bg-[var(--primary)] text-[var(--primary-foreground)] hover:bg-[color-mix(in_srgb,var(--primary)_88%,black)] disabled:bg-muted disabled:text-muted-foreground/40 dark:bg-[var(--primary)] dark:border-0 dark:text-[var(--primary-foreground)] dark:hover:bg-[color-mix(in_srgb,var(--primary)_88%,black)] dark:disabled:bg-muted",
-              )}
-              disabled={disabled}
-              variant="outline"
-              status={status}
-            />
-          </PromptInputTools>
-        </PromptInputFooter>
-        {!isNewThread && (
-          <div className="absolute right-0 -bottom-[17px] left-0 z-0 h-4 bg-background"></div>
-        )}
-      </PromptInput>
-      </div>
-      </div>
-
-      <Dialog open={confirmOpen} onOpenChange={setConfirmOpen}>
-        <DialogContent>
-          <DialogHeader>
-            <DialogTitle>{t.inputBox.followupConfirmTitle}</DialogTitle>
-            <DialogDescription>
-              {t.inputBox.followupConfirmDescription}
-            </DialogDescription>
-          </DialogHeader>
-          <DialogFooter>
-            <Button variant="outline" onClick={() => setConfirmOpen(false)}>
-              {t.common.cancel}
-            </Button>
-            <Button variant="secondary" onClick={confirmAppendAndSend}>
-              {t.inputBox.followupConfirmAppend}
-            </Button>
-            <Button onClick={confirmReplaceAndSend}>
-              {t.inputBox.followupConfirmReplace}
-            </Button>
-          </DialogFooter>
-        </DialogContent>
-      </Dialog>
-    </div>
-  );
-}
-
-function SuggestionList() {
-  const { t } = useI18n();
-  const { textInput } = usePromptInputController();
-  const handleSuggestionClick = useCallback(
-    (prompt: string | undefined) => {
-      if (!prompt) return;
-      textInput.setInput(prompt);
-      setTimeout(() => {
-        const textarea = document.querySelector<HTMLTextAreaElement>(
-          "textarea[name='message']",
-        );
-        if (textarea) {
-          const selStart = prompt.indexOf("[");
-          const selEnd = prompt.indexOf("]");
-          if (selStart !== -1 && selEnd !== -1) {
-            textarea.setSelectionRange(selStart, selEnd + 1);
-            textarea.focus();
-          }
-        }
-      }, 500);
-    },
-    [textInput],
-  );
-  return (
-    <Suggestions className="min-h-16 w-fit items-start">
-      <ConfettiButton
-        className="text-muted-foreground cursor-pointer rounded-full px-4 text-xs font-normal"
-        variant="outline"
-        size="sm"
-        onClick={() => handleSuggestionClick(t.inputBox.surpriseMePrompt)}
-      >
-        <SparklesIcon className="size-4" /> {t.inputBox.surpriseMe}
-      </ConfettiButton>
-      {t.inputBox.suggestions.map((suggestion) => (
-        <Suggestion
-          key={suggestion.suggestion}
-          icon={suggestion.icon}
-          suggestion={suggestion.suggestion}
-          onClick={() => handleSuggestionClick(suggestion.prompt)}
-        />
-      ))}
-      <DropdownMenu>
-        <DropdownMenuTrigger asChild>
-          <Suggestion icon={PlusIcon} suggestion={t.common.create} />
-        </DropdownMenuTrigger>
-        <DropdownMenuContent align="start">
-          <DropdownMenuGroup>
-            {t.inputBox.suggestionsCreate.map((suggestion, index) =>
-              "type" in suggestion && suggestion.type === "separator" ? (
-                <DropdownMenuSeparator key={index} />
-              ) : (
-                !("type" in suggestion) && (
-                  <DropdownMenuItem
-                    key={suggestion.suggestion}
-                    onClick={() => handleSuggestionClick(suggestion.prompt)}
-                  >
-                    {suggestion.icon && <suggestion.icon className="size-4" />}
-                    {suggestion.suggestion}
-                  </DropdownMenuItem>
-                )
-              ),
-            )}
-          </DropdownMenuGroup>
-        </DropdownMenuContent>
-      </DropdownMenu>
-    </Suggestions>
-  );
-}
-
-function AddAttachmentsButton({ className }: { className?: string }) {
-  const { t } = useI18n();
-  const attachments = usePromptInputAttachments();
-  return (
-    <Tooltip content={t.inputBox.addAttachments}>
-      <PromptInputButton
-        className={cn("px-2!", className)}
-        onClick={() => attachments.openFileDialog()}
-      >
-        <PaperclipIcon className="size-3" />
-      </PromptInputButton>
-    </Tooltip>
-  );
-}
+"use client";
+
+import type { ChatStatus } from "ai";
+import {
+  BrainCircuitIcon,
+  BookMarkedIcon,
+  CheckIcon,
+  FileTextIcon,
+  GraduationCapIcon,
+  LightbulbIcon,
+  PaperclipIcon,
+  PlusIcon,
+  ShrinkIcon,
+  SparklesIcon,
+  RocketIcon,
+  XIcon,
+  ZapIcon,
+} from "lucide-react";
+import { useSearchParams } from "next/navigation";
+import {
+  useCallback,
+  useEffect,
+  useMemo,
+  useRef,
+  useState,
+  type ComponentProps,
+} from "react";
+
+import {
+  PromptInput,
+  PromptInputActionMenu,
+  PromptInputActionMenuContent,
+  PromptInputActionMenuItem,
+  PromptInputActionMenuTrigger,
+  PromptInputAttachment,
+  PromptInputAttachments,
+  PromptInputBody,
+  PromptInputButton,
+  PromptInputFooter,
+  PromptInputSubmit,
+  PromptInputTextarea,
+  PromptInputTools,
+  usePromptInputAttachments,
+  usePromptInputController,
+  type PromptInputMessage,
+} from "@/components/ai-elements/prompt-input";
+import { Button } from "@/components/ui/button";
+import { ConfettiButton } from "@/components/ui/confetti-button";
+import {
+  Dialog,
+  DialogContent,
+  DialogDescription,
+  DialogFooter,
+  DialogHeader,
+  DialogTitle,
+} from "@/components/ui/dialog";
+import {
+  DropdownMenuGroup,
+  DropdownMenuLabel,
+  DropdownMenuSeparator,
+} from "@/components/ui/dropdown-menu";
+import { apiFetch } from "@/core/api";
+import { getBackendBaseURL } from "@/core/config";
+import { useI18n } from "@/core/i18n/hooks";
+import { useModels } from "@/core/models/hooks";
+import type { AgentThreadContext } from "@/core/threads";
+import { textOfMessage } from "@/core/threads/utils";
+import { desensitize } from "@/lib/desensitize";
+import { cn } from "@/lib/utils";
+
+import { AgentSelector } from "@/components/workspace/agents/agent-selector";
+
+import {
+  ModelSelector,
+  ModelSelectorContent,
+  ModelSelectorInput,
+  ModelSelectorItem,
+  ModelSelectorList,
+  ModelSelectorName,
+  ModelSelectorTrigger,
+} from "../ai-elements/model-selector";
+import { Suggestion, Suggestions } from "../ai-elements/suggestion";
+import {
+  DropdownMenu,
+  DropdownMenuContent,
+  DropdownMenuItem,
+  DropdownMenuTrigger,
+} from "../ui/dropdown-menu";
+
+import { useRecommendedQuestions } from "@/core/recommended-questions/hooks";
+import { useUserPreferences } from "@/core/user-preferences";
+import {
+  isThreadRagModeLocked,
+  lockThreadRagMode,
+  useLocalSettings,
+} from "@/core/settings";
+
+
+import { CustomPromptBar } from "./custom-prompt-menu";
+import { useThread } from "./messages/context";
+import { ModeHoverGuide } from "./mode-hover-guide";
+import { Tooltip } from "./tooltip";
+
+type InputMode = "flash" | "thinking" | "pro" | "ultra";
+
+
+function getResolvedMode(
+  mode: InputMode | undefined,
+  supportsThinking: boolean,
+): InputMode {
+  if (!supportsThinking && mode !== "flash") {
+    return "flash";
+  }
+  if (mode) {
+    return mode;
+  }
+  return supportsThinking ? "pro" : "flash";
+}
+
+export function InputBox({
+  className,
+  disabled,
+  autoFocus,
+  status = "ready",
+  context,
+  extraHeader,
+  isNewThread,
+  threadId,
+  initialValue,
+  onContextChange,
+  onFollowupsVisibilityChange,
+  onSubmit,
+  onStop,
+  promptPrefix,
+  agentSelector,
+  assistantId,
+  hideAdminRecommendedQuestions,
+  compact,
+  ...props
+}: Omit<ComponentProps<typeof PromptInput>, "onSubmit"> & {
+  assistantId?: string | null;
+  status?: ChatStatus;
+  /** 紧凑模式（正常 App）新对话落地态：去掉旋转彩边、用 CM 参考图的白卡输入框，
+   *  并在推荐问题上方加「试试这些高频业务场景」标题。仅落地态生效，默认 false。 */
+  compact?: boolean;
+  disabled?: boolean;
+  context: Omit<
+    AgentThreadContext,
+    "thread_id" | "is_plan_mode" | "thinking_enabled" | "subagent_enabled"
+  > & {
+    mode: "flash" | "thinking" | "pro" | "ultra" | undefined;
+    reasoning_effort?: "minimal" | "low" | "medium" | "high";
+    writing_mode?: boolean;
+    memory_injection_enabled?: boolean;
+  };
+  extraHeader?: React.ReactNode;
+  isNewThread?: boolean;
+  /** Suppress the global admin "recommended questions" row. The agent chat page
+   * sets this so only the agent's own common questions surface on its home. */
+  hideAdminRecommendedQuestions?: boolean;
+  threadId: string;
+  initialValue?: string;
+  onContextChange?: (
+    context: Omit<
+      AgentThreadContext,
+      "thread_id" | "is_plan_mode" | "thinking_enabled" | "subagent_enabled"
+    > & {
+      mode: "flash" | "thinking" | "pro" | "ultra" | undefined;
+      reasoning_effort?: "minimal" | "low" | "medium" | "high";
+      writing_mode?: boolean;
+      memory_injection_enabled?: boolean;
+    },
+  ) => void;
+  onFollowupsVisibilityChange?: (visible: boolean) => void;
+  onSubmit?: (message: PromptInputMessage) => void;
+  onStop?: () => void;
+  /**
+   * Optional toggle rendered next to the model selector. Used for the
+   * admin-configurable prompt-prefix switch on the chat input. Omitting the
+   * prop hides the button entirely.
+   */
+  promptPrefix?: {
+    label: string;
+    on: boolean;
+    onToggle: () => void;
+  };
+  /**
+   * Optional agent picker shown alongside the model selector. Only the new-chat
+   * homepage (``ChatPage``) passes this; ``AgentChatPage`` omits it because the
+   * thread is already bound to one agent.
+   */
+  agentSelector?: {
+    selectedAgentId: string | null;
+    onAgentChange: (agentId: string | null) => void;
+  };
+}) {
+  const { t } = useI18n();
+  const [localSettings, setLocalSettings] = useLocalSettings();
+  const searchParams = useSearchParams();
+  const [modelDialogOpen, setModelDialogOpen] = useState(false);
+  const { models } = useModels();
+  const { thread, isMock } = useThread();
+  const { textInput } = usePromptInputController();
+  const promptRootRef = useRef<HTMLDivElement | null>(null);
+
+  const [followups, setFollowups] = useState<string[]>([]);
+  const [followupsHidden, setFollowupsHidden] = useState(false);
+  const [followupsLoading, setFollowupsLoading] = useState(false);
+  const lastGeneratedForAiIdRef = useRef<string | null>(null);
+  const wasStreamingRef = useRef(false);
+
+  const [confirmOpen, setConfirmOpen] = useState(false);
+  const [pendingSuggestion, setPendingSuggestion] = useState<string | null>(
+    null,
+  );
+
+  const { questions: recommendedQuestions } = useRecommendedQuestions();
+  const userPreferences = useUserPreferences();
+  const showAdminRecommendedQuestions =
+    !hideAdminRecommendedQuestions &&
+    !userPreferences.hide_admin_prompt_prefix &&
+    recommendedQuestions.length > 0;
+
+  useEffect(() => {
+    if (models.length === 0) {
+      return;
+    }
+    const currentModel = models.find((m) => m.name === context.model_name);
+    const fallbackModel = currentModel ?? models[0]!;
+    const supportsThinking = fallbackModel.supports_thinking ?? false;
+    const nextModelName = fallbackModel.name;
+    const nextMode = getResolvedMode(context.mode, supportsThinking);
+
+    if (context.model_name === nextModelName && context.mode === nextMode) {
+      return;
+    }
+
+    onContextChange?.({
+      ...context,
+      model_name: nextModelName,
+      mode: nextMode,
+    });
+  }, [context, models, onContextChange]);
+
+  const selectedModel = useMemo(() => {
+    if (models.length === 0) {
+      return undefined;
+    }
+    return models.find((m) => m.name === context.model_name) ?? models[0];
+  }, [context.model_name, models]);
+
+  const resolvedModelName = selectedModel?.name;
+
+  const supportThinking = useMemo(
+    () => selectedModel?.supports_thinking ?? false,
+    [selectedModel],
+  );
+
+  const supportReasoningEffort = useMemo(
+    () => selectedModel?.supports_reasoning_effort ?? false,
+    [selectedModel],
+  );
+
+  const handleModelSelect = useCallback(
+    (model_name: string) => {
+      const model = models.find((m) => m.name === model_name);
+      if (!model) {
+        return;
+      }
+      onContextChange?.({
+        ...context,
+        model_name,
+        mode: getResolvedMode(context.mode, model.supports_thinking ?? false),
+        reasoning_effort: context.reasoning_effort,
+      });
+      setModelDialogOpen(false);
+    },
+    [onContextChange, context, models],
+  );
+
+  const handleModeSelect = useCallback(
+    (mode: InputMode) => {
+      onContextChange?.({
+        ...context,
+        mode: getResolvedMode(mode, supportThinking),
+        reasoning_effort:
+          mode === "ultra"
+            ? "high"
+            : mode === "pro"
+              ? "medium"
+              : mode === "thinking"
+                ? "low"
+                : "minimal",
+      });
+    },
+    [onContextChange, context, supportThinking],
+  );
+
+  const handleReasoningEffortSelect = useCallback(
+    (effort: "minimal" | "low" | "medium" | "high") => {
+      onContextChange?.({
+        ...context,
+        reasoning_effort: effort,
+      });
+    },
+    [onContextChange, context],
+  );
+
+  const handleWritingModeToggle = useCallback(() => {
+    onContextChange?.({
+      ...context,
+      writing_mode: !context.writing_mode,
+    });
+  }, [context, onContextChange]);
+
+  const memoryInjectionEnabled = context.memory_injection_enabled ?? true;
+  const handleMemoryInjectionToggle = useCallback(() => {
+    onContextChange?.({
+      ...context,
+      memory_injection_enabled: !memoryInjectionEnabled,
+    });
+  }, [context, memoryInjectionEnabled, onContextChange]);
+
+  const ragModeLocked = isThreadRagModeLocked(threadId);
+  const ragModeEnabled = localSettings.display.ragModeEnabled || ragModeLocked;
+  const handleRagModeToggle = useCallback(() => {
+    if (ragModeLocked) {
+      return;
+    }
+    setLocalSettings("display", { ragModeEnabled: !ragModeEnabled });
+  }, [ragModeEnabled, ragModeLocked, setLocalSettings]);
+
+  const contextCompressionEnabled =
+    localSettings.display.contextCompressionEnabled ?? false;
+  const handleContextCompressionToggle = useCallback(() => {
+    setLocalSettings("display", {
+      contextCompressionEnabled: !contextCompressionEnabled,
+    });
+  }, [contextCompressionEnabled, setLocalSettings]);
+
+  const handleSubmit = useCallback(
+    async (message: PromptInputMessage) => {
+      if (status === "streaming") {
+        onStop?.();
+        return;
+      }
+      if (!message.text) {
+        return;
+      }
+      setFollowups([]);
+      setFollowupsHidden(false);
+      setFollowupsLoading(false);
+      if (ragModeEnabled) {
+        lockThreadRagMode(threadId);
+      }
+
+      // Guard against submitting before the initial model auto-selection
+      // effect has flushed thread settings to storage/state.
+      if (resolvedModelName && context.model_name !== resolvedModelName) {
+        onContextChange?.({
+          ...context,
+          model_name: resolvedModelName,
+          mode: getResolvedMode(
+            context.mode,
+            selectedModel?.supports_thinking ?? false,
+          ),
+        });
+        setTimeout(() => onSubmit?.(message), 0);
+        return;
+      }
+
+      onSubmit?.(message);
+    },
+    [
+      context,
+      onContextChange,
+      onSubmit,
+      onStop,
+      resolvedModelName,
+      ragModeEnabled,
+      selectedModel?.supports_thinking,
+      status,
+      threadId,
+    ],
+  );
+
+  const requestFormSubmit = useCallback(() => {
+    const form = promptRootRef.current?.querySelector("form");
+    form?.requestSubmit();
+  }, []);
+
+  const handleFollowupClick = useCallback(
+    (suggestion: string) => {
+      if (status === "streaming") {
+        return;
+      }
+      const current = (textInput.value ?? "").trim();
+      if (current) {
+        setPendingSuggestion(suggestion);
+        setConfirmOpen(true);
+        return;
+      }
+      textInput.setInput(suggestion);
+      setFollowupsHidden(true);
+      setTimeout(() => requestFormSubmit(), 0);
+    },
+    [requestFormSubmit, status, textInput],
+  );
+
+  const confirmReplaceAndSend = useCallback(() => {
+    if (!pendingSuggestion) {
+      setConfirmOpen(false);
+      return;
+    }
+    textInput.setInput(pendingSuggestion);
+    setFollowupsHidden(true);
+    setConfirmOpen(false);
+    setPendingSuggestion(null);
+    setTimeout(() => requestFormSubmit(), 0);
+  }, [pendingSuggestion, requestFormSubmit, textInput]);
+
+  const confirmAppendAndSend = useCallback(() => {
+    if (!pendingSuggestion) {
+      setConfirmOpen(false);
+      return;
+    }
+    const current = (textInput.value ?? "").trim();
+    const next = current
+      ? `${current}\n${pendingSuggestion}`
+      : pendingSuggestion;
+    textInput.setInput(next);
+    setFollowupsHidden(true);
+    setConfirmOpen(false);
+    setPendingSuggestion(null);
+    setTimeout(() => requestFormSubmit(), 0);
+  }, [pendingSuggestion, requestFormSubmit, textInput]);
+
+  const showFollowups =
+    !disabled &&
+    !isNewThread &&
+    !followupsHidden &&
+    (followupsLoading || followups.length > 0);
+
+  const followupsVisibilityChangeRef = useRef(onFollowupsVisibilityChange);
+
+  useEffect(() => {
+    followupsVisibilityChangeRef.current = onFollowupsVisibilityChange;
+  }, [onFollowupsVisibilityChange]);
+
+  useEffect(() => {
+    followupsVisibilityChangeRef.current?.(showFollowups);
+  }, [showFollowups]);
+
+  useEffect(() => {
+    return () => followupsVisibilityChangeRef.current?.(false);
+  }, []);
+
+  useEffect(() => {
+    const streaming = status === "streaming";
+    const wasStreaming = wasStreamingRef.current;
+    wasStreamingRef.current = streaming;
+    if (!wasStreaming || streaming) {
+      return;
+    }
+
+    if (disabled || isMock) {
+      return;
+    }
+
+    const lastAi = [...thread.messages].reverse().find((m) => m.type === "ai");
+    const lastAiId = lastAi?.id ?? null;
+    if (!lastAiId || lastAiId === lastGeneratedForAiIdRef.current) {
+      return;
+    }
+    lastGeneratedForAiIdRef.current = lastAiId;
+
+    const recent = thread.messages
+      .filter((m) => m.type === "human" || m.type === "ai")
+      .map((m) => {
+        const role = m.type === "human" ? "user" : "assistant";
+        const content = textOfMessage(m) ?? "";
+        return { role, content };
+      })
+      .filter((m) => m.content.trim().length > 0)
+      .slice(-6);
+
+    if (recent.length === 0) {
+      return;
+    }
+
+    const controller = new AbortController();
+    setFollowupsHidden(false);
+    setFollowupsLoading(true);
+    setFollowups([]);
+
+    apiFetch(`${getBackendBaseURL()}/api/threads/${threadId}/suggestions`, {
+      method: "POST",
+      headers: { "Content-Type": "application/json" },
+      body: JSON.stringify({
+        messages: recent,
+        n: 3,
+        model_name: context.model_name ?? undefined,
+      }),
+      signal: controller.signal,
+    })
+      .then(async (res) => {
+        if (!res.ok) {
+          return { suggestions: [] as string[] };
+        }
+        return (await res.json()) as { suggestions?: string[] };
+      })
+      .then((data) => {
+        const suggestions = (data.suggestions ?? [])
+          .map((s) => (typeof s === "string" ? s.trim() : ""))
+          .filter((s) => s.length > 0)
+          .slice(0, 5);
+        setFollowups(suggestions);
+      })
+      .catch(() => {
+        setFollowups([]);
+      })
+      .finally(() => {
+        setFollowupsLoading(false);
+      });
+
+    return () => controller.abort();
+  }, [context.model_name, disabled, isMock, status, thread.messages, threadId]);
+
+  return (
+    <div ref={promptRootRef} className="relative flex flex-col gap-6">
+      {extraHeader && (
+        <div className="flex items-center justify-center">
+          {extraHeader}
+        </div>
+      )}
+
+      {isNewThread && searchParams.get("mode") !== "skill" && showAdminRecommendedQuestions && (
+        <div className="flex flex-wrap justify-center gap-2.5 px-4 mb-[10px]">
+          {recommendedQuestions.map((q) => (
+            <button
+              key={q.id}
+              type="button"
+              className="rounded-full border border-[#e4e4e7] bg-white px-4 py-2 text-sm text-[#18181b] shadow-none transition-colors hover:bg-[#f4f4f5] cursor-pointer dark:border-slate-600 dark:bg-slate-700/50 dark:text-slate-300 dark:hover:bg-slate-700"
+              onClick={() => {
+                textInput.setInput(q.content);
+                setFollowupsHidden(true);
+                setTimeout(() => requestFormSubmit(), 0);
+              }}
+            >
+              {desensitize(q.title)}
+            </button>
+          ))}
+        </div>
+      )}
+
+      {showFollowups && (
+        <div className="flex items-center justify-center pb-2">
+          <div className="flex items-center gap-2">
+            {followupsLoading ? (
+              <div className="text-muted-foreground bg-background/80 rounded-full border px-4 py-2 text-xs backdrop-blur-sm">
+                {t.inputBox.followupLoading}
+              </div>
+            ) : (
+              <Suggestions className="min-h-16 w-fit items-start">
+                {followups.map((s) => (
+                  <Suggestion
+                    key={s}
+                    suggestion={s}
+                    onClick={() => handleFollowupClick(s)}
+                  />
+                ))}
+                <Button
+                  aria-label={t.common.close}
+                  className="text-muted-foreground cursor-pointer rounded-full px-3 text-xs font-normal"
+                  variant="outline"
+                  size="sm"
+                  type="button"
+                  onClick={() => setFollowupsHidden(true)}
+                >
+                  <XIcon className="size-4" />
+                </Button>
+              </Suggestions>
+            )}
+          </div>
+        </div>
+      )}
+      <div className="flex flex-col gap-1.5">
+      {isNewThread && (
+        <CustomPromptBar
+          className="px-1"
+          currentAgentId={agentSelector?.selectedAgentId ?? assistantId ?? null}
+          onInsert={(content) => {
+            const current = textInput.value?.trim() ?? "";
+            textInput.setInput(current ? `${current}\n${content}` : content);
+            setTimeout(() => {
+              document
+                .querySelector<HTMLTextAreaElement>("textarea[name='message']")
+                ?.focus();
+            }, 0);
+          }}
+        />
+      )}
+      <div className={cn(isNewThread && !compact && "gemini-gradient-border gemini-shadow")}>
+      <PromptInput
+        className={cn(
+          "transition-all duration-300 ease-out",
+          // 紧凑落地态（CM 参考图）：表单作为白卡，内部 input-group 透明无边框，
+          // 避免双层描边；圆角对齐 2xl。
+          compact && isNewThread
+            ? "*:data-[slot='input-group']:rounded-2xl *:data-[slot='input-group']:border-0 *:data-[slot='input-group']:bg-transparent *:data-[slot='input-group']:shadow-none"
+            : "*:data-[slot='input-group']:rounded-3xl",
+          // 紧凑落地态（CM 参考图）：白卡 + 细边 + 轻阴影 + 主色聚焦环（无旋转彩边）
+          compact && isNewThread
+            ? "rounded-2xl border border-gray-200/70 bg-white overflow-hidden shadow-[0_4.5px_22px_rgba(0,0,0,0.025)] transition-all duration-300 focus-within:border-[color-mix(in_srgb,var(--primary)_35%,transparent)] focus-within:ring-4 focus-within:ring-[color-mix(in_srgb,var(--primary)_5%,transparent)] focus-within:shadow-[0_10px_35px_color-mix(in_srgb,var(--primary)_4%,transparent)] dark:border-slate-700 dark:bg-slate-800"
+            : // landing 态：透明，让外层 gemini-gradient-border 提供白卡 + 旋转描边
+              // 对话态：常规白卡 + 中性边框 + 轻阴影
+              isNewThread
+              ? "rounded-3xl bg-transparent border-0 shadow-none ring-0"
+              : "rounded-3xl bg-background border border-border shadow-sm transition-shadow hover:shadow-md focus-within:border-ring/40 focus-within:shadow-md focus-within:ring-1 focus-within:ring-ring/20 dark:bg-slate-800 dark:border-slate-700",
+          className,
+        )}
+        disabled={disabled}
+        globalDrop
+        multiple
+        onSubmit={handleSubmit}
+        {...props}
+      >
+        <PromptInputAttachments>
+          {(attachment) => <PromptInputAttachment data={attachment} />}
+        </PromptInputAttachments>
+        <PromptInputBody
+          className={cn(!compact && "absolute top-0 right-0 left-0 z-3")}
+        >
+          <PromptInputTextarea
+            className={cn(
+              "size-full px-2 py-1 text-[15px] leading-7 placeholder:text-muted-foreground/60",
+              isNewThread && "min-h-[110px]!",
+              compact && "min-h-[84px]! px-3 pt-3 pb-1 text-[14px]",
+            )}
+            disabled={disabled}
+            placeholder={t.inputBox.placeholder}
+            autoFocus={autoFocus}
+            defaultValue={initialValue}
+          />
+        </PromptInputBody>
+        <PromptInputFooter
+          className={cn(
+            "flex items-center border-t-0 px-2 pt-1 pb-1.5",
+            compact && "gap-2 px-3 pt-1.5 pb-3",
+          )}
+        >
+          <PromptInputTools
+            className={cn("gap-1.5", compact && "flex-1 flex-wrap gap-2")}
+          >
+            {/* TODO: Add more connectors here
+          <PromptInputActionMenu>
+            <PromptInputActionMenuTrigger className="px-2!" />
+            <PromptInputActionMenuContent>
+              <PromptInputActionAddAttachments
+                label={t.inputBox.addAttachments}
+              />
+            </PromptInputActionMenuContent>
+          </PromptInputActionMenu> */}
+            <AddAttachmentsButton
+              className={cn(
+                "rounded-full border-0 bg-transparent px-2.5! shadow-none text-muted-foreground hover:bg-muted/70 hover:text-foreground dark:text-slate-400 dark:hover:bg-slate-700/70 dark:hover:text-slate-100",
+                compact &&
+                  "gap-1.5! rounded-lg px-2.5! py-1! text-[11px] font-medium",
+              )}
+            />
+            <Tooltip
+              content={
+                context.writing_mode
+                  ? "写作模式已开启：最终输出会强制保存为 Markdown 文件"
+                  : "开启写作模式：最终输出保存为 Markdown 文件并进入画布"
+              }
+            >
+              <PromptInputButton
+                className={cn(
+                  "gap-1.5! rounded-full px-3! shadow-none",
+                  compact && "rounded-lg px-2.5! py-1! text-[11px] font-medium",
+                  context.writing_mode
+                    ? "border border-amber-300 bg-amber-50 text-amber-700 hover:bg-amber-100 dark:border-amber-700 dark:bg-amber-950/40 dark:text-amber-200"
+                    : compact
+                      ? "border border-transparent bg-muted/40 text-muted-foreground hover:bg-muted hover:text-foreground"
+                      : "border-0 bg-transparent text-muted-foreground hover:bg-muted/70 hover:text-foreground dark:text-slate-400 dark:hover:bg-slate-700/70 dark:hover:text-slate-100",
+                )}
+                onClick={handleWritingModeToggle}
+              >
+                <FileTextIcon className="size-3" />
+                <span className="text-xs font-normal">写作</span>
+              </PromptInputButton>
+            </Tooltip>
+            <Tooltip
+              content={
+                memoryInjectionEnabled
+                  ? "记忆注入已开启：对话前会注入历史记忆"
+                  : "记忆注入已关闭：本次对话不注入记忆"
+              }
+            >
+              <PromptInputButton
+                className={cn(
+                  "gap-1.5! rounded-full px-3! shadow-none",
+                  compact && "rounded-lg px-2.5! py-1! text-[11px] font-medium",
+                  memoryInjectionEnabled
+                    ? "border border-blue-300 bg-blue-50 text-blue-700 hover:bg-blue-100 dark:border-blue-700 dark:bg-blue-950/40 dark:text-blue-200"
+                    : compact
+                      ? "border border-transparent bg-muted/40 text-muted-foreground hover:bg-muted hover:text-foreground"
+                      : "border-0 bg-transparent text-muted-foreground hover:bg-muted/70 hover:text-foreground dark:text-slate-400 dark:hover:bg-slate-700/70 dark:hover:text-slate-100",
+                )}
+                onClick={handleMemoryInjectionToggle}
+              >
+                <BrainCircuitIcon className="size-3" />
+                <span className="text-xs font-normal">记忆</span>
+              </PromptInputButton>
+            </Tooltip>
+            <Tooltip
+              content={
+                ragModeEnabled
+                  ? "参考文献模式已开启：检索类结果会按编号引用展示"
+                  : "参考文献模式已关闭：点击开启"
+              }
+            >
+              <PromptInputButton
+                type="button"
+                aria-pressed={ragModeEnabled}
+                disabled={ragModeLocked || disabled}
+                className={cn(
+                  "gap-1.5! rounded-full px-3! shadow-none",
+                  compact && "rounded-lg px-2.5! py-1! text-[11px] font-medium",
+                  ragModeLocked && "cursor-not-allowed opacity-80",
+                  ragModeEnabled
+                    ? "border border-emerald-300 bg-emerald-50 text-emerald-700 hover:bg-emerald-100 dark:border-emerald-700 dark:bg-emerald-950/40 dark:text-emerald-200"
+                    : compact
+                      ? "border border-transparent bg-muted/40 text-muted-foreground hover:bg-muted hover:text-foreground"
+                      : "border-0 bg-transparent text-muted-foreground hover:bg-muted/70 hover:text-foreground dark:text-slate-400 dark:hover:bg-slate-700/70 dark:hover:text-slate-100",
+                )}
+                onClick={handleRagModeToggle}
+              >
+                {ragModeEnabled && (
+                  <CheckIcon className="size-3 text-emerald-700 dark:text-emerald-200" />
+                )}
+                <BookMarkedIcon className="size-3" />
+                <span className="text-xs font-normal">参考文献</span>
+              </PromptInputButton>
+            </Tooltip>
+            <Tooltip
+              content={
+                contextCompressionEnabled
+                  ? "上下文压缩已开启：超长对话会自动压缩历史以节省 token"
+                  : "上下文压缩已关闭：点击开启"
+              }
+            >
+              <PromptInputButton
+                type="button"
+                aria-pressed={contextCompressionEnabled}
+                disabled={disabled}
+                className={cn(
+                  "gap-1.5! rounded-full px-3! shadow-none",
+                  compact && "rounded-lg px-2.5! py-1! text-[11px] font-medium",
+                  contextCompressionEnabled
+                    ? "border border-violet-300 bg-violet-50 text-violet-700 hover:bg-violet-100 dark:border-violet-700 dark:bg-violet-950/40 dark:text-violet-200"
+                    : compact
+                      ? "border border-transparent bg-muted/40 text-muted-foreground hover:bg-muted hover:text-foreground"
+                      : "border-0 bg-transparent text-muted-foreground hover:bg-muted/70 hover:text-foreground dark:text-slate-400 dark:hover:bg-slate-700/70 dark:hover:text-slate-100",
+                )}
+                onClick={handleContextCompressionToggle}
+              >
+                {contextCompressionEnabled && (
+                  <CheckIcon className="size-3 text-violet-700 dark:text-violet-200" />
+                )}
+                <ShrinkIcon className="size-3" />
+                <span className="text-xs font-normal">上下文压缩</span>
+              </PromptInputButton>
+            </Tooltip>
+            <PromptInputActionMenu>
+              <ModeHoverGuide
+                mode={
+                  context.mode === "flash" ||
+                  context.mode === "thinking" ||
+                  context.mode === "pro" ||
+                  context.mode === "ultra"
+                    ? context.mode
+                    : "flash"
+                }
+              >
+                <PromptInputActionMenuTrigger
+                  className={cn(
+                    "gap-1.5! rounded-full border-0 bg-transparent px-3! text-muted-foreground shadow-none hover:bg-muted/70 hover:text-foreground dark:text-slate-400 dark:hover:bg-slate-700/70 dark:hover:text-slate-100",
+                    compact && "rounded-lg border px-2.5! py-1! text-[11px] font-medium",
+                    compact &&
+                      (context.mode === "thinking"
+                        ? "border-indigo-200 bg-indigo-50/60 text-indigo-700 hover:bg-indigo-100 dark:border-indigo-700 dark:bg-indigo-950/40 dark:text-indigo-200"
+                        : context.mode === "pro"
+                          ? "border-violet-200 bg-violet-50/60 text-violet-700 hover:bg-violet-100 dark:border-violet-700 dark:bg-violet-950/40 dark:text-violet-200"
+                          : context.mode === "ultra"
+                            ? "border-amber-200 bg-amber-50/60 text-amber-700 hover:bg-amber-100 dark:border-amber-700 dark:bg-amber-950/40 dark:text-amber-200"
+                            : "border-transparent bg-muted/40 hover:bg-muted hover:text-foreground"),
+                  )}
+                >
+                  <div>
+                    {context.mode === "flash" && <ZapIcon className="size-3" />}
+                    {context.mode === "thinking" && (
+                      <LightbulbIcon className="size-3" />
+                    )}
+                    {context.mode === "pro" && (
+                      <GraduationCapIcon className="size-3" />
+                    )}
+                    {context.mode === "ultra" && (
+                      <RocketIcon className="size-3 text-[#dabb5e]" />
+                    )}
+                  </div>
+                  <div
+                    className={cn(
+                      "text-xs font-normal",
+                      context.mode === "ultra" ? "golden-text" : "",
+                    )}
+                  >
+                    {(context.mode === "flash" && t.inputBox.flashMode) ||
+                      (context.mode === "thinking" &&
+                        t.inputBox.reasoningMode) ||
+                      (context.mode === "pro" && t.inputBox.proMode) ||
+                      (context.mode === "ultra" && t.inputBox.ultraMode)}
+                  </div>
+                </PromptInputActionMenuTrigger>
+              </ModeHoverGuide>
+              <PromptInputActionMenuContent className="w-80">
+                <DropdownMenuGroup>
+                  <DropdownMenuLabel className="text-muted-foreground text-xs">
+                    {t.inputBox.mode}
+                  </DropdownMenuLabel>
+                  <PromptInputActionMenu>
+                    <PromptInputActionMenuItem
+                      className={cn(
+                        context.mode === "flash"
+                          ? "text-accent-foreground"
+                          : "text-muted-foreground/65",
+                      )}
+                      onSelect={() => handleModeSelect("flash")}
+                    >
+                      <div className="flex flex-col gap-2">
+                        <div className="flex items-center gap-1 font-bold">
+                          <ZapIcon
+                            className={cn(
+                              "mr-2 size-4",
+                              context.mode === "flash" &&
+                                "text-accent-foreground",
+                            )}
+                          />
+                          {t.inputBox.flashMode}
+                        </div>
+                        <div className="pl-7 text-xs">
+                          {t.inputBox.flashModeDescription}
+                        </div>
+                      </div>
+                      {context.mode === "flash" ? (
+                        <CheckIcon className="ml-auto size-4" />
+                      ) : (
+                        <div className="ml-auto size-4" />
+                      )}
+                    </PromptInputActionMenuItem>
+                    {supportThinking && (
+                      <PromptInputActionMenuItem
+                        className={cn(
+                          context.mode === "thinking"
+                            ? "text-accent-foreground"
+                            : "text-muted-foreground/65",
+                        )}
+                        onSelect={() => handleModeSelect("thinking")}
+                      >
+                        <div className="flex flex-col gap-2">
+                          <div className="flex items-center gap-1 font-bold">
+                            <LightbulbIcon
+                              className={cn(
+                                "mr-2 size-4",
+                                context.mode === "thinking" &&
+                                  "text-accent-foreground",
+                              )}
+                            />
+                            {t.inputBox.reasoningMode}
+                          </div>
+                          <div className="pl-7 text-xs">
+                            {t.inputBox.reasoningModeDescription}
+                          </div>
+                        </div>
+                        {context.mode === "thinking" ? (
+                          <CheckIcon className="ml-auto size-4" />
+                        ) : (
+                          <div className="ml-auto size-4" />
+                        )}
+                      </PromptInputActionMenuItem>
+                    )}
+                    <PromptInputActionMenuItem
+                      className={cn(
+                        context.mode === "pro"
+                          ? "text-accent-foreground"
+                          : "text-muted-foreground/65",
+                      )}
+                      onSelect={() => handleModeSelect("pro")}
+                    >
+                      <div className="flex flex-col gap-2">
+                        <div className="flex items-center gap-1 font-bold">
+                          <GraduationCapIcon
+                            className={cn(
+                              "mr-2 size-4",
+                              context.mode === "pro" &&
+                                "text-accent-foreground",
+                            )}
+                          />
+                          {t.inputBox.proMode}
+                        </div>
+                        <div className="pl-7 text-xs">
+                          {t.inputBox.proModeDescription}
+                        </div>
+                      </div>
+                      {context.mode === "pro" ? (
+                        <CheckIcon className="ml-auto size-4" />
+                      ) : (
+                        <div className="ml-auto size-4" />
+                      )}
+                    </PromptInputActionMenuItem>
+                    <PromptInputActionMenuItem
+                      className={cn(
+                        context.mode === "ultra"
+                          ? "text-accent-foreground"
+                          : "text-muted-foreground/65",
+                      )}
+                      onSelect={() => handleModeSelect("ultra")}
+                    >
+                      <div className="flex flex-col gap-2">
+                        <div className="flex items-center gap-1 font-bold">
+                          <RocketIcon
+                            className={cn(
+                              "mr-2 size-4",
+                              context.mode === "ultra" && "text-[#dabb5e]",
+                            )}
+                          />
+                          <div
+                            className={cn(
+                              context.mode === "ultra" && "golden-text",
+                            )}
+                          >
+                            {t.inputBox.ultraMode}
+                          </div>
+                        </div>
+                        <div className="pl-7 text-xs">
+                          {t.inputBox.ultraModeDescription}
+                        </div>
+                      </div>
+                      {context.mode === "ultra" ? (
+                        <CheckIcon className="ml-auto size-4" />
+                      ) : (
+                        <div className="ml-auto size-4" />
+                      )}
+                    </PromptInputActionMenuItem>
+                  </PromptInputActionMenu>
+                </DropdownMenuGroup>
+              </PromptInputActionMenuContent>
+            </PromptInputActionMenu>
+            {supportReasoningEffort && context.mode !== "flash" && (
+              <PromptInputActionMenu>
+                <PromptInputActionMenuTrigger className={cn("gap-1.5! rounded-full border-0 bg-transparent px-3! text-muted-foreground shadow-none hover:bg-muted/70 hover:text-foreground dark:text-slate-400 dark:hover:bg-slate-700/70 dark:hover:text-slate-100", compact && "rounded-lg border border-border px-2.5! py-1! text-[13px]")}>
+                  <div className="text-xs font-normal">
+                    {t.inputBox.reasoningEffort}:
+                    {context.reasoning_effort === "minimal" &&
+                      " " + t.inputBox.reasoningEffortMinimal}
+                    {context.reasoning_effort === "low" &&
+                      " " + t.inputBox.reasoningEffortLow}
+                    {context.reasoning_effort === "medium" &&
+                      " " + t.inputBox.reasoningEffortMedium}
+                    {context.reasoning_effort === "high" &&
+                      " " + t.inputBox.reasoningEffortHigh}
+                  </div>
+                </PromptInputActionMenuTrigger>
+                <PromptInputActionMenuContent className="w-70">
+                  <DropdownMenuGroup>
+                    <DropdownMenuLabel className="text-muted-foreground text-xs">
+                      {t.inputBox.reasoningEffort}
+                    </DropdownMenuLabel>
+                    <PromptInputActionMenu>
+                      <PromptInputActionMenuItem
+                        className={cn(
+                          context.reasoning_effort === "minimal"
+                            ? "text-accent-foreground"
+                            : "text-muted-foreground/65",
+                        )}
+                        onSelect={() => handleReasoningEffortSelect("minimal")}
+                      >
+                        <div className="flex flex-col gap-2">
+                          <div className="flex items-center gap-1 font-bold">
+                            {t.inputBox.reasoningEffortMinimal}
+                          </div>
+                          <div className="pl-2 text-xs">
+                            {t.inputBox.reasoningEffortMinimalDescription}
+                          </div>
+                        </div>
+                        {context.reasoning_effort === "minimal" ? (
+                          <CheckIcon className="ml-auto size-4" />
+                        ) : (
+                          <div className="ml-auto size-4" />
+                        )}
+                      </PromptInputActionMenuItem>
+                      <PromptInputActionMenuItem
+                        className={cn(
+                          context.reasoning_effort === "low"
+                            ? "text-accent-foreground"
+                            : "text-muted-foreground/65",
+                        )}
+                        onSelect={() => handleReasoningEffortSelect("low")}
+                      >
+                        <div className="flex flex-col gap-2">
+                          <div className="flex items-center gap-1 font-bold">
+                            {t.inputBox.reasoningEffortLow}
+                          </div>
+                          <div className="pl-2 text-xs">
+                            {t.inputBox.reasoningEffortLowDescription}
+                          </div>
+                        </div>
+                        {context.reasoning_effort === "low" ? (
+                          <CheckIcon className="ml-auto size-4" />
+                        ) : (
+                          <div className="ml-auto size-4" />
+                        )}
+                      </PromptInputActionMenuItem>
+                      <PromptInputActionMenuItem
+                        className={cn(
+                          context.reasoning_effort === "medium" ||
+                            !context.reasoning_effort
+                            ? "text-accent-foreground"
+                            : "text-muted-foreground/65",
+                        )}
+                        onSelect={() => handleReasoningEffortSelect("medium")}
+                      >
+                        <div className="flex flex-col gap-2">
+                          <div className="flex items-center gap-1 font-bold">
+                            {t.inputBox.reasoningEffortMedium}
+                          </div>
+                          <div className="pl-2 text-xs">
+                            {t.inputBox.reasoningEffortMediumDescription}
+                          </div>
+                        </div>
+                        {context.reasoning_effort === "medium" ||
+                        !context.reasoning_effort ? (
+                          <CheckIcon className="ml-auto size-4" />
+                        ) : (
+                          <div className="ml-auto size-4" />
+                        )}
+                      </PromptInputActionMenuItem>
+                      <PromptInputActionMenuItem
+                        className={cn(
+                          context.reasoning_effort === "high"
+                            ? "text-accent-foreground"
+                            : "text-muted-foreground/65",
+                        )}
+                        onSelect={() => handleReasoningEffortSelect("high")}
+                      >
+                        <div className="flex flex-col gap-2">
+                          <div className="flex items-center gap-1 font-bold">
+                            {t.inputBox.reasoningEffortHigh}
+                          </div>
+                          <div className="pl-2 text-xs">
+                            {t.inputBox.reasoningEffortHighDescription}
+                          </div>
+                        </div>
+                        {context.reasoning_effort === "high" ? (
+                          <CheckIcon className="ml-auto size-4" />
+                        ) : (
+                          <div className="ml-auto size-4" />
+                        )}
+                      </PromptInputActionMenuItem>
+                    </PromptInputActionMenu>
+                  </DropdownMenuGroup>
+                </PromptInputActionMenuContent>
+              </PromptInputActionMenu>
+            )}
+          </PromptInputTools>
+          <PromptInputTools className={cn("gap-1.5", compact && "gap-2")}>
+            {agentSelector && (
+              <AgentSelector
+                selectedAgentId={agentSelector.selectedAgentId}
+                onAgentChange={agentSelector.onAgentChange}
+                disabled={disabled}
+                compact={compact}
+              />
+            )}
+            {promptPrefix && (
+              <Tooltip
+                content={
+                  promptPrefix.on
+                    ? `已开启：${promptPrefix.label}`
+                    : `点击开启：${promptPrefix.label}`
+                }
+              >
+                <PromptInputButton
+                  type="button"
+                  onClick={promptPrefix.onToggle}
+                  aria-pressed={promptPrefix.on}
+                  className={cn(
+                    "rounded-full px-3 shadow-none transition-colors",
+                    promptPrefix.on
+                      ? "border border-amber-300 bg-amber-50 text-amber-700 hover:bg-amber-100 dark:border-amber-700 dark:bg-amber-950/40 dark:text-amber-200"
+                      : "border-0 bg-transparent text-muted-foreground hover:bg-muted/70 hover:text-foreground",
+                  )}
+                >
+                  <span className="max-w-[160px] truncate text-xs font-normal">
+                    {promptPrefix.label}
+                  </span>
+                  {promptPrefix.on && (
+                    <CheckIcon className="ml-1 size-3.5 text-amber-700 dark:text-amber-200" />
+                  )}
+                </PromptInputButton>
+              </Tooltip>
+            )}
+            <ModelSelector
+              open={modelDialogOpen}
+              onOpenChange={setModelDialogOpen}
+            >
+              <ModelSelectorTrigger asChild>
+                <PromptInputButton
+                  className={cn(
+                    "rounded-full border-0 bg-transparent px-3 shadow-none text-muted-foreground hover:bg-muted/70 hover:text-foreground dark:text-slate-400 dark:hover:bg-slate-700/70 dark:hover:text-slate-100",
+                    compact &&
+                      "rounded-lg border border-transparent bg-gray-50/60 px-2.5! py-1! font-mono text-gray-600 hover:bg-gray-100/80 hover:text-gray-800 dark:bg-slate-700/50 dark:text-slate-300 dark:hover:bg-slate-700",
+                  )}
+                >
+                  <div className="flex min-w-0 flex-col items-start text-left">
+                    <ModelSelectorName
+                      className={cn(
+                        "text-xs font-normal",
+                        compact && "text-[11px]",
+                      )}
+                    >
+                      {selectedModel?.display_name}
+                    </ModelSelectorName>
+                  </div>
+                </PromptInputButton>
+              </ModelSelectorTrigger>
+              <ModelSelectorContent>
+                <ModelSelectorInput placeholder={t.inputBox.searchModels} />
+                <ModelSelectorList>
+                  {models.map((m) => (
+                    <ModelSelectorItem
+                      key={m.name}
+                      value={m.name}
+                      onSelect={() => handleModelSelect(m.name)}
+                    >
+                      <div className="flex min-w-0 flex-1 flex-col">
+                        <ModelSelectorName>{m.display_name}</ModelSelectorName>
+                        <span className="text-muted-foreground truncate text-[10px]">
+                          {m.model}
+                        </span>
+                      </div>
+                      {m.name === context.model_name ? (
+                        <CheckIcon className="ml-auto size-4" />
+                      ) : (
+                        <div className="ml-auto size-4" />
+                      )}
+                    </ModelSelectorItem>
+                  ))}
+                </ModelSelectorList>
+              </ModelSelectorContent>
+            </ModelSelector>
+            <PromptInputSubmit
+              className={cn(
+                "rounded-full border-border bg-background shadow-sm hover:bg-muted dark:bg-blue-500 dark:border-blue-500 dark:text-white dark:hover:bg-blue-400 dark:shadow-none",
+                compact &&
+                  "size-7 rounded-lg border-0 bg-[var(--primary)] text-[var(--primary-foreground)] hover:bg-[color-mix(in_srgb,var(--primary)_88%,black)] disabled:bg-muted disabled:text-muted-foreground/40 dark:bg-[var(--primary)] dark:border-0 dark:text-[var(--primary-foreground)] dark:hover:bg-[color-mix(in_srgb,var(--primary)_88%,black)] dark:disabled:bg-muted",
+              )}
+              disabled={disabled}
+              variant="outline"
+              status={status}
+            />
+          </PromptInputTools>
+        </PromptInputFooter>
+        {!isNewThread && (
+          <div className="absolute right-0 -bottom-[17px] left-0 z-0 h-4 bg-background"></div>
+        )}
+      </PromptInput>
+      </div>
+      </div>
+
+      <Dialog open={confirmOpen} onOpenChange={setConfirmOpen}>
+        <DialogContent>
+          <DialogHeader>
+            <DialogTitle>{t.inputBox.followupConfirmTitle}</DialogTitle>
+            <DialogDescription>
+              {t.inputBox.followupConfirmDescription}
+            </DialogDescription>
+          </DialogHeader>
+          <DialogFooter>
+            <Button variant="outline" onClick={() => setConfirmOpen(false)}>
+              {t.common.cancel}
+            </Button>
+            <Button variant="secondary" onClick={confirmAppendAndSend}>
+              {t.inputBox.followupConfirmAppend}
+            </Button>
+            <Button onClick={confirmReplaceAndSend}>
+              {t.inputBox.followupConfirmReplace}
+            </Button>
+          </DialogFooter>
+        </DialogContent>
+      </Dialog>
+    </div>
+  );
+}
+
+function SuggestionList() {
+  const { t } = useI18n();
+  const { textInput } = usePromptInputController();
+  const handleSuggestionClick = useCallback(
+    (prompt: string | undefined) => {
+      if (!prompt) return;
+      textInput.setInput(prompt);
+      setTimeout(() => {
+        const textarea = document.querySelector<HTMLTextAreaElement>(
+          "textarea[name='message']",
+        );
+        if (textarea) {
+          const selStart = prompt.indexOf("[");
+          const selEnd = prompt.indexOf("]");
+          if (selStart !== -1 && selEnd !== -1) {
+            textarea.setSelectionRange(selStart, selEnd + 1);
+            textarea.focus();
+          }
+        }
+      }, 500);
+    },
+    [textInput],
+  );
+  return (
+    <Suggestions className="min-h-16 w-fit items-start">
+      <ConfettiButton
+        className="text-muted-foreground cursor-pointer rounded-full px-4 text-xs font-normal"
+        variant="outline"
+        size="sm"
+        onClick={() => handleSuggestionClick(t.inputBox.surpriseMePrompt)}
+      >
+        <SparklesIcon className="size-4" /> {t.inputBox.surpriseMe}
+      </ConfettiButton>
+      {t.inputBox.suggestions.map((suggestion) => (
+        <Suggestion
+          key={suggestion.suggestion}
+          icon={suggestion.icon}
+          suggestion={suggestion.suggestion}
+          onClick={() => handleSuggestionClick(suggestion.prompt)}
+        />
+      ))}
+      <DropdownMenu>
+        <DropdownMenuTrigger asChild>
+          <Suggestion icon={PlusIcon} suggestion={t.common.create} />
+        </DropdownMenuTrigger>
+        <DropdownMenuContent align="start">
+          <DropdownMenuGroup>
+            {t.inputBox.suggestionsCreate.map((suggestion, index) =>
+              "type" in suggestion && suggestion.type === "separator" ? (
+                <DropdownMenuSeparator key={index} />
+              ) : (
+                !("type" in suggestion) && (
+                  <DropdownMenuItem
+                    key={suggestion.suggestion}
+                    onClick={() => handleSuggestionClick(suggestion.prompt)}
+                  >
+                    {suggestion.icon && <suggestion.icon className="size-4" />}
+                    {suggestion.suggestion}
+                  </DropdownMenuItem>
+                )
+              ),
+            )}
+          </DropdownMenuGroup>
+        </DropdownMenuContent>
+      </DropdownMenu>
+    </Suggestions>
+  );
+}
+
+function AddAttachmentsButton({ className }: { className?: string }) {
+  const { t } = useI18n();
+  const attachments = usePromptInputAttachments();
+  return (
+    <Tooltip content={t.inputBox.addAttachments}>
+      <PromptInputButton
+        className={cn("px-2!", className)}
+        onClick={() => attachments.openFileDialog()}
+      >
+        <PaperclipIcon className="size-3" />
+      </PromptInputButton>
+    </Tooltip>
+  );
+}
diff --git a/frontend-web/src/components/workspace/messages/message-list.tsx b/frontend-web/src/components/workspace/messages/message-list.tsx
index 5eb2ddde..68aa9429 100644
--- a/frontend-web/src/components/workspace/messages/message-list.tsx
+++ b/frontend-web/src/components/workspace/messages/message-list.tsx
@@ -7,6 +7,7 @@ import { useStickToBottomContext } from "use-stick-to-bottom";
 import {
   Conversation,
   ConversationContent,
+  ConversationScrollButton,
 } from "@/components/ai-elements/conversation";
 import { Button } from "@/components/ui/button";
 import { Textarea } from "@/components/ui/textarea";
@@ -94,6 +95,7 @@ export function MessageList({
   paddingBottom = MESSAGE_LIST_DEFAULT_PADDING_BOTTOM,
   tokenUsageEnabled = false,
   extraRehypePlugins,
+  scrollButtonClassName,
   onClarificationSubmit,
   onEditResubmit,
   onSetupAgentApprove,
@@ -108,6 +110,8 @@ export function MessageList({
   thread: BaseStream<AgentThreadState>;
   paddingBottom?: number;
   tokenUsageEnabled?: boolean;
+  /** 额外定位「回到最新」悬浮按钮——不同页面输入框高度不同，需上移让位。 */
+  scrollButtonClassName?: string;
   extraRehypePlugins?: ReturnType<typeof useRehypeSplitWordsIntoSpans>;
   onClarificationSubmit?: (answer: string) => void | Promise<void>;
   onEditResubmit?: (text: string) => void;
@@ -577,6 +581,16 @@ export function MessageList({
         <div style={{ height: `${paddingBottom}px` }} />
         </MessageMetricsProvider>
       </ConversationContent>
+      {/* ChatGPT 式「回到最新」悬浮钮：仅在向上滚离底部时出现，点击平滑回到
+          最新问答。默认贴底，调用方可用 scrollButtonClassName 上移避开输入框。 */}
+      <ConversationScrollButton
+        className={cn(
+          "pointer-events-auto z-20 size-9 shadow-md",
+          scrollButtonClassName,
+        )}
+        aria-label="回到最新"
+        title="回到最新"
+      />
     </Conversation>
   );
 }
diff --git a/frontend-web/src/components/workspace/settings/appearance-settings-page.tsx b/frontend-web/src/components/workspace/settings/appearance-settings-page.tsx
index ae309232..3c3201e2 100644
--- a/frontend-web/src/components/workspace/settings/appearance-settings-page.tsx
+++ b/frontend-web/src/components/workspace/settings/appearance-settings-page.tsx
@@ -35,7 +35,7 @@ import { cn } from "@/lib/utils";
 import PrivateSquareSettingsPage from "@/pages/PrivateSquareSettingsPage";
 import PublishSquaresSettingsPage from "@/pages/PublishSquaresSettingsPage";
 import PromptPrefixSettingsPage from "@/pages/PromptPrefixSettingsPage";
-import RecommendedQuestionsPage from "@/pages/RecommendedQuestionsPage";
+import SkillCompressionSettingsPage from "@/pages/SkillCompressionSettingsPage";
 
 import { SettingsSection } from "./settings-section";
 
@@ -457,14 +457,14 @@ export function AppearanceSettingsPage({ onClose }: { onClose?: () => void }) {
 
       {isAdmin && (
         <>
-          <Separator />
-          <RecommendedQuestionsPage embedded />
           <Separator />
           <PromptPrefixSettingsPage embedded />
           <Separator />
           <PrivateSquareSettingsPage embedded />
           <Separator />
           <PublishSquaresSettingsPage embedded />
+          <Separator />
+          <SkillCompressionSettingsPage embedded />
         </>
       )}
     </div>
diff --git a/frontend-web/src/components/workspace/settings/skill-settings-page.tsx b/frontend-web/src/components/workspace/settings/skill-settings-page.tsx
index 916c83ca..67db57a4 100644
--- a/frontend-web/src/components/workspace/settings/skill-settings-page.tsx
+++ b/frontend-web/src/components/workspace/settings/skill-settings-page.tsx
@@ -2,6 +2,7 @@
 
 import {
   AlertCircle,
+  Anchor,
   ArchiveIcon,
   ArrowDownUpIcon,
   BookmarkIcon,
@@ -77,6 +78,7 @@ import {
   useGenerateSkillNameZh,
   useInstallSkillUpload,
   useReconcileCustomSkills,
+  useSetSkillAlwaysOn,
   useSetSkillFeatured,
   useSetSkillNameZh,
   useSetSkillPinned,
@@ -84,8 +86,13 @@ import {
   useSkills,
   useValidateSkillUpload,
 } from "@/core/skills/hooks";
-import { usePublishSquares, useViewMode } from "@/core/system-settings/hooks";
+import {
+  usePublishSquares,
+  useSkillCompressionSettings,
+  useViewMode,
+} from "@/core/system-settings/hooks";
 import type { Skill } from "@/core/skills/type";
+import { formatDateTime } from "@/core/utils/datetime";
 import { env } from "@/env";
 import { desensitize } from "@/lib/desensitize";
 import { cn } from "@/lib/utils";
@@ -175,13 +182,22 @@ function SkillSettingsList({ onClose }: { onClose?: () => void }) {
   const fillNamesMutation = useGenerateMissingSkillNameZh();
   // Optional model override for the admin "一键补充中文名" action (null = default).
   const [fillModel, setFillModel] = useState<string | null>(null);
-  // Name + tag filtering is resolved server-side via the /api/skills query.
-  const { skills, isLoading, error } = useSkills({ search, tagIds, sort });
-
   const storedAuth = useMemo(() => getStoredAuth(), []);
   const currentUserId = storedAuth?.user_id ?? null;
   const isAdmin = storedAuth?.system_role === "admin";
 
+  // Name + tag filtering is resolved server-side via the /api/skills query.
+  // A 岗位 (position) no longer *restricts* the list — it grants skills into the
+  // member's 我的技能 (each such skill comes back with ``favorited: true``), so
+  // the 广场 always shows every visible skill and the 我的 tab includes the
+  // position-granted ones. ``scope`` is kept for API compatibility only.
+  const { skills, isLoading, error } = useSkills({
+    search,
+    tagIds,
+    sort,
+    scope: isAdmin ? undefined : "mine",
+  });
+
   // 把可见技能分类成 UI 关心的几个桶。
   // "All" = 内置 + 已发布的技能（仅 active 状态);自己的技能也要自己发布了才出现。
   // "Mine" = 严格属于我自己的自定义技能（仅 active 状态)。
@@ -209,7 +225,9 @@ function SkillSettingsList({ onClose }: { onClose?: () => void }) {
       if (isBuiltin || !!skill.published) {
         all.push(skill);
       }
-      if (isOwnedByMe) {
+      // "我的" = skills I own OR skills granted into my collection by my 岗位
+      // (position) — mirroring how favorited agents land in 我的智能体.
+      if (isOwnedByMe || !!skill.favorited) {
         mine.push(skill);
       }
     }
@@ -593,6 +611,8 @@ export function SkillRow({
   const { data: cardPublishSquares } = usePublishSquares();
   const setPinned = useSetSkillPinned();
   const setFeatured = useSetSkillFeatured();
+  const setAlwaysOn = useSetSkillAlwaysOn();
+  const { data: compression } = useSkillCompressionSettings();
   const deleteSkill = useDeleteCustomSkill();
   const duplicateSkillMutation = useDuplicateSkill();
   const [confirmDeleteOpen, setConfirmDeleteOpen] = useState(false);
@@ -612,6 +632,17 @@ export function SkillRow({
   // Admin can edit/delete/unpublish anyone's skill; non-admin only their own.
   const canManage = isCustom && (isOwned || isAdmin);
   const canTogglePublish = canManage && !isBuiltin;
+  // 常驻(免压缩):内置技能仅管理员可设;自定义技能其所有者或管理员可设(与后端一致)。
+  const canSetAlwaysOn = isAdmin || isOwned;
+
+  async function handleToggleAlwaysOn(next: boolean) {
+    try {
+      await setAlwaysOn.mutateAsync({ skillName: skill.name, alwaysOn: next });
+      toast.success(next ? "已设为常驻(免压缩)" : "已取消常驻");
+    } catch (err) {
+      toast.error(err instanceof Error ? err.message : String(err));
+    }
+  }
 
   async function handleTogglePublished(next: boolean) {
     // Publishing: prompt for which square to publish into (when configured).
@@ -806,6 +837,15 @@ export function SkillRow({
           {t.settings.skills.badgePinned}
         </Badge>
       )}
+      {skill.always_on && (
+        <Badge
+          variant="outline"
+          className="border-amber-500/30 text-amber-700 dark:text-amber-300 text-xs"
+          title="常驻(免压缩):开启技能压缩后,本技能仍全文注入系统提示词并被正常检索"
+        >
+          常驻
+        </Badge>
+      )}
     </div>
   );
 
@@ -848,6 +888,14 @@ export function SkillRow({
             : t.settings.skills.useCountInlineZero}
         </div>
       )}
+      {formatDateTime(skill.created_at) && (
+        <div className="text-muted-foreground mt-1 flex items-center gap-1 text-xs">
+          <ClockIcon className="size-3.5 shrink-0" />
+          <span className="truncate">
+            创建时间：{formatDateTime(skill.created_at)}
+          </span>
+        </div>
+      )}
       {(skill.tags?.length ?? 0) > 0 && (
         <TagChipList tags={skill.tags!} className="mt-2" />
       )}
@@ -907,6 +955,27 @@ export function SkillRow({
           />
         </Button>
       )}
+      {compression?.enabled && canSetAlwaysOn && (
+        <Button
+          size="icon"
+          variant="ghost"
+          className={cn(
+            "h-8 w-8 shrink-0",
+            skill.always_on && "text-amber-600 hover:text-amber-700",
+          )}
+          disabled={setAlwaysOn.isPending}
+          onClick={() => void handleToggleAlwaysOn(!skill.always_on)}
+          title={
+            skill.always_on
+              ? "取消常驻(将被压缩,改由 search_skills 检索)"
+              : "设为常驻(免压缩,始终全文注入提示词)"
+          }
+        >
+          <Anchor
+            className={cn("size-3.5", skill.always_on && "fill-current")}
+          />
+        </Button>
+      )}
       {canTogglePublish && (
         <Button
           size="icon"
diff --git a/frontend-web/src/components/workspace/skills/skill-detail-dialog.tsx b/frontend-web/src/components/workspace/skills/skill-detail-dialog.tsx
index 733cbf27..e64f7e43 100644
--- a/frontend-web/src/components/workspace/skills/skill-detail-dialog.tsx
+++ b/frontend-web/src/components/workspace/skills/skill-detail-dialog.tsx
@@ -26,6 +26,7 @@ import {
   DialogTitle,
 } from "@/components/ui/dialog";
 import { Input } from "@/components/ui/input";
+import { Switch } from "@/components/ui/switch";
 import { Tabs, TabsContent, TabsList, TabsTrigger } from "@/components/ui/tabs";
 import { Textarea } from "@/components/ui/textarea";
 import { MarkdownContent } from "@/components/workspace/messages/markdown-content";
@@ -37,12 +38,14 @@ import { getStoredAuth } from "@/core/auth";
 import { useI18n } from "@/core/i18n/hooks";
 import {
   useGenerateSkillNameZh,
+  useSetSkillAlwaysOn,
   useSetSkillDetail,
   useSetSkillNameZh,
   useSkill,
   useSkillContent,
   useSkillFiles,
 } from "@/core/skills/hooks";
+import { useSkillCompressionSettings } from "@/core/system-settings/hooks";
 import type { Skill } from "@/core/skills/type";
 import { reasoningPlugins } from "@/core/streamdown";
 import { desensitize } from "@/lib/desensitize";
@@ -105,6 +108,8 @@ export function SkillDetailDialog({
   const setDetail = useSetSkillDetail();
   const setNameZh = useSetSkillNameZh();
   const generateNameZh = useGenerateSkillNameZh();
+  const setAlwaysOn = useSetSkillAlwaysOn();
+  const { data: compression } = useSkillCompressionSettings();
 
   const [editing, setEditing] = useState(false);
   const [draft, setDraft] = useState("");
@@ -327,6 +332,15 @@ export function SkillDetailDialog({
                       {tx.badgePublished}
                     </Badge>
                   )}
+                  {skill?.always_on && compression?.enabled && (
+                    <Badge
+                      variant="outline"
+                      className="border-amber-500/30 text-xs text-amber-700 dark:text-amber-300"
+                      title="常驻(免压缩):本技能始终全文注入系统提示词"
+                    >
+                      常驻
+                    </Badge>
+                  )}
                 </div>
               )}
               <DialogDescription className="mt-1 line-clamp-2 text-sm leading-relaxed">
@@ -372,6 +386,31 @@ export function SkillDetailDialog({
             </TabsList>
 
             <div className="flex shrink-0 items-center gap-1">
+              {canManage && compression?.enabled && skill && (
+                <div
+                  className="mr-1 flex items-center gap-1.5"
+                  title="开启后,即使启用了技能压缩,本技能仍会全文注入系统提示词并被正常检索(免压缩)"
+                >
+                  <span className="text-muted-foreground text-xs">常驻(免压缩)</span>
+                  <Switch
+                    checked={!!skill.always_on}
+                    disabled={setAlwaysOn.isPending}
+                    onCheckedChange={(next) =>
+                      setAlwaysOn.mutate(
+                        { skillName: skill.name, alwaysOn: next },
+                        {
+                          onSuccess: () =>
+                            toast.success(next ? "已设为常驻(免压缩)" : "已取消常驻"),
+                          onError: (e) =>
+                            toast.error(
+                              e instanceof Error ? e.message : "更新常驻状态失败",
+                            ),
+                        },
+                      )
+                    }
+                  />
+                </div>
+              )}
               {selectedTab === "detail" && canManage && !editing && (
                 <Button
                   size="sm"
diff --git a/frontend-web/src/components/workspace/tags/resource-tag-dialog.tsx b/frontend-web/src/components/workspace/tags/resource-tag-dialog.tsx
new file mode 100644
index 00000000..cf44de1f
--- /dev/null
+++ b/frontend-web/src/components/workspace/tags/resource-tag-dialog.tsx
@@ -0,0 +1,98 @@
+"use client";
+
+import { CheckIcon, Loader2 } from "lucide-react";
+import { useState } from "react";
+import { toast } from "sonner";
+
+import { Button } from "@/components/ui/button";
+import {
+  Dialog,
+  DialogContent,
+  DialogFooter,
+  DialogHeader,
+  DialogTitle,
+} from "@/components/ui/dialog";
+import { resolveTagColor } from "@/components/workspace/tags/tag-colors";
+import { useResourceTags, useSetResourceTags, useTags, type TagType } from "@/core/tags";
+import { cn } from "@/lib/utils";
+
+/** Generic admin dialog to assign a resource's tags (multi-select chips). */
+export function ResourceTagDialog({
+  open,
+  onOpenChange,
+  targetType,
+  targetId,
+  title = "打标签",
+}: {
+  open: boolean;
+  onOpenChange: (open: boolean) => void;
+  targetType: TagType;
+  targetId: string;
+  title?: string;
+}) {
+  const { data: tags } = useTags(targetType);
+  const { data: current } = useResourceTags(targetType, targetId, open);
+  const setTags = useSetResourceTags();
+  const [selected, setSelected] = useState<string[]>([]);
+
+  const [syncedFor, setSyncedFor] = useState<string | null>(null);
+  const syncKey = `${open}:${targetId}:${current ? "ready" : "loading"}`;
+  if (open && current && syncedFor !== syncKey) {
+    setSelected(current.map((t) => t.id));
+    setSyncedFor(syncKey);
+  }
+  if (!open && syncedFor !== null) setSyncedFor(null);
+
+  const toggle = (id: string) =>
+    setSelected((prev) => (prev.includes(id) ? prev.filter((x) => x !== id) : [...prev, id]));
+
+  const handleSave = async () => {
+    try {
+      await setTags.mutateAsync({ targetType, targetId, tagIds: selected });
+      onOpenChange(false);
+    } catch (e) {
+      toast.error(String(e instanceof Error ? e.message : e));
+    }
+  };
+
+  return (
+    <Dialog open={open} onOpenChange={onOpenChange}>
+      <DialogContent>
+        <DialogHeader>
+          <DialogTitle>{title}</DialogTitle>
+        </DialogHeader>
+        {!tags || tags.length === 0 ? (
+          <p className="text-muted-foreground text-sm">暂无可用标签，请先到「标签管理」创建对应类型的标签。</p>
+        ) : (
+          <div className="flex flex-wrap gap-1.5">
+            {tags.map((tag) => {
+              const checked = selected.includes(tag.id);
+              const color = resolveTagColor(tag.color);
+              return (
+                <button
+                  key={tag.id}
+                  type="button"
+                  onClick={() => toggle(tag.id)}
+                  className={cn(
+                    "inline-flex items-center gap-1 rounded-full border px-3 py-1 text-xs font-medium transition-all",
+                    color.chip,
+                    checked ? "ring-foreground/40 ring-2 ring-offset-1" : "opacity-45 hover:opacity-90",
+                  )}
+                >
+                  {checked && <CheckIcon className="size-3" />}
+                  {tag.name}
+                </button>
+              );
+            })}
+          </div>
+        )}
+        <DialogFooter>
+          <Button variant="outline" onClick={() => onOpenChange(false)}>取消</Button>
+          <Button onClick={handleSave} disabled={setTags.isPending}>
+            {setTags.isPending && <Loader2 className="size-4 animate-spin" />}保存
+          </Button>
+        </DialogFooter>
+      </DialogContent>
+    </Dialog>
+  );
+}
diff --git a/frontend-web/src/components/workspace/workspace-nav-menu.tsx b/frontend-web/src/components/workspace/workspace-nav-menu.tsx
index fe14ee4f..73a56ac4 100644
--- a/frontend-web/src/components/workspace/workspace-nav-menu.tsx
+++ b/frontend-web/src/components/workspace/workspace-nav-menu.tsx
@@ -4,11 +4,13 @@ import {
   BellIcon,
   BotIcon,
   BrainIcon,
+  BriefcaseIcon,
   CalendarClockIcon,
   ChevronsUpDown,
   DatabaseZapIcon,
   GaugeIcon,
   HistoryIcon,
+  MessageCircleQuestionIcon,
   ScrollTextIcon,
   Settings2Icon,
   SettingsIcon,
@@ -216,6 +218,26 @@ export function WorkspaceNavMenu() {
                     <TagsIcon />
                     {t.adminTags.title}
                   </DropdownMenuItem>
+                  {isAdmin && (
+                    <DropdownMenuItem
+                      onClick={() => navigate("/page/workspace/admin/positions")}
+                    >
+                      <BriefcaseIcon />
+                      岗位管理
+                    </DropdownMenuItem>
+                  )}
+                  {isAdmin && (
+                    <DropdownMenuItem
+                      onClick={() =>
+                        navigate(
+                          "/page/workspace/settings/recommended-questions",
+                        )
+                      }
+                    >
+                      <MessageCircleQuestionIcon />
+                      推荐问题配置
+                    </DropdownMenuItem>
+                  )}
                   {isAdmin && logSystem?.enabled && logSystem.url && (
                     <DropdownMenuItem
                       onClick={() =>
diff --git a/frontend-web/src/components/workspace/workspace-sidebar.tsx b/frontend-web/src/components/workspace/workspace-sidebar.tsx
index de7afebd..39206d90 100644
--- a/frontend-web/src/components/workspace/workspace-sidebar.tsx
+++ b/frontend-web/src/components/workspace/workspace-sidebar.tsx
@@ -12,7 +12,7 @@ import { useTheme } from "@/contexts/ThemeContext";
 import { cn } from "@/lib/utils";
 import { usePathname } from "next/navigation";
 
-import { FrequentAgentsList } from "./frequent-agents-list";
+import { GeneralQaSidebar } from "./general-qa-sidebar";
 import { RecentChatList } from "./recent-chat-list";
 import { RecentRoundtableList } from "./recent-roundtable-list";
 import { RecentWritingList } from "./recent-writing-list";
@@ -30,7 +30,8 @@ export function WorkspaceSidebar({
   const pathname = usePathname();
   const useWhiteSidebar = compactChrome && theme === "light";
   // 通用问答（全量模式）左栏：新建回答 → 常用智能体 → 聊天记录（需求 0616 #1）。
-  // 仅在通用问答（/page/workspace/chats*）下启用三块式布局；其余 workspace 页保持原样。
+  // 仅在通用问答（/page/workspace/chats*）下启用三块式布局；点击常用智能体后是
+  // 「原地嵌入」（停留在 /chats/new?agent=...，仅右侧问答区换成智能体对话），左侧栏不变。
   const isGeneralQa = pathname.startsWith("/page/workspace/chats");
 
   // 紧凑模式：用「外层定宽容器 + transition-[width]」做展开/收起过渡动画。
@@ -58,13 +59,16 @@ export function WorkspaceSidebar({
           {...props}
         >
           <SidebarContent>
-            <WorkspaceNavChatList showCollapseTrigger compact />
-            {isGeneralQa && <FrequentAgentsList />}
-            <RecentChatList
-              compact
-              limit={isGeneralQa ? 10 : undefined}
-              moreHref={isGeneralQa ? "/page/workspace/chats" : undefined}
-            />
+            {isGeneralQa ? (
+              // 全量版通用问答（/page/workspace/chats*）：新建对话侧栏独立成
+              // GeneralQaSidebar（按 0616 样式参考图重画），与其余 workspace 页解耦。
+              <GeneralQaSidebar />
+            ) : (
+              <>
+                <WorkspaceNavChatList showCollapseTrigger compact />
+                <RecentChatList compact />
+              </>
+            )}
           </SidebarContent>
           <SidebarRail />
         </Sidebar>
diff --git a/frontend-web/src/core/auth/task-deeplink.ts b/frontend-web/src/core/auth/task-deeplink.ts
new file mode 100644
index 00000000..603e75c4
--- /dev/null
+++ b/frontend-web/src/core/auth/task-deeplink.ts
@@ -0,0 +1,175 @@
+/**
+ * 外部系统跳转登录时的「任务深链」：登录 URL 带 ``taskId`` + ``goPath`` 时，
+ * 拉取任务详情并据 goPath 把用户直接送到对应的智能体问答 / 多智能体会商页。
+ *
+ * 流程：
+ *   1. 调 consumer 接口 ``GET {VITE_AIPI_CONSUMER_URL}/taskAnalyseSearch/cop-task-detail?taskId=``，
+ *      请求头 ``Authorization: Bearer admin``（由 service 拦截器统一注入）。
+ *   2. 把返回的 data（id/overview/taskDirection/taskName/taskContent）整包存到 sessionStorage。
+ *   3. 按 goPath 解析出业务码 + 模式：
+ *        - ``3qfx`` / ``rwfx`` / ``xdfx``       → 单智能体问答（用「配置管理」里该业务配置的智能体）
+ *        - ``3qfx-A`` / ``rwfx-A`` / ``xdfx-A`` → 多智能体会商（用该业务配置的业务链条）
+ *      携带 taskName / taskContent 作为开场信息。
+ *
+ * 业务码映射与「配置管理」(`/api/business-mapping`) 一致：
+ *   3Q=三情分析(3QFX)、6BF=任务分析(任务FX)、7BF=行动分析(XDFX)。
+ */
+
+import { requestCrudService } from "@/service/request-axios-service";
+import { listBusinessMappings } from "@/strategy-components/api/business-mapping";
+
+/** consumer 接口返回的任务详情（data 字段）。 */
+export interface CopTaskDetail {
+  id: string | number | null;
+  overview: string;
+  taskDirection: string;
+  taskName: string;
+  taskContent: string;
+}
+
+/** 任务详情整包暂存的 sessionStorage key（目标页/后续可读取）。 */
+export const COP_TASK_DETAIL_STORAGE_KEY = "login.copTaskDetail";
+
+/** goPath 基础码 → 业务码（配置管理 businessCode）。 */
+const GOPATH_BUSINESS_CODE: Record<string, string> = {
+  "3qfx": "3Q", // 三情分析
+  rwfx: "6BF", // 任务分析（任务FX）
+  xdfx: "7BF", // 行动分析（XDFX）
+};
+
+/**
+ * 业务码 → 「任务管理」布局下的路由前缀（与 sidebar-menu.ts 的 COH 中任务菜单路径一致）。
+ * 会商 / 智能体问答都挂在此前缀下，保证侧边栏停留在 任务管理 → COH中任务 → 对应业务 高亮。
+ */
+const BUSINESS_TASK_BASE: Record<string, string> = {
+  "3Q": "/page/strategy/task-management/plan-task-sentiment",
+  "6BF": "/page/strategy/task-management/plan-task/plan-task-fx",
+  "7BF": "/page/strategy/task-management/plan-task/plan-task-action",
+};
+
+export interface ParsedGoPath {
+  businessCode: string;
+  /** true = 多智能体会商（``-A`` 后缀）；false = 单智能体问答。 */
+  roundtable: boolean;
+}
+
+/** 解析 goPath（大小写不敏感）；``-A`` 后缀表示多智能体会商。无法识别返回 null。 */
+export function parseGoPath(goPath: string): ParsedGoPath | null {
+  const raw = goPath.trim().toLowerCase();
+  if (!raw) return null;
+  const roundtable = raw.endsWith("-a");
+  const base = roundtable ? raw.slice(0, -2) : raw;
+  const businessCode = GOPATH_BUSINESS_CODE[base];
+  if (!businessCode) return null;
+  return { businessCode, roundtable };
+}
+
+function consumerBaseUrl(): string {
+  const value = import.meta.env.VITE_AIPI_CONSUMER_URL;
+  return typeof value === "string" ? value.replace(/\/+$/, "") : "";
+}
+
+/** 拉取任务详情；缺少 consumer 地址 / 返回不合法时返回 null。 */
+export async function fetchCopTaskDetail(taskId: string): Promise<CopTaskDetail | null> {
+  const base = consumerBaseUrl();
+  if (!base || !taskId) return null;
+  // service 拦截器已统一注入 `Authorization: Bearer admin` 并返回 response.data（整包 body）。
+  const body = await requestCrudService.get(`${base}/taskAnalyseSearch/cop-task-detail`, { taskId });
+  const data = body && typeof body === "object" ? (body as { data?: unknown }).data : null;
+  if (!data || typeof data !== "object") return null;
+  const d = data as Record<string, unknown>;
+  return {
+    id: (d.id as string | number | null) ?? null,
+    overview: typeof d.overview === "string" ? d.overview : "",
+    taskDirection: typeof d.taskDirection === "string" ? d.taskDirection : "",
+    taskName: typeof d.taskName === "string" ? d.taskName : "",
+    taskContent: typeof d.taskContent === "string" ? d.taskContent : "",
+  };
+}
+
+/** 把任务详情整包暂存到 sessionStorage（隐私模式/quota 异常时静默失败）。 */
+export function persistCopTaskDetail(detail: CopTaskDetail): void {
+  try {
+    sessionStorage.setItem(COP_TASK_DETAIL_STORAGE_KEY, JSON.stringify(detail));
+  } catch {
+    /* noop */
+  }
+}
+
+/** 读取暂存的任务详情。 */
+export function readCopTaskDetail(): CopTaskDetail | null {
+  try {
+    const raw = sessionStorage.getItem(COP_TASK_DETAIL_STORAGE_KEY);
+    return raw ? (JSON.parse(raw) as CopTaskDetail) : null;
+  } catch {
+    return null;
+  }
+}
+
+export interface TaskDeeplinkTarget {
+  path: string;
+  /** react-router navigate 的 state（多智能体会商用 roundtablePrefill）。 */
+  state?: Record<string, unknown>;
+}
+
+/** 用 taskName / taskContent 拼出开场信息文本。 */
+function buildOpeningText(detail: CopTaskDetail): string {
+  return [
+    detail.taskName ? `任务名称：${detail.taskName}` : "",
+    detail.taskContent ? `任务内容：${detail.taskContent}` : "",
+  ]
+    .filter(Boolean)
+    .join("\n");
+}
+
+/**
+ * 据 goPath + 任务详情解析跳转目标。
+ *   - 多智能体会商：返回带 ``roundtablePrefill``（businessCode 让会商页自动选用配置的业务链条）的 state。
+ *   - 单智能体问答：解析「配置管理」里该业务配置的智能体；把开场信息写入
+ *     ``sessionStorage.pendingAgentMessage``（AgentChatPage 自动发出）。
+ * 无法识别 goPath / 未配置智能体时返回 null（调用方回退到默认首页）。
+ */
+export async function resolveTaskDeeplinkTarget(
+  goPath: string,
+  detail: CopTaskDetail,
+): Promise<TaskDeeplinkTarget | null> {
+  const parsed = parseGoPath(goPath);
+  if (!parsed) return null;
+  const taskBase = BUSINESS_TASK_BASE[parsed.businessCode];
+  if (!taskBase) return null;
+
+  if (parsed.roundtable) {
+    // 多智能体会商：嵌进对应业务的「任务管理」布局；businessCode 让会商页自动选用配置的业务链条。
+    return {
+      path: `${taskBase}/roundtable`,
+      state: {
+        roundtablePrefill: {
+          taskTitle: detail.taskName,
+          taskGoal: detail.taskContent,
+          businessCode: parsed.businessCode,
+          autoStart: true,
+        },
+      },
+    };
+  }
+
+  let agentId: string | null = null;
+  try {
+    const mappings = await listBusinessMappings();
+    agentId = mappings.find((m) => m.businessCode === parsed.businessCode)?.agentId ?? null;
+  } catch {
+    agentId = null;
+  }
+  if (!agentId) return null;
+
+  try {
+    sessionStorage.setItem(
+      "pendingAgentMessage",
+      JSON.stringify({ agentId, text: buildOpeningText(detail) }),
+    );
+  } catch {
+    /* noop */
+  }
+  // 单智能体问答：同样嵌进对应业务的「任务管理」布局。
+  return { path: `${taskBase}/agent/${agentId}/chats/new` };
+}
diff --git a/frontend-web/src/core/knowledge/api.ts b/frontend-web/src/core/knowledge/api.ts
index 182f34f5..32968ecf 100644
--- a/frontend-web/src/core/knowledge/api.ts
+++ b/frontend-web/src/core/knowledge/api.ts
@@ -4,8 +4,11 @@ import { getBackendBaseURL } from "@/core/config";
 import type {
   BatchIngestPayload,
   BatchIngestResult,
+  ExtractTemplateScope,
   GraphExportResult,
   IngestThreadPayload,
+  KnowledgeExtractTemplate,
+  KnowledgeFolder,
   KnowledgeListParams,
   KnowledgeListResponse,
   KnowledgeNote,
@@ -35,6 +38,7 @@ export async function listKnowledgeNotes(params: KnowledgeListParams = {}) {
   if (params.entity) qs.set("entity", params.entity);
   if (params.source_type) qs.set("source_type", params.source_type);
   if (params.status) qs.set("status", params.status);
+  if (params.folder !== undefined) qs.set("folder", params.folder);
   qs.set("limit", String(params.limit ?? 20));
   qs.set("offset", String(params.offset ?? 0));
   return parseResponse<KnowledgeListResponse>(
@@ -54,6 +58,7 @@ export async function createKnowledgeNote(data: {
   summary?: string | null;
   tags?: string[];
   status?: string;
+  folder?: string | null;
 }) {
   const res = await parseResponse<{ note: KnowledgeNote }>(
     await apiFetch(`${baseUrl()}/notes`, {
@@ -73,6 +78,7 @@ export async function updateKnowledgeNote(
     summary: string | null;
     tags: string[];
     status: string;
+    folder: string | null;
   }>,
 ) {
   return parseResponse<KnowledgeNote>(
@@ -138,7 +144,14 @@ export interface IngestFileResult {
 /** Import a file (md/txt/pdf/word/ppt/excel) into the KB as wiki note(s). */
 export async function ingestKnowledgeFile(
   file: File,
-  opts: { title?: string; tags?: string; status?: string; model_name?: string | null } = {},
+  opts: {
+    title?: string;
+    tags?: string;
+    status?: string;
+    model_name?: string | null;
+    template_id?: string | null;
+    folder?: string | null;
+  } = {},
 ) {
   const form = new FormData();
   form.append("file", file);
@@ -146,6 +159,8 @@ export async function ingestKnowledgeFile(
   if (opts.tags) form.append("tags", opts.tags);
   if (opts.status) form.append("status", opts.status);
   if (opts.model_name) form.append("model_name", opts.model_name);
+  if (opts.template_id) form.append("template_id", opts.template_id);
+  if (opts.folder) form.append("folder", opts.folder);
   // No Content-Type header — the browser sets the multipart boundary itself.
   return parseResponse<IngestFileResult>(
     await apiFetch(`${baseUrl()}/ingest/file`, { method: "POST", body: form }),
@@ -284,3 +299,98 @@ export async function removeGraphBlacklist(entryId: string) {
     }),
   );
 }
+
+// --- folders (user-managed directories) ---
+
+export async function listKnowledgeFolders() {
+  const res = await parseResponse<{ items: KnowledgeFolder[] }>(
+    await apiFetch(`${baseUrl()}/folders`),
+  );
+  return res.items;
+}
+
+export async function createKnowledgeFolder(path: string) {
+  const res = await parseResponse<{ folder: KnowledgeFolder }>(
+    await apiFetch(`${baseUrl()}/folders`, {
+      method: "POST",
+      headers: { "Content-Type": "application/json" },
+      body: JSON.stringify({ path }),
+    }),
+  );
+  return res.folder;
+}
+
+export async function renameKnowledgeFolder(folderId: string, path: string) {
+  const res = await parseResponse<{ folder: KnowledgeFolder }>(
+    await apiFetch(`${baseUrl()}/folders/${encodeURIComponent(folderId)}`, {
+      method: "PUT",
+      headers: { "Content-Type": "application/json" },
+      body: JSON.stringify({ path }),
+    }),
+  );
+  return res.folder;
+}
+
+export async function deleteKnowledgeFolder(folderId: string, reassignTo?: string) {
+  const qs = reassignTo ? `?reassign_to=${encodeURIComponent(reassignTo)}` : "";
+  return parseResponse<{ ok: boolean }>(
+    await apiFetch(`${baseUrl()}/folders/${encodeURIComponent(folderId)}${qs}`, {
+      method: "DELETE",
+    }),
+  );
+}
+
+// --- extraction prompt templates ---
+
+export async function listExtractTemplates(scope?: ExtractTemplateScope, enabledOnly = false) {
+  const qs = new URLSearchParams();
+  if (scope) qs.set("scope", scope);
+  if (enabledOnly) qs.set("enabled_only", "true");
+  const q = qs.toString();
+  const res = await parseResponse<{ items: KnowledgeExtractTemplate[] }>(
+    await apiFetch(`${baseUrl()}/extract-templates${q ? `?${q}` : ""}`),
+  );
+  return res.items;
+}
+
+export type ExtractTemplateInput = {
+  name: string;
+  system_prompt: string;
+  scope?: ExtractTemplateScope;
+  description?: string | null;
+  is_default?: boolean;
+  enabled?: boolean;
+};
+
+export async function createExtractTemplate(data: ExtractTemplateInput) {
+  const res = await parseResponse<{ template: KnowledgeExtractTemplate }>(
+    await apiFetch(`${baseUrl()}/extract-templates`, {
+      method: "POST",
+      headers: { "Content-Type": "application/json" },
+      body: JSON.stringify(data),
+    }),
+  );
+  return res.template;
+}
+
+export async function updateExtractTemplate(
+  templateId: string,
+  data: Partial<ExtractTemplateInput>,
+) {
+  const res = await parseResponse<{ template: KnowledgeExtractTemplate }>(
+    await apiFetch(`${baseUrl()}/extract-templates/${encodeURIComponent(templateId)}`, {
+      method: "PUT",
+      headers: { "Content-Type": "application/json" },
+      body: JSON.stringify(data),
+    }),
+  );
+  return res.template;
+}
+
+export async function deleteExtractTemplate(templateId: string) {
+  return parseResponse<{ ok: boolean }>(
+    await apiFetch(`${baseUrl()}/extract-templates/${encodeURIComponent(templateId)}`, {
+      method: "DELETE",
+    }),
+  );
+}
diff --git a/frontend-web/src/core/knowledge/hooks.ts b/frontend-web/src/core/knowledge/hooks.ts
index 2c2e1b46..b5a7eac4 100644
--- a/frontend-web/src/core/knowledge/hooks.ts
+++ b/frontend-web/src/core/knowledge/hooks.ts
@@ -7,7 +7,11 @@ import {
 
 import {
   addGraphBlacklist,
+  createExtractTemplate,
+  createKnowledgeFolder,
   createKnowledgeNote,
+  deleteExtractTemplate,
+  deleteKnowledgeFolder,
   deleteKnowledgeNote,
   exportKnowledgeGraph,
   fetchVaultPage,
@@ -18,14 +22,18 @@ import {
   ingestThread,
   ingestThreadsBatch,
   listDuplicates,
+  listExtractTemplates,
+  listKnowledgeFolders,
   listKnowledgeNotes,
   mergeKnowledgeNotes,
   removeGraphBlacklist,
+  renameKnowledgeFolder,
   resolveNoteWikilinks,
   searchKnowledge,
+  updateExtractTemplate,
   updateKnowledgeNote,
 } from "./api";
-import type { KnowledgeListParams } from "./types";
+import type { ExtractTemplateScope, KnowledgeListParams } from "./types";
 
 const ROOT_KEY = ["knowledge"] as const;
 
@@ -172,3 +180,62 @@ export function useKnowledgeDuplicates(enabled: boolean) {
     refetchOnWindowFocus: false,
   });
 }
+
+// --- folders (user-managed directories) ---
+
+export function useKnowledgeFolders() {
+  return useQuery({
+    queryKey: [...ROOT_KEY, "folders"],
+    queryFn: listKnowledgeFolders,
+    refetchOnWindowFocus: false,
+  });
+}
+
+export function useFolderActions() {
+  const queryClient = useQueryClient();
+  // Folder changes can move notes around → invalidate the whole KB tree.
+  const invalidate = () => queryClient.invalidateQueries({ queryKey: ROOT_KEY });
+  return {
+    create: useMutation({ mutationFn: createKnowledgeFolder, onSuccess: invalidate }),
+    rename: useMutation({
+      mutationFn: ({ folderId, path }: { folderId: string; path: string }) =>
+        renameKnowledgeFolder(folderId, path),
+      onSuccess: invalidate,
+    }),
+    remove: useMutation({
+      mutationFn: ({ folderId, reassignTo }: { folderId: string; reassignTo?: string }) =>
+        deleteKnowledgeFolder(folderId, reassignTo),
+      onSuccess: invalidate,
+    }),
+  };
+}
+
+// --- extraction prompt templates ---
+
+export function useExtractTemplates(scope?: ExtractTemplateScope, enabledOnly = false) {
+  return useQuery({
+    queryKey: [...ROOT_KEY, "extract-templates", scope ?? "all", enabledOnly],
+    queryFn: () => listExtractTemplates(scope, enabledOnly),
+    refetchOnWindowFocus: false,
+  });
+}
+
+export function useExtractTemplateActions() {
+  const queryClient = useQueryClient();
+  const invalidate = () =>
+    queryClient.invalidateQueries({ queryKey: [...ROOT_KEY, "extract-templates"] });
+  return {
+    create: useMutation({ mutationFn: createExtractTemplate, onSuccess: invalidate }),
+    update: useMutation({
+      mutationFn: ({
+        templateId,
+        data,
+      }: {
+        templateId: string;
+        data: Parameters<typeof updateExtractTemplate>[1];
+      }) => updateExtractTemplate(templateId, data),
+      onSuccess: invalidate,
+    }),
+    remove: useMutation({ mutationFn: deleteExtractTemplate, onSuccess: invalidate }),
+  };
+}
diff --git a/frontend-web/src/core/knowledge/types.ts b/frontend-web/src/core/knowledge/types.ts
index 7699dae4..d43ed079 100644
--- a/frontend-web/src/core/knowledge/types.ts
+++ b/frontend-web/src/core/knowledge/types.ts
@@ -31,6 +31,8 @@ export interface KnowledgeNote {
   summary?: string | null;
   content_md: string;
   vault_path?: string | null;
+  /** User-assigned directory (slash-joined path); null/undefined = unfiled. */
+  folder?: string | null;
   source_type: string;
   source_id?: string | null;
   status: KnowledgeStatus;
@@ -58,10 +60,37 @@ export interface KnowledgeListParams {
   entity?: string;
   source_type?: string;
   status?: string;
+  /** Filter to a directory (and its sub-folders); "" = unfiled only. */
+  folder?: string;
   limit?: number;
   offset?: number;
 }
 
+/** A user-managed knowledge directory. */
+export interface KnowledgeFolder {
+  id: string;
+  path: string;
+  name: string;
+  sort_order?: number;
+}
+
+export type ExtractTemplateScope = "thread" | "document" | "both";
+
+/** An editable knowledge-extraction prompt template. */
+export interface KnowledgeExtractTemplate {
+  id: string;
+  name: string;
+  description?: string | null;
+  scope: ExtractTemplateScope;
+  system_prompt: string;
+  is_default: boolean;
+  enabled: boolean;
+  created_by?: string | null;
+  updated_by?: string | null;
+  created_at?: string | null;
+  updated_at?: string | null;
+}
+
 export type WikilinkKind = "note" | "vault" | "missing";
 
 export interface WikilinkResolution {
@@ -105,6 +134,10 @@ export interface IngestThreadPayload {
   background?: boolean;
   /** Optional model override for knowledge extraction (null = configured default). */
   model_name?: string | null;
+  /** Optional extraction-prompt template id (null = scope default). */
+  template_id?: string | null;
+  /** Optional directory to file the result under (slash-joined path). */
+  folder?: string | null;
 }
 
 export interface BatchIngestPayload {
@@ -114,12 +147,25 @@ export interface BatchIngestPayload {
   include_sources?: boolean;
   skip_existing?: boolean;
   mode?: "summary" | "full";
+  /**
+   * When true (default), the backend runs the per-thread extraction loop in a
+   * background task and returns immediately (`queued:true`) instead of blocking
+   * the request until every thread is processed — which otherwise outlives the
+   * nginx gateway timeout and surfaces as a generic "request failed".
+   */
+  background?: boolean;
   /** Optional model override for knowledge extraction (null = configured default). */
   model_name?: string | null;
+  /** Optional extraction-prompt template id (null = scope default). */
+  template_id?: string | null;
+  /** Optional directory to file results under (slash-joined path). */
+  folder?: string | null;
 }
 
 export interface BatchIngestResult {
   ok: boolean;
+  /** True when the backend queued the batch as a background task (see payload). */
+  queued?: boolean;
   processed: number;
   created: number;
   skipped: number;
diff --git a/frontend-web/src/core/page-layout/sidebar-menu.ts b/frontend-web/src/core/page-layout/sidebar-menu.ts
index 25f2a51f..ca1c76ff 100644
--- a/frontend-web/src/core/page-layout/sidebar-menu.ts
+++ b/frontend-web/src/core/page-layout/sidebar-menu.ts
@@ -13,6 +13,7 @@ import {
   BookOpen,
   Bot,
   Brain,
+  Briefcase,
   CalendarClock,
   CircleHelp,
   ClipboardList,
@@ -52,7 +53,7 @@ export type MenuItem = {
 export const sidebarMenuItems: MenuItem[] = [
   {
     id: "qa-center",
-    label: "问答中心",
+    label: "问答管理",
     icon: HelpCircle,
     children: [
       { id: "general-qa", label: "通用问答", icon: CircleHelp, path: "/page/workspace/chats/new" },
@@ -65,11 +66,11 @@ export const sidebarMenuItems: MenuItem[] = [
       // { id: "expert-qa", label: "专家问答", icon: Brain, path: "/page/strategy/qa/expert-qa" },
       {
         id: "qa-zzjc",
-        label: "zzJC支持",
+        label: "JC支持",
         icon: Bot,
-        path: "/page/strategy/agents/6bf/chats/new",
+        path: "/page/strategy/qa/zzjc",
       },
-      { id: "qa-zzxd", label: "zzXD应用", icon: Layout, path: "/page/strategy/agents/10d69686718349688e58a8790cb61024/chats/new", },
+      { id: "qa-zzxd", label: "XD应用", icon: Layout, path: "/page/strategy/qa/zzxd" },
       { id: "qa-research", label: "课题研究", icon: FileText, path: "/page/canvas/ai-writing" },
     ],
   },
@@ -102,24 +103,24 @@ export const sidebarMenuItems: MenuItem[] = [
         label: "COH中任务",
         icon: Activity,
         children: [
-          {
-            id: "coh-task-3qfx",
-            label: "3QFX",
-            icon: ListTodo,
-            path: "/page/strategy/task-management/plan-task-sentiment",
-          },
-          {
-            id: "coh-task-fx",
-            label: "任务FX",
-            icon: ListTodo,
-            path: "/page/strategy/task-management/plan-task/plan-task-fx",
-          },
-          {
-            id: "coh-task-xdfx",
-            label: "XDFX",
-            icon: ListTodo,
-            path: "/page/strategy/task-management/plan-task/plan-task-action",
-          },
+          // {
+          //   id: "coh-task-3qfx",
+          //   label: "3QFX",
+          //   icon: ListTodo,
+          //   path: "/page/strategy/task-management/plan-task-sentiment",
+          // },
+          // {
+          //   id: "coh-task-fx",
+          //   label: "任务FX",
+          //   icon: ListTodo,
+          //   path: "/page/strategy/task-management/plan-task/plan-task-fx",
+          // },
+          // {
+          //   id: "coh-task-xdfx",
+          //   label: "XDFX",
+          //   icon: ListTodo,
+          //   path: "/page/strategy/task-management/plan-task/plan-task-action",
+          // },
         ],
       },
       {
@@ -241,6 +242,13 @@ export const sidebarMenuItems: MenuItem[] = [
         path: "/page/strategy/admin/users",
         adminOnly: true,
       },
+      {
+        id: "mgmt-positions",
+        label: "岗位管理",
+        icon: Briefcase,
+        path: "/page/strategy/admin/positions",
+        adminOnly: true,
+      },
       {
         id: "mgmt-admin-scheduled-tasks",
         label: "定时任务管理",
diff --git a/frontend-web/src/core/positions/api.ts b/frontend-web/src/core/positions/api.ts
new file mode 100644
index 00000000..be1a8521
--- /dev/null
+++ b/frontend-web/src/core/positions/api.ts
@@ -0,0 +1,130 @@
+import { apiFetch } from "@/core/api";
+import { getBackendBaseURL } from "@/core/config";
+
+import type {
+  Position,
+  PositionCreateInput,
+  PositionListResponse,
+  PositionMember,
+  PositionUpdateInput,
+  PositionUser,
+  TagBindings,
+  TagBindingsResponse,
+} from "./types";
+
+async function parseDetail(res: Response, fallback: string): Promise<Error> {
+  const data = (await res.json().catch(() => ({}))) as { detail?: unknown };
+  const detail = typeof data.detail === "string" && data.detail ? data.detail : null;
+  return new Error(detail ?? `${fallback} (HTTP ${res.status})`);
+}
+
+const base = () => `${getBackendBaseURL()}/api/positions`;
+
+export async function listPositions(): Promise<Position[]> {
+  const res = await apiFetch(base());
+  if (!res.ok) throw await parseDetail(res, "Failed to list positions");
+  const data = (await res.json()) as PositionListResponse;
+  return data.positions;
+}
+
+export async function createPosition(input: PositionCreateInput): Promise<Position> {
+  const res = await apiFetch(base(), {
+    method: "POST",
+    headers: { "Content-Type": "application/json" },
+    body: JSON.stringify(input),
+  });
+  if (!res.ok) throw await parseDetail(res, "Failed to create position");
+  return (await res.json()) as Position;
+}
+
+export async function updatePosition(id: string, input: PositionUpdateInput): Promise<Position> {
+  const res = await apiFetch(`${base()}/${encodeURIComponent(id)}`, {
+    method: "PUT",
+    headers: { "Content-Type": "application/json" },
+    body: JSON.stringify(input),
+  });
+  if (!res.ok) throw await parseDetail(res, "Failed to update position");
+  return (await res.json()) as Position;
+}
+
+export async function deletePosition(id: string): Promise<void> {
+  const res = await apiFetch(`${base()}/${encodeURIComponent(id)}`, { method: "DELETE" });
+  if (!res.ok) throw await parseDetail(res, "Failed to delete position");
+}
+
+export async function getTagBindings(positionId: string): Promise<TagBindings> {
+  const res = await apiFetch(`${base()}/${encodeURIComponent(positionId)}/tag-bindings`);
+  if (!res.ok) throw await parseDetail(res, "Failed to load tag bindings");
+  const data = (await res.json()) as TagBindingsResponse;
+  return data.bindings;
+}
+
+export async function setTagBindings(positionId: string, bindings: TagBindings): Promise<TagBindings> {
+  const res = await apiFetch(`${base()}/${encodeURIComponent(positionId)}/tag-bindings`, {
+    method: "PUT",
+    headers: { "Content-Type": "application/json" },
+    body: JSON.stringify({ bindings }),
+  });
+  if (!res.ok) throw await parseDetail(res, "Failed to save tag bindings");
+  const data = (await res.json()) as TagBindingsResponse;
+  return data.bindings;
+}
+
+export async function listMembers(positionId: string): Promise<PositionMember[]> {
+  const res = await apiFetch(`${base()}/${encodeURIComponent(positionId)}/members`);
+  if (!res.ok) throw await parseDetail(res, "Failed to list members");
+  const data = (await res.json()) as { members: PositionMember[] };
+  return data.members;
+}
+
+export async function addMembers(positionId: string, userIds: string[]): Promise<PositionMember[]> {
+  const res = await apiFetch(`${base()}/${encodeURIComponent(positionId)}/members`, {
+    method: "POST",
+    headers: { "Content-Type": "application/json" },
+    body: JSON.stringify({ user_ids: userIds }),
+  });
+  if (!res.ok) throw await parseDetail(res, "Failed to add members");
+  const data = (await res.json()) as { members: PositionMember[] };
+  return data.members;
+}
+
+export async function removeMembers(positionId: string, userIds: string[]): Promise<PositionMember[]> {
+  const res = await apiFetch(`${base()}/${encodeURIComponent(positionId)}/members`, {
+    method: "DELETE",
+    headers: { "Content-Type": "application/json" },
+    body: JSON.stringify({ user_ids: userIds }),
+  });
+  if (!res.ok) throw await parseDetail(res, "Failed to remove members");
+  const data = (await res.json()) as { members: PositionMember[] };
+  return data.members;
+}
+
+export async function resyncPosition(positionId: string): Promise<number> {
+  const res = await apiFetch(`${base()}/${encodeURIComponent(positionId)}/resync`, { method: "POST" });
+  if (!res.ok) throw await parseDetail(res, "Failed to resync position");
+  const data = (await res.json()) as { synced: number };
+  return data.synced;
+}
+
+export async function getMyPosition(): Promise<Position | null> {
+  const res = await apiFetch(`${base()}/me`);
+  if (!res.ok) throw await parseDetail(res, "Failed to load my position");
+  const data = (await res.json()) as { position: Position | null };
+  return data.position ?? null;
+}
+
+export async function listPositionUsers(): Promise<PositionUser[]> {
+  const res = await apiFetch(`${base()}/users`);
+  if (!res.ok) throw await parseDetail(res, "Failed to list users");
+  const data = (await res.json()) as { users: PositionUser[] };
+  return data.users;
+}
+
+export async function setUserPosition(userId: string, positionId: string | null): Promise<void> {
+  const res = await apiFetch(`${base()}/users/${encodeURIComponent(userId)}`, {
+    method: "PUT",
+    headers: { "Content-Type": "application/json" },
+    body: JSON.stringify({ position_id: positionId }),
+  });
+  if (!res.ok) throw await parseDetail(res, "Failed to set user position");
+}
diff --git a/frontend-web/src/core/positions/hooks.ts b/frontend-web/src/core/positions/hooks.ts
new file mode 100644
index 00000000..ed0835f3
--- /dev/null
+++ b/frontend-web/src/core/positions/hooks.ts
@@ -0,0 +1,131 @@
+import { useMutation, useQuery, useQueryClient } from "@tanstack/react-query";
+
+import {
+  addMembers,
+  createPosition,
+  deletePosition,
+  getMyPosition,
+  getTagBindings,
+  listMembers,
+  listPositionUsers,
+  listPositions,
+  removeMembers,
+  resyncPosition,
+  setTagBindings,
+  setUserPosition,
+  updatePosition,
+} from "./api";
+import type {
+  PositionCreateInput,
+  PositionUpdateInput,
+  TagBindings,
+} from "./types";
+
+const positionKeys = {
+  root: ["positions"] as const,
+  list: ["positions", "list"] as const,
+  bindings: (id: string) => ["positions", "bindings", id] as const,
+  members: (id: string) => ["positions", "members", id] as const,
+  users: ["positions", "users"] as const,
+  me: ["positions", "me"] as const,
+};
+
+export function usePositions() {
+  return useQuery({ queryKey: positionKeys.list, queryFn: listPositions });
+}
+
+/** The caller's own effective position (explicit, else default). */
+export function useMyPosition() {
+  return useQuery({ queryKey: positionKeys.me, queryFn: getMyPosition });
+}
+
+export function usePositionUsers() {
+  return useQuery({ queryKey: positionKeys.users, queryFn: listPositionUsers });
+}
+
+export function usePositionTagBindings(positionId: string | null) {
+  return useQuery({
+    queryKey: positionKeys.bindings(positionId ?? ""),
+    queryFn: () => getTagBindings(positionId as string),
+    enabled: !!positionId,
+  });
+}
+
+export function usePositionMembers(positionId: string | null) {
+  return useQuery({
+    queryKey: positionKeys.members(positionId ?? ""),
+    queryFn: () => listMembers(positionId as string),
+    enabled: !!positionId,
+  });
+}
+
+function useInvalidate() {
+  const qc = useQueryClient();
+  return () => void qc.invalidateQueries({ queryKey: positionKeys.root });
+}
+
+export function useCreatePosition() {
+  const invalidate = useInvalidate();
+  return useMutation({
+    mutationFn: (input: PositionCreateInput) => createPosition(input),
+    onSuccess: invalidate,
+  });
+}
+
+export function useUpdatePosition() {
+  const invalidate = useInvalidate();
+  return useMutation({
+    mutationFn: ({ id, input }: { id: string; input: PositionUpdateInput }) => updatePosition(id, input),
+    onSuccess: invalidate,
+  });
+}
+
+export function useDeletePosition() {
+  const invalidate = useInvalidate();
+  return useMutation({
+    mutationFn: (id: string) => deletePosition(id),
+    onSuccess: invalidate,
+  });
+}
+
+export function useSetTagBindings() {
+  const qc = useQueryClient();
+  return useMutation({
+    mutationFn: ({ positionId, bindings }: { positionId: string; bindings: TagBindings }) =>
+      setTagBindings(positionId, bindings),
+    onSuccess: (_data, { positionId }) => {
+      void qc.invalidateQueries({ queryKey: positionKeys.bindings(positionId) });
+    },
+  });
+}
+
+export function useAddMembers() {
+  const qc = useQueryClient();
+  return useMutation({
+    mutationFn: ({ positionId, userIds }: { positionId: string; userIds: string[] }) =>
+      addMembers(positionId, userIds),
+    onSuccess: () => void qc.invalidateQueries({ queryKey: positionKeys.root }),
+  });
+}
+
+export function useRemoveMembers() {
+  const qc = useQueryClient();
+  return useMutation({
+    mutationFn: ({ positionId, userIds }: { positionId: string; userIds: string[] }) =>
+      removeMembers(positionId, userIds),
+    onSuccess: () => void qc.invalidateQueries({ queryKey: positionKeys.root }),
+  });
+}
+
+export function useResyncPosition() {
+  return useMutation({ mutationFn: (positionId: string) => resyncPosition(positionId) });
+}
+
+export function useSetUserPosition() {
+  const qc = useQueryClient();
+  return useMutation({
+    mutationFn: ({ userId, positionId }: { userId: string; positionId: string | null }) =>
+      setUserPosition(userId, positionId),
+    onSuccess: () => void qc.invalidateQueries({ queryKey: positionKeys.root }),
+  });
+}
diff --git a/frontend-web/src/core/positions/index.ts b/frontend-web/src/core/positions/index.ts
new file mode 100644
index 00000000..ad2dff6f
--- /dev/null
+++ b/frontend-web/src/core/positions/index.ts
@@ -0,0 +1,3 @@
+export * from "./types";
+export * from "./api";
+export * from "./hooks";
diff --git a/frontend-web/src/core/positions/types.ts b/frontend-web/src/core/positions/types.ts
new file mode 100644
index 00000000..3e3ff5f1
--- /dev/null
+++ b/frontend-web/src/core/positions/types.ts
@@ -0,0 +1,82 @@
+import type { TagType } from "@/core/tags/types";
+
+/** Resource types a 岗位 (position) can grant, by tag. */
+export type PositionResourceType =
+  | "agent"
+  | "skill"
+  | "scheduled_task"
+  | "recommended_question";
+
+export const POSITION_RESOURCE_TYPES: PositionResourceType[] = [
+  "agent",
+  "skill",
+  "scheduled_task",
+  "recommended_question",
+];
+
+// A position's resource types are exactly the tag types — reuse the labels.
+export const POSITION_RESOURCE_LABELS: Record<PositionResourceType, string> = {
+  agent: "智能体",
+  skill: "技能",
+  scheduled_task: "定时任务",
+  recommended_question: "推荐问题",
+};
+
+export const POSITION_RESOURCE_HINT: Record<PositionResourceType, string> = {
+  agent: "不配置标签 = 默认加载全部内置智能体；配置后只加载所选标签下的智能体。",
+  skill: "不配置标签 = 不向「我的技能」添加任何技能；配置后将所选标签下的技能加入成员的「我的技能」（技能广场仍展示全部技能）。",
+  scheduled_task: "不配置标签 = 不订阅任何定时任务；配置后订阅所选标签下的定时任务。",
+  recommended_question: "不配置标签 = 展示全部推荐问题；配置后只展示所选标签下的推荐问题（聊天框上方的推荐问题）。",
+};
+
+export interface Position {
+  id: string;
+  name: string;
+  description: string;
+  is_default: boolean;
+  member_count: number;
+  created_at: string | null;
+  updated_at: string | null;
+}
+
+export interface PositionListResponse {
+  positions: Position[];
+  resource_types: PositionResourceType[];
+}
+
+export interface PositionCreateInput {
+  name: string;
+  description?: string;
+  is_default?: boolean;
+}
+
+export interface PositionUpdateInput {
+  name?: string;
+  description?: string;
+  is_default?: boolean;
+}
+
+/** resource_type -> [tag_id] */
+export type TagBindings = Partial<Record<PositionResourceType, string[]>>;
+
+export interface TagBindingsResponse {
+  position_id: string;
+  bindings: TagBindings;
+}
+
+export interface PositionMember {
+  id: string;
+  email: string;
+}
+
+export interface PositionUser {
+  id: string;
+  email: string;
+  system_role: string;
+  position_id: string | null;
+}
+
+// The resource types double as tag types for the binding pickers.
+export function resourceTypeToTagType(rt: PositionResourceType): TagType {
+  return rt as TagType;
+}
diff --git a/frontend-web/src/core/recommended-questions/api.ts b/frontend-web/src/core/recommended-questions/api.ts
index 0d72dd89..ad840f7f 100644
--- a/frontend-web/src/core/recommended-questions/api.ts
+++ b/frontend-web/src/core/recommended-questions/api.ts
@@ -9,8 +9,10 @@ import type {
 
 const BASE = () => `${getBackendBaseURL()}/api/recommended-questions`;
 
-export async function listRecommendedQuestions(): Promise<RecommendedQuestion[]> {
-  const res = await apiFetch(BASE());
+export async function listRecommendedQuestions(all = false): Promise<RecommendedQuestion[]> {
+  // ``all=true`` is the admin management view: every question, unfiltered by the
+  // caller's 岗位 (position). The chat input uses the default (filtered) view.
+  const res = await apiFetch(`${BASE()}${all ? "?all=1" : ""}`);
   if (!res.ok) throw new Error(`Failed to load questions: ${res.statusText}`);
   const data = (await res.json()) as { questions: RecommendedQuestion[] };
   return data.questions;
diff --git a/frontend-web/src/core/recommended-questions/hooks.ts b/frontend-web/src/core/recommended-questions/hooks.ts
index 538f4a33..b55353ac 100644
--- a/frontend-web/src/core/recommended-questions/hooks.ts
+++ b/frontend-web/src/core/recommended-questions/hooks.ts
@@ -10,11 +10,21 @@ import {
 import type { CreateQuestionRequest, UpdateQuestionRequest } from "./index";
 
 const QK = ["recommended-questions"] as const;
+const ADMIN_QK = ["recommended-questions", "admin"] as const;
 
 export function useRecommendedQuestions() {
   const { data, isLoading, error } = useQuery({
     queryKey: QK,
-    queryFn: listRecommendedQuestions,
+    queryFn: () => listRecommendedQuestions(false),
+  });
+  return { questions: data ?? [], isLoading, error };
+}
+
+/** Admin management view — every question, unfiltered by 岗位. */
+export function useAdminRecommendedQuestions() {
+  const { data, isLoading, error } = useQuery({
+    queryKey: ADMIN_QK,
+    queryFn: () => listRecommendedQuestions(true),
   });
   return { questions: data ?? [], isLoading, error };
 }
diff --git a/frontend-web/src/core/recommended-questions/index.ts b/frontend-web/src/core/recommended-questions/index.ts
index a7a5b286..40c36723 100644
--- a/frontend-web/src/core/recommended-questions/index.ts
+++ b/frontend-web/src/core/recommended-questions/index.ts
@@ -1,8 +1,12 @@
+import type { Tag } from "@/core/tags";
+
 export interface RecommendedQuestion {
   id: string;
   title: string;
   content: string;
   sort_order: number;
+  /** Tags assigned to this question (按岗位分发标签). */
+  tags?: Tag[];
   created_at?: string;
   updated_at?: string;
 }
diff --git a/frontend-web/src/core/settings/local.ts b/frontend-web/src/core/settings/local.ts
index cb78eba0..b9c86665 100644
--- a/frontend-web/src/core/settings/local.ts
+++ b/frontend-web/src/core/settings/local.ts
@@ -7,6 +7,7 @@ export const DEFAULT_LOCAL_SETTINGS: LocalSettings = {
   display: {
     streamSpeedMeterEnabled: true,
     ragModeEnabled: false,
+    contextCompressionEnabled: false,
     citationCleanupWords: [],
     fontScale: 1,
   },
@@ -43,6 +44,12 @@ export interface LocalSettings {
      * messages render tool results with the legacy generic search UI only.
      */
     ragModeEnabled: boolean;
+    /**
+     * Per-user switch for context compression (上下文压缩). When enabled, long
+     * conversations are summarized transiently before the model call to save
+     * tokens. Default off; toggled from the chat input box (below 参考文献).
+     */
+    contextCompressionEnabled: boolean;
     /**
      * Admin-maintained words or short phrases removed from reference snippets
      * and dialogs, one entry per line in the settings UI.
diff --git a/frontend-web/src/core/skills/api.ts b/frontend-web/src/core/skills/api.ts
index fd2947aa..e68c2f89 100644
--- a/frontend-web/src/core/skills/api.ts
+++ b/frontend-web/src/core/skills/api.ts
@@ -9,6 +9,8 @@ export interface ListSkillsParams {
   sort?: "time_desc" | "time_asc" | "activity_desc" | "activity_asc" | "usage_desc" | "usage_asc" | "name_asc" | "name_desc";
   /** Keep only skills filed under this publish square (发布广场). */
   square?: string;
+  /** ``"mine"`` restricts to the skills the caller's 岗位 (position) makes visible. */
+  scope?: "mine";
 }
 
 export async function loadSkills(params: ListSkillsParams = {}) {
@@ -17,6 +19,7 @@ export async function loadSkills(params: ListSkillsParams = {}) {
   for (const id of params.tagIds ?? []) qs.append("tag_ids", id);
   if (params.sort) qs.set("sort", params.sort);
   if (params.square) qs.set("square", params.square);
+  if (params.scope) qs.set("scope", params.scope);
   const query = qs.toString();
   const skills = await apiFetch(
     `${getBackendBaseURL()}/api/skills${query ? `?${query}` : ""}`,
@@ -62,6 +65,26 @@ export async function enableSkill(skillName: string, enabled: boolean) {
   return response.json();
 }
 
+export async function setSkillAlwaysOn(
+  skillName: string,
+  alwaysOn: boolean,
+): Promise<Skill> {
+  const response = await apiFetch(
+    `${getBackendBaseURL()}/api/skills/${encodeURIComponent(skillName)}/always-on`,
+    {
+      method: "PUT",
+      headers: {
+        "Content-Type": "application/json",
+      },
+      body: JSON.stringify({ always_on: alwaysOn }),
+    },
+  );
+  if (!response.ok) {
+    throw await parseDetailedError(response, "Failed to update skill 常驻 flag");
+  }
+  return response.json() as Promise<Skill>;
+}
+
 export async function setSkillDisplay(
   skillName: string,
   display: SkillDisplayConfig | null,
diff --git a/frontend-web/src/core/skills/hooks.ts b/frontend-web/src/core/skills/hooks.ts
index b4c5e4e8..4460242e 100644
--- a/frontend-web/src/core/skills/hooks.ts
+++ b/frontend-web/src/core/skills/hooks.ts
@@ -18,6 +18,7 @@ import {
   installSkillUpload,
   reconcileCustomSkills,
   saveSkillFile,
+  setSkillAlwaysOn,
   setSkillDetail,
   setSkillNameZh,
   setSkillDisplay,
@@ -42,9 +43,10 @@ export function useSkills(params: ListSkillsParams = {}) {
   const search = params.search ?? "";
   const tagIds = params.tagIds ?? [];
   const sort = params.sort ?? "time_desc";
+  const scope = params.scope;
   const { data, isLoading, error } = useQuery({
-    queryKey: ["skills", search, [...tagIds].sort(), sort],
-    queryFn: () => loadSkills({ search, tagIds, sort }),
+    queryKey: ["skills", search, [...tagIds].sort(), sort, scope ?? "all"],
+    queryFn: () => loadSkills({ search, tagIds, sort, scope }),
     placeholderData: keepPreviousData,
   });
   return { skills: data ?? EMPTY_SKILLS, isLoading, error };
@@ -170,6 +172,24 @@ export function useEnableSkill() {
   });
 }
 
+export function useSetSkillAlwaysOn() {
+  const queryClient = useQueryClient();
+  return useMutation({
+    mutationFn: ({ skillName, alwaysOn }: { skillName: string; alwaysOn: boolean }) =>
+      setSkillAlwaysOn(skillName, alwaysOn),
+    onSuccess: (data, { skillName }) => {
+      queryClient.setQueryData(["skill", skillName], data);
+      queryClient.setQueriesData<Skill[]>({ queryKey: ["skills"] }, (current) =>
+        Array.isArray(current)
+          ? current.map((item) => (item.name === skillName ? { ...item, ...data } : item))
+          : current,
+      );
+      void queryClient.invalidateQueries({ queryKey: ["skills"] });
+      void queryClient.invalidateQueries({ queryKey: ["skill", skillName] });
+    },
+  });
+}
+
 export function useValidateSkillUpload() {
   return useMutation({
     mutationFn: (file: File) => validateSkillUpload(file),
diff --git a/frontend-web/src/core/skills/type.ts b/frontend-web/src/core/skills/type.ts
index 00bc6cca..dc0ba9da 100644
--- a/frontend-web/src/core/skills/type.ts
+++ b/frontend-web/src/core/skills/type.ts
@@ -48,6 +48,9 @@ export interface Skill {
   category: string;
   license: string;
   enabled: boolean;
+  // 常驻(免压缩) —— 开启技能压缩后,标记此项的技能仍以全文形式注入系统提示词,
+  // 而非被压缩成索引 + search_skills 检索。
+  always_on?: boolean;
   // Multitenancy fields — owner_user_id is null for built-in / public skills.
   // ``builtin`` mirrors that nullness for explicit branching on the UI side.
   owner_user_id?: string | null;
@@ -58,6 +61,9 @@ export interface Skill {
   /** Publish square (发布广场) the skill is filed under once published. "" = default square. */
   square_id?: string;
   builtin?: boolean;
+  // True when this skill is in the caller's 我的技能 collection — e.g. granted by
+  // their 岗位 (position). Owned skills are detected separately via owner_user_id.
+  favorited?: boolean;
   // Long-form detail/introduction authored by the skill owner or an admin.
   // Distinct from ``description`` (parsed from SKILL.md); supports Markdown.
   detail?: string | null;
diff --git a/frontend-web/src/core/system-settings/api.ts b/frontend-web/src/core/system-settings/api.ts
index 20fd57f1..26d854f6 100644
--- a/frontend-web/src/core/system-settings/api.ts
+++ b/frontend-web/src/core/system-settings/api.ts
@@ -247,6 +247,89 @@ export async function updateCitationDisplaySettings(
   return (await response.json()) as CitationDisplaySettings;
 }
 
+export interface SkillCompressionSettings {
+  enabled: boolean;
+  search_top_n: number;
+  keep_index_in_prompt: boolean;
+}
+
+export interface SkillCompressionUpdate {
+  enabled?: boolean;
+  search_top_n?: number;
+  keep_index_in_prompt?: boolean;
+}
+
+export async function loadSkillCompressionSettings(): Promise<SkillCompressionSettings> {
+  const response = await apiFetch(
+    `${getBackendBaseURL()}/api/system-settings/skill-compression`,
+  );
+  if (!response.ok) {
+    throw new Error(
+      `Failed to load skill compression settings: ${response.status}`,
+    );
+  }
+  return (await response.json()) as SkillCompressionSettings;
+}
+
+export async function updateSkillCompressionSettings(
+  body: SkillCompressionUpdate,
+): Promise<SkillCompressionSettings> {
+  const response = await apiFetch(
+    `${getBackendBaseURL()}/api/system-settings/skill-compression`,
+    {
+      method: "PUT",
+      headers: { "Content-Type": "application/json" },
+      body: JSON.stringify(body),
+    },
+  );
+  if (!response.ok) {
+    const detail = await response
+      .json()
+      .catch(() => ({ detail: "Failed to update skill compression" }));
+    throw new Error(detail.detail ?? "Failed to update skill compression");
+  }
+  return (await response.json()) as SkillCompressionSettings;
+}
+
+/** 「按类型聚合」模块（zzJC / zzXD）选中的智能体标签配置（需求 0616 #3/#4）。 */
+export interface AgentModuleConfig {
+  key: string;
+  tag_ids: string[];
+}
+
+export async function loadAgentModuleConfig(
+  key: string,
+): Promise<AgentModuleConfig> {
+  const response = await apiFetch(
+    `${getBackendBaseURL()}/api/system-settings/agent-modules/${encodeURIComponent(key)}`,
+  );
+  if (!response.ok) {
+    throw new Error(`Failed to load agent module config: ${response.status}`);
+  }
+  return (await response.json()) as AgentModuleConfig;
+}
+
+export async function updateAgentModuleConfig(
+  key: string,
+  tagIds: string[],
+): Promise<AgentModuleConfig> {
+  const response = await apiFetch(
+    `${getBackendBaseURL()}/api/system-settings/agent-modules/${encodeURIComponent(key)}`,
+    {
+      method: "PUT",
+      headers: { "Content-Type": "application/json" },
+      body: JSON.stringify({ tag_ids: tagIds }),
+    },
+  );
+  if (!response.ok) {
+    const detail = await response
+      .json()
+      .catch(() => ({ detail: "Failed to update agent module config" }));
+    throw new Error(detail.detail ?? "Failed to update agent module config");
+  }
+  return (await response.json()) as AgentModuleConfig;
+}
+
 export interface LogSystemSettings {
   enabled: boolean;
   url: string;
diff --git a/frontend-web/src/core/system-settings/hooks.ts b/frontend-web/src/core/system-settings/hooks.ts
index 05bf91a0..0060b22a 100644
--- a/frontend-web/src/core/system-settings/hooks.ts
+++ b/frontend-web/src/core/system-settings/hooks.ts
@@ -3,24 +3,29 @@ import { useMutation, useQuery, useQueryClient } from "@tanstack/react-query";
 import { useLocalSettings } from "@/core/settings";
 
 import {
+  loadAgentModuleConfig,
   loadAppearanceSettings,
   loadCitationDisplaySettings,
   loadLeaderboardSettings,
   loadPrivateSquareStatus,
   loadPromptPrefixSettings,
   loadPublishSquares,
+  loadSkillCompressionSettings,
+  updateAgentModuleConfig,
   updateAppearanceSettings,
   updateCitationDisplaySettings,
   updateLeaderboardSettings,
   updatePrivateSquareSettings,
   updatePromptPrefixSettings,
   updatePublishSquares,
+  updateSkillCompressionSettings,
   type AppearanceUpdate,
   type CitationDisplayUpdate,
   type LeaderboardSettingsUpdate,
   type PrivateSquareUpdate,
   type PromptPrefixUpdate,
   type PublishSquaresUpdate,
+  type SkillCompressionUpdate,
   type LayoutMode,
   type ViewMode,
 } from "./api";
@@ -31,6 +36,7 @@ const PUBLISH_SQUARES_KEY = ["systemSettings", "publishSquares"] as const;
 const APPEARANCE_KEY = ["systemSettings", "appearance"] as const;
 const CITATION_DISPLAY_KEY = ["systemSettings", "citationDisplay"] as const;
 const LEADERBOARD_KEY = ["systemSettings", "leaderboard"] as const;
+const SKILL_COMPRESSION_KEY = ["systemSettings", "skillCompression"] as const;
 
 export function usePromptPrefixSettings() {
   return useQuery({
@@ -165,6 +171,29 @@ export function useLayoutMode(): [
   return [mode, setMode];
 }
 
+/** 「按类型聚合」模块（zzJC / zzXD）选中标签配置（需求 0616 #3/#4）。 */
+const agentModuleKey = (key: string) =>
+  ["systemSettings", "agentModule", key] as const;
+
+export function useAgentModuleConfig(key: string) {
+  return useQuery({
+    queryKey: agentModuleKey(key),
+    queryFn: () => loadAgentModuleConfig(key),
+    staleTime: 60_000,
+    refetchOnWindowFocus: false,
+  });
+}
+
+export function useUpdateAgentModuleConfig(key: string) {
+  const queryClient = useQueryClient();
+  return useMutation({
+    mutationFn: (tagIds: string[]) => updateAgentModuleConfig(key, tagIds),
+    onSuccess: (data) => {
+      queryClient.setQueryData(agentModuleKey(key), data);
+    },
+  });
+}
+
 export function useCitationDisplaySettings() {
   return useQuery({
     queryKey: CITATION_DISPLAY_KEY,
@@ -186,6 +215,29 @@ export function useUpdateCitationDisplaySettings() {
   });
 }
 
+/** Skill compression (技能压缩) settings. Readable by any logged-in user — the
+ * skill detail page reads it to decide whether to show the 常驻(免压缩) toggle. */
+export function useSkillCompressionSettings() {
+  return useQuery({
+    queryKey: SKILL_COMPRESSION_KEY,
+    queryFn: () => loadSkillCompressionSettings(),
+    staleTime: 60_000,
+    refetchOnWindowFocus: false,
+    retry: false,
+  });
+}
+
+export function useUpdateSkillCompressionSettings() {
+  const queryClient = useQueryClient();
+  return useMutation({
+    mutationFn: (body: SkillCompressionUpdate) =>
+      updateSkillCompressionSettings(body),
+    onSuccess: (data) => {
+      queryClient.setQueryData(SKILL_COMPRESSION_KEY, data);
+    },
+  });
+}
+
 export function useLeaderboardSettings() {
   return useQuery({
     queryKey: LEADERBOARD_KEY,
diff --git a/frontend-web/src/core/tags/hooks.ts b/frontend-web/src/core/tags/hooks.ts
index 42861feb..950eb082 100644
--- a/frontend-web/src/core/tags/hooks.ts
+++ b/frontend-web/src/core/tags/hooks.ts
@@ -83,6 +83,7 @@ export function useSetResourceTags() {
       void queryClient.invalidateQueries({ queryKey: ["skills"] });
       void queryClient.invalidateQueries({ queryKey: ["agents"] });
       void queryClient.invalidateQueries({ queryKey: ["scheduled-tasks"] });
+      void queryClient.invalidateQueries({ queryKey: ["recommended-questions"] });
     },
   });
 }
diff --git a/frontend-web/src/core/tags/types.ts b/frontend-web/src/core/tags/types.ts
index 17350f3e..a06f4c69 100644
--- a/frontend-web/src/core/tags/types.ts
+++ b/frontend-web/src/core/tags/types.ts
@@ -1,6 +1,23 @@
-export type TagType = "agent" | "skill" | "scheduled_task";
+export type TagType =
+  | "agent"
+  | "skill"
+  | "scheduled_task"
+  | "recommended_question";
 
-export const TAG_TYPES: TagType[] = ["agent", "skill", "scheduled_task"];
+export const TAG_TYPES: TagType[] = [
+  "agent",
+  "skill",
+  "scheduled_task",
+  "recommended_question",
+];
+
+// Human labels for the tag types (used by the tag management + 岗位 pages).
+export const TAG_TYPE_LABELS: Record<TagType, string> = {
+  agent: "智能体",
+  skill: "技能",
+  scheduled_task: "定时任务",
+  recommended_question: "推荐问题",
+};
 
 export interface Tag {
   id: string;
diff --git a/frontend-web/src/core/threads/hooks.ts b/frontend-web/src/core/threads/hooks.ts
index 36891e07..687529cc 100644
--- a/frontend-web/src/core/threads/hooks.ts
+++ b/frontend-web/src/core/threads/hooks.ts
@@ -343,6 +343,23 @@ export function useThreadStream({
       ) {
         const e = event as { type: "llm_retry"; message: string };
         toast(e.message);
+        return;
+      }
+
+      // Context compression about to run a (blocking) summarization LLM call —
+      // surface a transient hint so the wait before the answer streams isn't a
+      // silent dead air. Emitted by DeerFlowSummarizationMiddleware.
+      if (
+        typeof event === "object" &&
+        event !== null &&
+        "type" in event &&
+        event.type === "context_compacting" &&
+        "message" in event &&
+        typeof event.message === "string" &&
+        event.message.trim()
+      ) {
+        const e = event as { type: "context_compacting"; message: string };
+        toast(e.message);
       }
     },
     onError(error) {
@@ -647,6 +664,9 @@ export function useThreadStream({
           lockThreadRagMode(threadId);
         }
 
+        const contextCompressionEnabled =
+          localSettings.display.contextCompressionEnabled ?? false;
+
         const mergedContext = {
           ...context,
           ...extraContext,
@@ -698,6 +718,7 @@ export function useThreadStream({
               memory_injection_enabled:
                 mergedContext.memory_injection_enabled ?? true,
               rag_mode_enabled: ragModeEnabled,
+              summarization_enabled: contextCompressionEnabled,
               thread_id: threadId,
             },
           },
diff --git a/frontend-web/src/core/threads/types.ts b/frontend-web/src/core/threads/types.ts
index 475269a7..905fd77a 100644
--- a/frontend-web/src/core/threads/types.ts
+++ b/frontend-web/src/core/threads/types.ts
@@ -20,6 +20,8 @@ export interface AgentThreadContext extends Record<string, unknown> {
   writing_artifact_path?: string;
   memory_injection_enabled?: boolean;
   rag_mode_enabled?: boolean;
+  /** Per-request switch for context compression (上下文压缩). Default off. */
+  summarization_enabled?: boolean;
   /** Per-conversation switch for knowledge-base RAG injection (phase 2). */
   knowledge_rag_enabled?: boolean;
   agent_id?: string;
diff --git a/frontend-web/src/core/utils/datetime.ts b/frontend-web/src/core/utils/datetime.ts
index 3ec7203d..fea6af99 100644
--- a/frontend-web/src/core/utils/datetime.ts
+++ b/frontend-web/src/core/utils/datetime.ts
@@ -25,3 +25,18 @@ export function formatTimeAgo(date: Date | string | number, locale?: Locale) {
     locale: getDateFnsLocale(effectiveLocale),
   });
 }
+
+/**
+ * Format a created/updated timestamp as `YYYY-MM-DD HH:mm`.
+ * Returns an empty string for missing or unparseable values.
+ */
+export function formatDateTime(value?: Date | string | number | null): string {
+  if (value === null || value === undefined || value === "") return "";
+  const date = value instanceof Date ? value : new Date(value);
+  if (Number.isNaN(date.getTime())) return "";
+  const pad = (n: number) => String(n).padStart(2, "0");
+  return (
+    `${date.getFullYear()}-${pad(date.getMonth() + 1)}-${pad(date.getDate())} ` +
+    `${pad(date.getHours())}:${pad(date.getMinutes())}`
+  );
+}
diff --git a/frontend-web/src/core/utils/markdown.ts b/frontend-web/src/core/utils/markdown.ts
index 96060708..c5b80575 100644
--- a/frontend-web/src/core/utils/markdown.ts
+++ b/frontend-web/src/core/utils/markdown.ts
@@ -100,9 +100,20 @@ function candidateTextFromHref(href: string): string {
 
 /**
  * Injects a small runtime script into generated HTML so that clicking an
- * in-page anchor (`<a href="#section">`) scrolls within the iframe instead of
- * navigating it. The preview iframe loads from a sandboxed `blob:` URL, where a
- * fragment navigation blanks the frame; intercepting the click avoids that.
+ * in-page anchor (`<a href="#section">`) works inside the preview iframe.
+ *
+ * The preview loads from a sandboxed `blob:` URL (opaque origin). Such a frame
+ * is FORBIDDEN from navigating to its own `blob:` URL, so changing the hash
+ * throws "Not allowed to load local resource: blob:…#page-x". And because CSS
+ * `:target` only updates on a real fragment navigation, the html-multipage
+ * skill's `:target`-based sub-page switching can never fire here.
+ *
+ * So we never touch the hash. Instead the handler emulates the skill's
+ * show/hide in JS: a "page" is a `.page` inside a `.pages` container; clicking
+ * a link to one shows it and hides its siblings (mirroring the
+ * `.page:target` + `~ #page-home` rules via inline `display`, which outranks
+ * the skill's class-level CSS). Plain in-page anchors (TOC headings, etc.) just
+ * smooth-scroll as before.
  */
 export function injectInPageAnchorHandler(html: string): string {
   const script = `<script>(function(){
@@ -112,6 +123,22 @@ export function injectInPageAnchorHandler(html: string): string {
     var esc=window.CSS&&CSS.escape?CSS.escape(id):id.replace(/["\\\\]/g,"\\\\$&");
     try{return document.querySelector("#"+esc+",[name=\\""+esc+"\\"]");}catch(e){return null;}
   }
+  function find(raw){
+    var id=decode(raw);
+    return document.getElementById(id)||document.getElementById(raw)||
+      document.getElementsByName(id)[0]||document.getElementsByName(raw)[0]||
+      bySelector(id)||bySelector(raw);
+  }
+  function switchPage(target){
+    var container=target.parentElement;
+    while(container&&!(container.classList&&container.classList.contains("pages"))){container=container.parentElement;}
+    var pages=(container||document).querySelectorAll(".page");
+    if(!pages.length)return false;
+    for(var i=0;i<pages.length;i++){pages[i].style.display="none";}
+    target.style.display="block";
+    if(window.scrollTo){window.scrollTo(0,0);}
+    return true;
+  }
   document.addEventListener("click",function(e){
     var el=e.target;
     while(el&&el.tagName!=="A"){el=el.parentElement;}
@@ -119,16 +146,10 @@ export function injectInPageAnchorHandler(html: string): string {
     var href=el.getAttribute("href");
     if(!href||href.charAt(0)!=="#"||href.length<2)return;
     e.preventDefault();
-    var id=decode(href.slice(1));
-    var raw=href.slice(1);
-    var target=
-      document.getElementById(id)||
-      document.getElementById(raw)||
-      document.getElementsByName(id)[0]||
-      document.getElementsByName(raw)[0]||
-      bySelector(id)||
-      bySelector(raw);
-    if(target){target.scrollIntoView({behavior:"smooth",block:"start"});}
+    var target=find(href.slice(1));
+    if(!target)return;
+    if(target.classList&&target.classList.contains("page")&&switchPage(target))return;
+    target.scrollIntoView({behavior:"smooth",block:"start"});
   },true);
 })();</script>`;
   if (/<\/body>/i.test(html)) {
diff --git a/frontend-web/src/lib/desensitize.ts b/frontend-web/src/lib/desensitize.ts
index 3778cc98..a6e6bd0f 100644
--- a/frontend-web/src/lib/desensitize.ts
+++ b/frontend-web/src/lib/desensitize.ts
@@ -4,6 +4,9 @@
  * value: 脱敏后替换词
  */
 export const DESENSITIZE_MAP: Record<string, string> = {
+  方案: "FA",
+  设计: "SJ",
+  任务: "RW", 
   态势: "TS",
   滨海作战体系: "第一张网",
   滨海作战部队: "第一张网",
@@ -72,6 +75,26 @@ export function desensitize(text: string): string {
   return text.replace(_regex, (matched) => DESENSITIZE_MAP[matched] ?? matched);
 }
 
+/**
+ * 深度脱敏：递归遍历对象 / 数组，对其中所有字符串值做 {@link desensitize}，
+ * 其余类型原样保留。返回**新值**（不修改入参），用于对结构化数据（如大屏 report-json
+ * 解析出的对象）整体脱敏后再渲染。
+ *
+ * 注意：仅用于「展示」路径——不要用它改写要回传后端 / 落库的数据。
+ */
+export function desensitizeDeep<T>(value: T): T {
+  if (typeof value === "string") return desensitize(value) as unknown as T;
+  if (Array.isArray(value)) return value.map((v) => desensitizeDeep(v)) as unknown as T;
+  if (value && typeof value === "object") {
+    const out: Record<string, unknown> = {};
+    for (const [k, v] of Object.entries(value as Record<string, unknown>)) {
+      out[k] = desensitizeDeep(v);
+    }
+    return out as T;
+  }
+  return value;
+}
+
 /**
  * 运行时动态追加脱敏词，追加后自动重编译正则。
  *
diff --git a/frontend-web/src/open-canvas/api/ai-writing-research-templates.ts b/frontend-web/src/open-canvas/api/ai-writing-research-templates.ts
new file mode 100644
index 00000000..05d9e1a8
--- /dev/null
+++ b/frontend-web/src/open-canvas/api/ai-writing-research-templates.ts
@@ -0,0 +1,75 @@
+import { getBackendBaseURL } from '@/core/config';
+import { apiFetch } from '@/core/api';
+
+/** 课题研究模板：一组写作表单预设（需求 0616 #5）。 */
+export interface ResearchTemplate {
+  id: string;
+  name: string;
+  description: string | null;
+  /** 引用的文章类型 key（对应 ArticleType.key）；null = 不设。 */
+  article_type_key: string | null;
+  outline: string | null;
+  material_source: 'general' | 'notebook';
+  word_count: number;
+  audience: string | null;
+  strict_mode: boolean;
+  topic: string | null;
+  sort_order: number;
+  created_at: string;
+  updated_at: string;
+}
+
+export interface CreateResearchTemplateRequest {
+  name: string;
+  description?: string | null;
+  article_type_key?: string | null;
+  outline?: string | null;
+  material_source?: 'general' | 'notebook';
+  word_count?: number;
+  audience?: string | null;
+  strict_mode?: boolean;
+  topic?: string | null;
+  sort_order?: number;
+}
+
+export type UpdateResearchTemplateRequest = Partial<CreateResearchTemplateRequest>;
+
+const BASE = () => `${getBackendBaseURL()}/api/ai-writing/research-templates`;
+
+export async function listResearchTemplates(): Promise<ResearchTemplate[]> {
+  const resp = await apiFetch(BASE());
+  if (!resp.ok) throw new Error(`获取研究模板失败: ${resp.status}`);
+  return resp.json();
+}
+
+export async function createResearchTemplate(
+  data: CreateResearchTemplateRequest,
+): Promise<ResearchTemplate> {
+  const resp = await apiFetch(BASE(), {
+    method: 'POST',
+    headers: { 'Content-Type': 'application/json' },
+    body: JSON.stringify(data),
+  });
+  if (!resp.ok) throw new Error(`创建研究模板失败: ${resp.status}`);
+  return resp.json();
+}
+
+export async function updateResearchTemplate(
+  id: string,
+  data: UpdateResearchTemplateRequest,
+): Promise<ResearchTemplate> {
+  const resp = await apiFetch(`${BASE()}/${id}`, {
+    method: 'PATCH',
+    headers: { 'Content-Type': 'application/json' },
+    body: JSON.stringify(data),
+  });
+  if (resp.status === 404) throw new Error(`研究模板不存在`);
+  if (!resp.ok) throw new Error(`更新研究模板失败: ${resp.status}`);
+  return resp.json();
+}
+
+export async function deleteResearchTemplate(id: string): Promise<void> {
+  const resp = await apiFetch(`${BASE()}/${id}`, { method: 'DELETE' });
+  if (resp.status === 404) throw new Error(`研究模板不存在`);
+  if (!resp.ok) throw new Error(`删除研究模板失败: ${resp.status}`);
+}
diff --git a/frontend-web/src/open-canvas/components/ai-writing/AIWritingForm.tsx b/frontend-web/src/open-canvas/components/ai-writing/AIWritingForm.tsx
index e6ca5d91..aa6a7d60 100644
--- a/frontend-web/src/open-canvas/components/ai-writing/AIWritingForm.tsx
+++ b/frontend-web/src/open-canvas/components/ai-writing/AIWritingForm.tsx
@@ -1,4 +1,4 @@
-import { useEffect, useMemo, useState } from 'react';
+import { useEffect, useMemo } from 'react';
 import { PenLine, SlidersHorizontal } from 'lucide-react';
 
 import { Button } from '@/components/ui/button';
@@ -55,8 +55,6 @@ export function AIWritingForm({
   onSubmit,
   disabled,
   hideAdvancedFields = false,
-  initialNotebookId,
-  initialDocumentIds,
   studioToken,
 }: AIWritingFormProps) {
   const {
@@ -68,13 +66,12 @@ export function AIWritingForm({
     wordCountInput, setWordCountInput,
     strictMode, setStrictMode,
     materialSource, setMaterialSource,
+    // 笔记来源（search_space_id + document_id 列表）已上提到 WritingFormContext，
+    // 以便左栏「笔记本」点击可直接设为素材来源（需求 0616 #5）。
+    selectedNotebookId, setSelectedNotebookId,
+    selectedDocumentIds, setSelectedDocumentIds,
     modelName, setModelName,
   } = useWritingForm();
-  // 笔记来源：选中的 search_space_id 和多个 document_id
-  const [selectedNotebookId, setSelectedNotebookId] = useState<number | null>(
-    initialNotebookId ?? null,
-  );
-  const [selectedDocumentIds, setSelectedDocumentIds] = useState<number[]>(initialDocumentIds ?? []);
 
   const { models, isLoading: modelsLoading } = useModels();
   const { notebooks, isLoading: notebooksLoading } = useStudioNotebooks(!studioToken);
@@ -146,8 +143,10 @@ export function AIWritingForm({
   }
 
   function toggleDocument(docId: number) {
-    setSelectedDocumentIds((prev) =>
-      prev.includes(docId) ? prev.filter((id) => id !== docId) : [...prev, docId],
+    setSelectedDocumentIds(
+      selectedDocumentIds.includes(docId)
+        ? selectedDocumentIds.filter((id) => id !== docId)
+        : [...selectedDocumentIds, docId],
     );
   }
 
diff --git a/frontend-web/src/open-canvas/components/ai-writing/ResearchSidebar.tsx b/frontend-web/src/open-canvas/components/ai-writing/ResearchSidebar.tsx
new file mode 100644
index 00000000..d464b872
--- /dev/null
+++ b/frontend-web/src/open-canvas/components/ai-writing/ResearchSidebar.tsx
@@ -0,0 +1,283 @@
+"use client";
+
+import { BookOpen, FileText, LayoutGrid, Loader2, Plus, Settings2 } from "lucide-react";
+import { useEffect, useState } from "react";
+import { useNavigate } from "react-router-dom";
+import { useQueryClient } from "@tanstack/react-query";
+
+import { Button } from "@/components/ui/button";
+import {
+  Dialog,
+  DialogContent,
+  DialogHeader,
+  DialogTitle,
+} from "@/components/ui/dialog";
+import { getStoredAuth } from "@/core/auth";
+import { createSearchSpace } from "@/core/studio/api";
+import { useStudioNotebooks } from "@/core/studio/hooks";
+import { cn } from "@/lib/utils";
+
+import { useResearchTemplates } from "../../hooks/useResearchTemplates";
+import type { ResearchTemplate } from "../../api/ai-writing-research-templates";
+import { useWritingForm } from "../../contexts/WritingFormContext";
+import { useAIWriting } from "../../contexts/AIWritingContext";
+
+/**
+ * 课题研究左栏（需求 0616 #5）。
+ *  - 上半「研究模板」= 研究模板实体（``useResearchTemplates``）；点击一键预填整张写作表单
+ *    （文章类型 / 大纲 / 搜索来源 / 字数 / 受众 / 严格模式 / 主题）并展开配置。顶部「配置」
+ *    入口进入模板配置页维护这些模板。
+ *  - 下半「笔记本」= 我的空间（``useStudioNotebooks``）；点击在右侧内嵌打开该笔记本页。
+ */
+export function ResearchSidebar({
+  viewingNotebookId,
+  onOpenNotebook,
+  onOpenAllSpaces,
+}: {
+  /** 当前在右侧查看的笔记本 id（高亮用）。 */
+  viewingNotebookId?: number | null;
+  /** 点击笔记本 → 在右侧内嵌打开该笔记本页（替换右侧内容）。 */
+  onOpenNotebook: (notebookId: number) => void;
+  /** 点击「所有空间」→ 在右侧内嵌打开空间列表页。 */
+  onOpenAllSpaces: () => void;
+}) {
+  const navigate = useNavigate();
+  const { templates, refresh } = useResearchTemplates();
+  const { startWriting } = useAIWriting();
+  const {
+    notebooks,
+    isLoading: notebooksLoading,
+    isError: notebooksError,
+  } = useStudioNotebooks();
+  const {
+    setIntent,
+    setArticleType,
+    setOutlineText,
+    setShowOutline,
+    setMaterialSource,
+    setWordCountInput,
+    setAudience,
+    setStrictMode,
+    setFormExpanded,
+  } = useWritingForm();
+  const queryClient = useQueryClient();
+
+  // 新建空间弹窗（与 workspace-nav-studio 同款）。
+  const [createOpen, setCreateOpen] = useState(false);
+  const [name, setName] = useState("");
+  const [creating, setCreating] = useState(false);
+  const [createError, setCreateError] = useState("");
+
+  useEffect(() => {
+    void refresh();
+  }, [refresh]);
+
+  const handleCreate = async () => {
+    const trimmed = name.trim();
+    if (!trimmed) return;
+    setCreating(true);
+    setCreateError("");
+    try {
+      const email = getStoredAuth()?.email ?? "";
+      const token = queryClient.getQueryData<string>(["studio", "token", email]);
+      if (!token) throw new Error("未登录或 token 已过期，请刷新页面");
+      await createSearchSpace(trimmed, token);
+      setCreateOpen(false);
+      setName("");
+      await queryClient.invalidateQueries({ queryKey: ["studio", "notebooks"] });
+    } catch (e) {
+      setCreateError(e instanceof Error ? e.message : "创建失败");
+    } finally {
+      setCreating(false);
+    }
+  };
+
+  // 点击研究模板 → ①回填整张写作表单（保持写作态左栏显示与运行配置一致）
+  // ②直接以模板预设进入 AI 写作流程（startWriting）。
+  const handleTemplateClick = (tpl: ResearchTemplate) => {
+    const intent = tpl.topic?.trim() ?? "";
+    const articleType = tpl.article_type_key || "analysis";
+    const outline = tpl.outline?.trim() || "";
+    const audience = tpl.audience?.trim() || "普通读者";
+    const wordCount = tpl.word_count > 0 ? tpl.word_count : 800;
+    const writingMode = tpl.strict_mode ? "strict" : "loose";
+
+    // 回填表单上下文（写作态左栏据此展示）。
+    if (intent) setIntent(intent);
+    if (tpl.article_type_key) setArticleType(tpl.article_type_key);
+    if (outline) {
+      setOutlineText(outline);
+      setShowOutline(true);
+    }
+    setMaterialSource(tpl.material_source);
+    setWordCountInput(String(wordCount));
+    if (tpl.audience?.trim()) setAudience(audience);
+    setStrictMode(tpl.strict_mode);
+    setFormExpanded(true);
+
+    // 无主题/意图无法发起写作 → 仅回填表单，由用户补充后手动开始。
+    if (!intent) return;
+    void startWriting({
+      userIntent: intent,
+      userOutline: outline || undefined,
+      articleType,
+      targetAudience: audience,
+      wordCountTarget: wordCount,
+      keywordCount: 1,
+      writingMode,
+      materialSource: tpl.material_source,
+    });
+  };
+
+  return (
+    <aside className="flex h-full w-64 shrink-0 flex-col overflow-hidden border-r border-border/70 bg-card/60 backdrop-blur-sm">
+      {/* 上半：研究模板 */}
+      <div className="flex min-h-0 flex-1 flex-col overflow-y-auto px-2 py-3">
+        <div className="px-2 mb-1.5 flex items-center justify-between">
+          <h2 className="text-muted-foreground/70 text-[10px] font-semibold tracking-[0.12em] uppercase select-none">
+            研究模板
+          </h2>
+          <button
+            type="button"
+            onClick={() => navigate("/page/canvas/research-templates")}
+            title="配置研究模板"
+            aria-label="配置研究模板"
+            className="text-muted-foreground/70 hover:text-foreground rounded p-0.5 transition-colors"
+          >
+            <Settings2 className="size-3.5" />
+          </button>
+        </div>
+        {templates.length === 0 ? (
+          <p className="px-2 py-1.5 text-xs text-muted-foreground/70">
+            暂无研究模板，点右上角配置
+          </p>
+        ) : (
+          <div className="flex flex-col gap-1">
+            {templates.map((tpl) => (
+              <button
+                key={tpl.id}
+                type="button"
+                onClick={() => handleTemplateClick(tpl)}
+                title={tpl.description ?? tpl.name}
+                className="flex items-start gap-2 rounded-lg px-2.5 py-1.5 text-left text-muted-foreground transition-colors hover:bg-sidebar-accent/60 hover:text-foreground"
+              >
+                <FileText className="mt-0.5 size-3.5 shrink-0 text-muted-foreground/70" />
+                <span className="min-w-0">
+                  <span className="block truncate text-[13px] font-medium">{tpl.name}</span>
+                  {tpl.description && (
+                    <span className="text-muted-foreground/70 line-clamp-1 block text-[11px]">
+                      {tpl.description}
+                    </span>
+                  )}
+                </span>
+              </button>
+            ))}
+          </div>
+        )}
+      </div>
+
+      {/* 下半：笔记本（我的空间） */}
+      <div className="flex min-h-0 flex-1 flex-col overflow-y-auto border-t border-border/60 px-2 py-3">
+        <h2 className="text-muted-foreground/70 px-2 mb-1.5 text-[10px] font-semibold tracking-[0.12em] uppercase select-none">
+          笔记本
+        </h2>
+        <div className="flex flex-col gap-0.5">
+          {/* 新建空间 */}
+          <button
+            type="button"
+            onClick={() => {
+              setName("");
+              setCreateError("");
+              setCreateOpen(true);
+            }}
+            className="flex items-center gap-2 rounded-lg px-2.5 py-1.5 text-left text-[13px] text-muted-foreground transition-colors hover:bg-sidebar-accent/60 hover:text-foreground"
+          >
+            <Plus className="size-3.5 shrink-0 text-muted-foreground/70" />
+            <span>新建空间</span>
+          </button>
+
+          {notebooksLoading ? (
+            <p className="px-2 py-1.5 text-xs text-muted-foreground/70">加载中…</p>
+          ) : notebooksError ? (
+            <p className="px-2 py-1.5 text-xs text-amber-600">
+              笔记本服务暂不可用（检查空间服务地址 / 账号）
+            </p>
+          ) : notebooks.length === 0 ? (
+            <p className="px-2 py-1.5 text-xs text-muted-foreground/70">暂无笔记本</p>
+          ) : (
+            notebooks.map((nb) => {
+              const active = viewingNotebookId === nb.id;
+              return (
+                <button
+                  key={nb.id}
+                  type="button"
+                  onClick={() => onOpenNotebook(nb.id)}
+                  title={nb.description || nb.name}
+                  className={cn(
+                    "flex items-center gap-2 rounded-lg px-2.5 py-1.5 text-left text-[13px] transition-colors",
+                    active
+                      ? "bg-primary/10 text-primary"
+                      : "text-muted-foreground hover:bg-sidebar-accent/60 hover:text-foreground",
+                  )}
+                >
+                  <BookOpen className="size-3.5 shrink-0 text-muted-foreground/70" />
+                  <span className="truncate">{nb.name}</span>
+                </button>
+              );
+            })
+          )}
+
+          {/* 所有空间 */}
+          <button
+            type="button"
+            onClick={onOpenAllSpaces}
+            className="flex items-center gap-2 rounded-lg px-2.5 py-1.5 text-left text-[13px] text-muted-foreground transition-colors hover:bg-sidebar-accent/60 hover:text-foreground"
+          >
+            <LayoutGrid className="size-3.5 shrink-0 text-muted-foreground/70" />
+            <span>所有空间</span>
+          </button>
+        </div>
+      </div>
+
+      {/* 新建空间弹窗 */}
+      <Dialog open={createOpen} onOpenChange={(v) => !creating && setCreateOpen(v)}>
+        <DialogContent className="sm:max-w-md">
+          <DialogHeader>
+            <DialogTitle>新建空间</DialogTitle>
+          </DialogHeader>
+          <div className="flex flex-col gap-4 pt-2">
+            <input
+              autoFocus
+              value={name}
+              onChange={(e) => setName(e.target.value)}
+              onKeyDown={(e) => {
+                if (e.key === "Enter") void handleCreate();
+              }}
+              placeholder="空间名称"
+              disabled={creating}
+              className="w-full rounded-lg border border-border bg-background px-3 py-2 text-sm outline-none focus:ring-2 focus:ring-primary disabled:opacity-50"
+            />
+            {createError && <p className="text-xs text-destructive">{createError}</p>}
+            <div className="flex justify-end gap-2">
+              <Button
+                variant="outline"
+                size="sm"
+                disabled={creating}
+                onClick={() => setCreateOpen(false)}
+              >
+                取消
+              </Button>
+              <Button
+                size="sm"
+                disabled={!name.trim() || creating}
+                onClick={() => void handleCreate()}
+              >
+                {creating ? <Loader2 className="size-4 animate-spin" /> : "创建"}
+              </Button>
+            </div>
+          </div>
+        </DialogContent>
+      </Dialog>
+    </aside>
+  );
+}
diff --git a/frontend-web/src/open-canvas/components/ai-writing/WritingSetupChat.tsx b/frontend-web/src/open-canvas/components/ai-writing/WritingSetupChat.tsx
index b1207024..e620b0e2 100644
--- a/frontend-web/src/open-canvas/components/ai-writing/WritingSetupChat.tsx
+++ b/frontend-web/src/open-canvas/components/ai-writing/WritingSetupChat.tsx
@@ -28,7 +28,7 @@ import { uuid } from '@/core/utils/uuid';
 import { cn } from '@/lib/utils';
 
 import { useArticleTypes } from '../../hooks/useArticleTypes';
-import { useWritingForm } from '../../contexts/WritingFormContext';
+import { useWritingFormOptional } from '../../contexts/WritingFormContext';
 import { WritingWelcome } from './WritingWelcome';
 
 /**
@@ -42,13 +42,25 @@ import { WritingWelcome } from './WritingWelcome';
  *
  * 这条线程只负责「帮用户把表单聊清」，与真正写作的 ai_writing graph 完全解耦。
  */
-export function WritingSetupChat() {
+export function WritingSetupChat({
+  onApprove: onApproveProp,
+  title = '写作助手',
+  hint = '描述你想写什么，我来帮你填好右侧配置',
+}: {
+  /** 审批卡「应用」回调；缺省回填写作表单（需 WritingFormProvider）。传入则脱离表单上下文，
+   *  用于复用到「新建研究模板」等场景，把建议字段写进调用方的表单。 */
+  onApprove?: (payload: SetupWritingApprovalPayload) => void;
+  title?: string;
+  hint?: string;
+} = {}) {
   const threadId = useMemo(() => uuid(), []);
   const [threadPrepared, setThreadPrepared] = useState(false);
   const [chatModel, setChatModel] = useState('');
   const { models, isLoading: modelsLoading } = useModels();
   const { articleTypes } = useArticleTypes();
-  const { applyProposal } = useWritingForm();
+  // 缺省回填写作表单；onApprove 传入时脱离 WritingFormProvider（如模板创建页）。
+  const formCtx = useWritingFormOptional();
+  const applyProposal = onApproveProp ?? formCtx?.applyProposal;
 
   // 首个模型可用时默认选中
   useEffect(() => {
@@ -98,8 +110,8 @@ export function WritingSetupChat() {
 
   const handleApprove = useCallback(
     async (payload: SetupWritingApprovalPayload) => {
-      applyProposal(payload);
-      toast.success('已应用到左侧表单，可继续编辑或开始写作');
+      applyProposal?.(payload);
+      toast.success('已应用到表单，可继续编辑');
     },
     [applyProposal],
   );
@@ -119,8 +131,8 @@ export function WritingSetupChat() {
           {!showWelcome && (
             <header className="flex shrink-0 items-center gap-2 border-b px-4 py-3">
               <Sparkles className="size-4 text-primary" />
-              <h2 className="text-sm font-semibold">写作助手</h2>
-              <span className="text-muted-foreground text-xs">描述你想写什么，我来帮你填好右侧配置</span>
+              <h2 className="text-sm font-semibold">{title}</h2>
+              <span className="text-muted-foreground text-xs">{hint}</span>
             </header>
           )}
 
diff --git a/frontend-web/src/open-canvas/components/ai-writing/WritingWelcome.tsx b/frontend-web/src/open-canvas/components/ai-writing/WritingWelcome.tsx
index 30110ee4..6c0cd0e5 100644
--- a/frontend-web/src/open-canvas/components/ai-writing/WritingWelcome.tsx
+++ b/frontend-web/src/open-canvas/components/ai-writing/WritingWelcome.tsx
@@ -39,7 +39,7 @@ export function WritingWelcome() {
         <div className="flex flex-1 flex-col justify-center gap-6">
           <div>
             <h2 className="text-2xl font-bold leading-snug tracking-tight text-slate-800 dark:text-white">
-              欢迎使用「妙笔 AI」智力打字协作工坊
+              欢迎使用「ai写作」智力打字协作工坊
             </h2>
             <p className="mt-2 max-w-xl text-sm leading-relaxed text-slate-500 dark:text-white/60">
               这是一个专业的自动化写作系统。你只需在下方对话框中
diff --git a/frontend-web/src/open-canvas/contexts/WritingFormContext.tsx b/frontend-web/src/open-canvas/contexts/WritingFormContext.tsx
index a5bfbe8c..a05dc175 100644
--- a/frontend-web/src/open-canvas/contexts/WritingFormContext.tsx
+++ b/frontend-web/src/open-canvas/contexts/WritingFormContext.tsx
@@ -36,6 +36,12 @@ export interface WritingFormValue {
   setStrictMode: (v: boolean) => void;
   materialSource: 'general' | 'notebook';
   setMaterialSource: (v: 'general' | 'notebook') => void;
+  /** 我的空间（笔记本）来源：选中的 search_space_id；上提到 Context 以便左栏「笔记本」点击直接设为素材来源。 */
+  selectedNotebookId: number | null;
+  setSelectedNotebookId: (v: number | null) => void;
+  /** 选中笔记本下的文档 id 列表（缩小检索范围）。 */
+  selectedDocumentIds: number[];
+  setSelectedDocumentIds: (v: number[]) => void;
   modelName: string;
   setModelName: (v: string) => void;
   /** 把模型建议（setup_writing 审批 payload）回填到表单各字段。 */
@@ -56,7 +62,17 @@ export function useWritingFormOptional(): WritingFormValue | null {
   return useContext(WritingFormContext);
 }
 
-export function WritingFormProvider({ children, initialMaterialSource }: { children: ReactNode; initialMaterialSource?: 'general' | 'notebook' }) {
+export function WritingFormProvider({
+  children,
+  initialMaterialSource,
+  initialNotebookId,
+  initialDocumentIds,
+}: {
+  children: ReactNode;
+  initialMaterialSource?: 'general' | 'notebook';
+  initialNotebookId?: number | null;
+  initialDocumentIds?: number[];
+}) {
   const [intent, setIntent] = useState('');
   const [formExpanded, setFormExpanded] = useState(false);
   const [showOutline, setShowOutline] = useState(true);
@@ -66,6 +82,8 @@ export function WritingFormProvider({ children, initialMaterialSource }: { child
   const [audience, setAudience] = useState('普通读者');
   const [strictMode, setStrictMode] = useState(false);
   const [materialSource, setMaterialSource] = useState<'general' | 'notebook'>(initialMaterialSource ?? 'general');
+  const [selectedNotebookId, setSelectedNotebookId] = useState<number | null>(initialNotebookId ?? null);
+  const [selectedDocumentIds, setSelectedDocumentIds] = useState<number[]>(initialDocumentIds ?? []);
   const [modelName, setModelName] = useState('');
 
   const applyProposal = useCallback((payload: SetupWritingApprovalPayload) => {
@@ -100,10 +118,12 @@ export function WritingFormProvider({ children, initialMaterialSource }: { child
       audience, setAudience,
       strictMode, setStrictMode,
       materialSource, setMaterialSource,
+      selectedNotebookId, setSelectedNotebookId,
+      selectedDocumentIds, setSelectedDocumentIds,
       modelName, setModelName,
       applyProposal,
     }),
-    [intent, formExpanded, showOutline, outlineText, articleType, wordCountInput, audience, strictMode, materialSource, modelName, applyProposal],
+    [intent, formExpanded, showOutline, outlineText, articleType, wordCountInput, audience, strictMode, materialSource, selectedNotebookId, selectedDocumentIds, modelName, applyProposal],
   );
 
   return <WritingFormContext.Provider value={value}>{children}</WritingFormContext.Provider>;
diff --git a/frontend-web/src/open-canvas/hooks/useResearchTemplates.ts b/frontend-web/src/open-canvas/hooks/useResearchTemplates.ts
new file mode 100644
index 00000000..6ae76d28
--- /dev/null
+++ b/frontend-web/src/open-canvas/hooks/useResearchTemplates.ts
@@ -0,0 +1,60 @@
+import { useCallback, useState } from 'react';
+
+import {
+  createResearchTemplate,
+  deleteResearchTemplate,
+  listResearchTemplates,
+  updateResearchTemplate,
+} from '../api/ai-writing-research-templates';
+import type {
+  CreateResearchTemplateRequest,
+  ResearchTemplate,
+  UpdateResearchTemplateRequest,
+} from '../api/ai-writing-research-templates';
+
+export function useResearchTemplates() {
+  const [templates, setTemplates] = useState<ResearchTemplate[]>([]);
+  const [loading, setLoading] = useState(false);
+  const [error, setError] = useState<string | null>(null);
+
+  const refresh = useCallback(async () => {
+    setLoading(true);
+    setError(null);
+    try {
+      setTemplates(await listResearchTemplates());
+    } catch (e) {
+      setError(e instanceof Error ? e.message : '加载失败');
+    } finally {
+      setLoading(false);
+    }
+  }, []);
+
+  const create = useCallback(
+    async (data: CreateResearchTemplateRequest): Promise<ResearchTemplate> => {
+      const created = await createResearchTemplate(data);
+      setTemplates((prev) =>
+        [...prev, created].sort((a, b) => a.sort_order - b.sort_order),
+      );
+      return created;
+    },
+    [],
+  );
+
+  const update = useCallback(
+    async (id: string, data: UpdateResearchTemplateRequest): Promise<ResearchTemplate> => {
+      const updated = await updateResearchTemplate(id, data);
+      setTemplates((prev) =>
+        prev.map((t) => (t.id === id ? updated : t)).sort((a, b) => a.sort_order - b.sort_order),
+      );
+      return updated;
+    },
+    [],
+  );
+
+  const remove = useCallback(async (id: string): Promise<void> => {
+    await deleteResearchTemplate(id);
+    setTemplates((prev) => prev.filter((t) => t.id !== id));
+  }, []);
+
+  return { templates, loading, error, refresh, create, update, remove };
+}
diff --git a/frontend-web/src/open-canvas/pages/AIWritingPage.tsx b/frontend-web/src/open-canvas/pages/AIWritingPage.tsx
index 2e86de1c..c403bb45 100644
--- a/frontend-web/src/open-canvas/pages/AIWritingPage.tsx
+++ b/frontend-web/src/open-canvas/pages/AIWritingPage.tsx
@@ -81,8 +81,8 @@
  *         -d '{"retention_days": 0}'
  * ──────────────────────────────────────────────────────────────────────────
  */
-import { useRef } from 'react';
-import { PanelRightOpen } from 'lucide-react';
+import { useRef, useState } from 'react';
+import { ArrowLeft, PanelRightOpen } from 'lucide-react';
 import { TooltipProvider } from '@/components/ui/tooltip';
 import { cn } from '@/lib/utils';
 import type { GroupImperativeHandle } from 'react-resizable-panels';
@@ -93,6 +93,13 @@ import { CanvasUIProvider } from '../contexts/CanvasUIContext';
 import { AIWritingPanel } from '../components/ai-writing/AIWritingPanel';
 import { AIWritingDraftPanel } from '../components/ai-writing/AIWritingDraftPanel';
 import { WritingSetupChat } from '../components/ai-writing/WritingSetupChat';
+import { ResearchSidebar } from '../components/ai-writing/ResearchSidebar';
+import StudioNotebookDetailPage from '@/pages/StudioNotebookDetailPage';
+import StudioNotebooksPage from '@/pages/StudioNotebooksPage';
+import { ChatRuntime } from '@/pages/WorkspaceRoutes';
+
+/** 右侧内嵌视图：笔记本详情 / 所有空间列表 / null（写作） */
+type RightView = { kind: 'notebook'; id: number } | { kind: 'allSpaces' } | null;
 import { AIWritingProvider, useAIWriting } from '../contexts/AIWritingContext';
 import { AIWritingHistoryProvider, useAIWritingHistory } from '../contexts/AIWritingHistoryContext';
 import { WritingFormProvider, useWritingForm } from '../contexts/WritingFormContext';
@@ -210,6 +217,81 @@ function AIWritingLayout({ hideAdvancedFields, initialNotebookId, initialDocumen
   );
 }
 
+/**
+ * 写作页主体（侧栏 + 内容区）。放在 Provider 内部，以便读 useAIWriting 状态：
+ * 进入写作（status !== idle 或查看历史）后自动收起左侧研究模板栏，让写作区更宽。
+ */
+function AIWritingBody({ hideAdvancedFields, initialNotebookId, initialDocumentIds, studioToken }: { hideAdvancedFields: boolean } & NotebookProps) {
+  const { state } = useAIWriting();
+  const { isHistoryMode } = useAIWritingHistory();
+
+  // 需求 0616 #5：左栏点「笔记本 / 所有空间」→ 在右侧内嵌打开对应页，替换右侧内容
+  // （写作布局不卸载，仅 display:none 隐藏，避免 WritingSetupChat 重挂丢线程）。
+  const [rightView, setRightView] = useState<RightView>(null);
+  const overlayActive = rightView != null;
+  const viewingNotebookId = rightView?.kind === 'notebook' ? rightView.id : null;
+
+  // 进入写作（开始生成 / 查看历史）后收起研究模板侧栏；嵌入态(studioToken)始终不展示。
+  const showSidebar = !studioToken && state.status === 'idle' && !isHistoryMode;
+
+  return (
+    <div className="flex-1 min-h-0 overflow-hidden flex">
+      {/* 需求 0616 #5：课题研究左栏（研究模板 + 笔记本）。进入写作后收起。 */}
+      {showSidebar && (
+        <ResearchSidebar
+          viewingNotebookId={viewingNotebookId}
+          onOpenNotebook={(id) => setRightView({ kind: 'notebook', id })}
+          onOpenAllSpaces={() => setRightView({ kind: 'allSpaces' })}
+        />
+      )}
+      {/* 右侧：写作布局常驻（仅 display:none 隐藏，避免 WritingSetupChat 重挂丢线程），
+          查看笔记本 / 所有空间时在其上叠加内嵌页。 */}
+      <div className="relative flex-1 min-h-0 overflow-hidden">
+        <div className={cn('absolute inset-0', overlayActive && 'hidden')}>
+          <AIWritingLayout hideAdvancedFields={hideAdvancedFields} initialNotebookId={initialNotebookId} initialDocumentIds={initialDocumentIds} studioToken={studioToken} />
+        </div>
+        {overlayActive && (
+          <div className="absolute inset-0 flex flex-col bg-background">
+            <div className="flex shrink-0 items-center gap-1 border-b border-border px-4 py-2">
+              <button
+                type="button"
+                onClick={() => setRightView(null)}
+                className="flex items-center gap-1.5 rounded-md px-2 py-1 text-sm text-muted-foreground transition-colors hover:bg-muted hover:text-foreground"
+              >
+                <ArrowLeft className="size-4" />
+                返回写作
+              </button>
+              {rightView.kind === 'notebook' && (
+                <button
+                  type="button"
+                  onClick={() => setRightView({ kind: 'allSpaces' })}
+                  className="ml-1 rounded-md px-2 py-1 text-sm text-muted-foreground transition-colors hover:bg-muted hover:text-foreground"
+                >
+                  所有空间
+                </button>
+              )}
+            </div>
+            <div className="min-h-0 flex-1 overflow-hidden">
+              {rightView.kind === 'notebook' ? (
+                // 笔记本详情页内部用 ChatBox(useArtifacts) → 需 ChatRuntime 提供 Provider。
+                <ChatRuntime>
+                  <StudioNotebookDetailPage key={rightView.id} notebookId={rightView.id} />
+                </ChatRuntime>
+              ) : (
+                <div className="h-full overflow-y-auto">
+                  <StudioNotebooksPage
+                    onOpenNotebook={(id) => setRightView({ kind: 'notebook', id })}
+                  />
+                </div>
+              )}
+            </div>
+          </div>
+        )}
+      </div>
+    </div>
+  );
+}
+
 export default function AIWritingPage({
   hideAdvancedFields = false,
   initialNotebookId,
@@ -227,12 +309,19 @@ export default function AIWritingPage({
       <CanvasUIProvider>
       <AIWritingHistoryProvider>
       <AIWritingProvider>
-      <WritingFormProvider initialMaterialSource={studioToken ?? initialNotebookId != null ? 'notebook' : undefined}>
+      <WritingFormProvider
+        initialMaterialSource={studioToken ?? initialNotebookId != null ? 'notebook' : undefined}
+        initialNotebookId={initialNotebookId}
+        initialDocumentIds={initialDocumentIds}
+      >
       <TooltipProvider delayDuration={300}>
         <div className="h-full flex flex-col overflow-hidden bg-background canvas-writing-bg">
-          <div className="flex-1 min-h-0 overflow-hidden">
-            <AIWritingLayout hideAdvancedFields={hideAdvancedFields} initialNotebookId={initialNotebookId} initialDocumentIds={initialDocumentIds} studioToken={studioToken} />
-          </div>
+          <AIWritingBody
+            hideAdvancedFields={hideAdvancedFields}
+            initialNotebookId={initialNotebookId}
+            initialDocumentIds={initialDocumentIds}
+            studioToken={studioToken}
+          />
         </div>
       </TooltipProvider>
       </WritingFormProvider>
diff --git a/frontend-web/src/open-canvas/pages/ResearchTemplateEditorPage.tsx b/frontend-web/src/open-canvas/pages/ResearchTemplateEditorPage.tsx
new file mode 100644
index 00000000..adcdf976
--- /dev/null
+++ b/frontend-web/src/open-canvas/pages/ResearchTemplateEditorPage.tsx
@@ -0,0 +1,300 @@
+import { useEffect, useState } from 'react';
+import { useNavigate, useParams } from 'react-router-dom';
+import { ArrowLeft, Loader2 } from 'lucide-react';
+
+import { Button } from '@/components/ui/button';
+import type { SetupWritingApprovalPayload } from '@/components/workspace/messages';
+
+import { useResearchTemplates } from '../hooks/useResearchTemplates';
+import { useArticleTypes } from '../hooks/useArticleTypes';
+import { WritingSetupChat } from '../components/ai-writing/WritingSetupChat';
+import type {
+  CreateResearchTemplateRequest,
+  ResearchTemplate,
+} from '../api/ai-writing-research-templates';
+
+const LIST_PATH = '/page/canvas/research-templates';
+
+const INPUT_CLASS =
+  'flex h-9 w-full rounded-md border border-input bg-background px-3 py-1 text-sm shadow-sm ' +
+  'placeholder:text-muted-foreground focus-visible:outline-none focus-visible:ring-1 focus-visible:ring-ring disabled:opacity-50';
+const TEXTAREA_CLASS =
+  'flex w-full rounded-md border border-input bg-background px-3 py-2 text-sm shadow-sm ' +
+  'placeholder:text-muted-foreground focus-visible:outline-none focus-visible:ring-1 focus-visible:ring-ring resize-none';
+const LABEL_CLASS = 'mb-1 block text-xs font-medium text-muted-foreground';
+
+interface FormState {
+  name: string;
+  description: string;
+  article_type_key: string;
+  outline: string;
+  material_source: 'general' | 'notebook';
+  word_count: string;
+  audience: string;
+  strict_mode: boolean;
+  topic: string;
+  sort_order: string;
+}
+
+const EMPTY_FORM: FormState = {
+  name: '',
+  description: '',
+  article_type_key: '',
+  outline: '',
+  material_source: 'general',
+  word_count: '800',
+  audience: '',
+  strict_mode: false,
+  topic: '',
+  sort_order: '0',
+};
+
+function toForm(t: ResearchTemplate): FormState {
+  return {
+    name: t.name,
+    description: t.description ?? '',
+    article_type_key: t.article_type_key ?? '',
+    outline: t.outline ?? '',
+    material_source: t.material_source,
+    word_count: String(t.word_count),
+    audience: t.audience ?? '',
+    strict_mode: t.strict_mode,
+    topic: t.topic ?? '',
+    sort_order: String(t.sort_order),
+  };
+}
+
+function toPayload(form: FormState): CreateResearchTemplateRequest {
+  return {
+    name: form.name.trim(),
+    description: form.description.trim() || null,
+    article_type_key: form.article_type_key || null,
+    outline: form.outline.trim() || null,
+    material_source: form.material_source,
+    word_count: Math.max(100, parseInt(form.word_count, 10) || 800),
+    audience: form.audience.trim() || null,
+    strict_mode: form.strict_mode,
+    topic: form.topic.trim() || null,
+    sort_order: parseInt(form.sort_order, 10) || 0,
+  };
+}
+
+export default function ResearchTemplateEditorPage() {
+  const navigate = useNavigate();
+  const { id } = useParams<{ id: string }>();
+  const editing = !!id;
+
+  const { templates, refresh, create, update } = useResearchTemplates();
+  const { articleTypes, refresh: refreshTypes } = useArticleTypes();
+
+  const [form, setForm] = useState<FormState>(EMPTY_FORM);
+  // 新建即就绪；编辑需等列表加载后回填一次。
+  const [initialized, setInitialized] = useState(!editing);
+  const [formError, setFormError] = useState<string | null>(null);
+  const [saving, setSaving] = useState(false);
+
+  useEffect(() => {
+    void refresh();
+    void refreshTypes();
+  }, [refresh, refreshTypes]);
+
+  useEffect(() => {
+    if (!editing || initialized) return;
+    const t = templates.find((x) => x.id === id);
+    if (t) {
+      setForm(toForm(t));
+      setInitialized(true);
+    }
+  }, [editing, initialized, templates, id]);
+
+  const set = <K extends keyof FormState>(key: K, value: FormState[K]) =>
+    setForm((f) => ({ ...f, [key]: value }));
+
+  // 问答助手「应用」→ 把 setup_writing 建议映射进模板表单字段（与 AI 写作同款机制）。
+  const applyProposalToForm = (p: SetupWritingApprovalPayload) =>
+    setForm((f) => ({
+      ...f,
+      name: f.name || (p.user_intent?.trim() ? p.user_intent.trim().slice(0, 20) : f.name),
+      topic: p.user_intent?.trim() || f.topic,
+      article_type_key: p.article_type?.trim() || f.article_type_key,
+      outline: p.outline?.trim() || f.outline,
+      word_count:
+        typeof p.word_count_target === 'number' && p.word_count_target > 0
+          ? String(p.word_count_target)
+          : f.word_count,
+      audience: p.target_audience?.trim() || f.audience,
+      strict_mode:
+        p.writing_mode === 'strict' ? true : p.writing_mode === 'loose' ? false : f.strict_mode,
+      material_source: p.material_source
+        ? p.material_source === 'notebook'
+          ? 'notebook'
+          : 'general'
+        : f.material_source,
+    }));
+
+  const handleSave = async () => {
+    if (!form.name.trim()) {
+      setFormError('请填写模板名称');
+      return;
+    }
+    setSaving(true);
+    setFormError(null);
+    try {
+      const payload = toPayload(form);
+      if (editing && id) await update(id, payload);
+      else await create(payload);
+      navigate(LIST_PATH);
+    } catch (e) {
+      setFormError(e instanceof Error ? e.message : '保存失败');
+      setSaving(false);
+    }
+  };
+
+  return (
+    <div className="flex h-full flex-col bg-background">
+      <header className="flex shrink-0 items-center justify-between border-b border-border px-5 py-3">
+        <div className="flex items-center gap-2">
+          <Button variant="ghost" size="icon-sm" onClick={() => navigate(LIST_PATH)}>
+            <ArrowLeft className="size-4" />
+          </Button>
+          <div>
+            <h1 className="text-base font-semibold">
+              {editing ? '编辑研究模板' : '新建研究模板'}
+            </h1>
+            <p className="text-xs text-muted-foreground">
+              左侧用问答描述需求，自动填好右侧字段；也可手动编辑。
+            </p>
+          </div>
+        </div>
+        <div className="flex items-center gap-2">
+          {formError && <span className="text-sm text-destructive">{formError}</span>}
+          <Button variant="outline" size="sm" onClick={() => navigate(LIST_PATH)} disabled={saving}>
+            取消
+          </Button>
+          <Button size="sm" onClick={() => void handleSave()} disabled={saving}>
+            {saving && <Loader2 className="size-4 animate-spin" />}
+            保存
+          </Button>
+        </div>
+      </header>
+
+      <div className="flex min-h-0 flex-1">
+        {/* 左：模板助手问答（同 AI 写作 setup 机制）→ 应用到右侧表单。问答区更宽（约 7:3）。 */}
+        <div className="min-w-0 flex-[7] border-r border-border">
+          <WritingSetupChat
+            onApprove={applyProposalToForm}
+            title="模板助手"
+            hint="描述你想要的研究模板，我来帮你填好右侧字段"
+          />
+        </div>
+
+        {/* 右：模板表单 */}
+        <div className="grid min-w-0 flex-[3] grid-cols-2 content-start gap-4 overflow-y-auto p-6">
+          <div className="col-span-2">
+            <label className={LABEL_CLASS}>模板名称 *</label>
+            <input
+              className={INPUT_CLASS}
+              value={form.name}
+              onChange={(e) => set('name', e.target.value)}
+              placeholder="如：深度舆情分析"
+            />
+          </div>
+          <div className="col-span-2">
+            <label className={LABEL_CLASS}>描述</label>
+            <input
+              className={INPUT_CLASS}
+              value={form.description}
+              onChange={(e) => set('description', e.target.value)}
+              placeholder="模板用途简述"
+            />
+          </div>
+          <div>
+            <label className={LABEL_CLASS}>文章类型</label>
+            <select
+              className={INPUT_CLASS}
+              value={form.article_type_key}
+              onChange={(e) => set('article_type_key', e.target.value)}
+            >
+              <option value="">不设</option>
+              {articleTypes.map((t) => (
+                <option key={t.key} value={t.key}>
+                  {t.label}
+                </option>
+              ))}
+            </select>
+          </div>
+          <div>
+            <label className={LABEL_CLASS}>搜索来源</label>
+            <select
+              className={INPUT_CLASS}
+              value={form.material_source}
+              onChange={(e) => set('material_source', e.target.value as 'general' | 'notebook')}
+            >
+              <option value="general">通用检索</option>
+              <option value="notebook">我的空间（笔记本）</option>
+            </select>
+          </div>
+          <div>
+            <label className={LABEL_CLASS}>目标字数</label>
+            <input
+              type="number"
+              className={INPUT_CLASS}
+              value={form.word_count}
+              onChange={(e) => set('word_count', e.target.value)}
+            />
+          </div>
+          <div>
+            <label className={LABEL_CLASS}>目标受众</label>
+            <input
+              className={INPUT_CLASS}
+              value={form.audience}
+              onChange={(e) => set('audience', e.target.value)}
+              placeholder="如：决策者"
+            />
+          </div>
+          <div className="col-span-2">
+            <label className={LABEL_CLASS}>主题 / 意图</label>
+            <textarea
+              className={TEXTAREA_CLASS}
+              rows={2}
+              value={form.topic}
+              onChange={(e) => set('topic', e.target.value)}
+              placeholder="点击模板时预填到写作意图输入框"
+            />
+          </div>
+          <div className="col-span-2">
+            <label className={LABEL_CLASS}>默认大纲</label>
+            <textarea
+              className={TEXTAREA_CLASS}
+              rows={6}
+              value={form.outline}
+              onChange={(e) => set('outline', e.target.value)}
+              placeholder={'一、…\n二、…\n三、…'}
+            />
+          </div>
+          <div className="flex items-center gap-2">
+            <input
+              id="rt-strict"
+              type="checkbox"
+              checked={form.strict_mode}
+              onChange={(e) => set('strict_mode', e.target.checked)}
+              className="size-4"
+            />
+            <label htmlFor="rt-strict" className="text-sm">
+              严格写作模式
+            </label>
+          </div>
+          <div>
+            <label className={LABEL_CLASS}>排序（小在前）</label>
+            <input
+              type="number"
+              className={INPUT_CLASS}
+              value={form.sort_order}
+              onChange={(e) => set('sort_order', e.target.value)}
+            />
+          </div>
+        </div>
+      </div>
+    </div>
+  );
+}
diff --git a/frontend-web/src/open-canvas/pages/ResearchTemplatesConfigPage.tsx b/frontend-web/src/open-canvas/pages/ResearchTemplatesConfigPage.tsx
new file mode 100644
index 00000000..ad9ebfa1
--- /dev/null
+++ b/frontend-web/src/open-canvas/pages/ResearchTemplatesConfigPage.tsx
@@ -0,0 +1,148 @@
+import { useEffect, useState } from 'react';
+import { useNavigate } from 'react-router-dom';
+import { ArrowLeft, FileText, Loader2, Pencil, Plus, Trash2 } from 'lucide-react';
+
+import { Button } from '@/components/ui/button';
+import {
+  Dialog,
+  DialogContent,
+  DialogFooter,
+  DialogHeader,
+  DialogTitle,
+} from '@/components/ui/dialog';
+
+import { useResearchTemplates } from '../hooks/useResearchTemplates';
+import { useArticleTypes } from '../hooks/useArticleTypes';
+
+const NEW_PATH = '/page/canvas/research-templates/new';
+const editPath = (id: string) => `/page/canvas/research-templates/edit/${id}`;
+
+export default function ResearchTemplatesConfigPage() {
+  const navigate = useNavigate();
+  const { templates, loading, error, refresh, remove } = useResearchTemplates();
+  const { articleTypes, refresh: refreshTypes } = useArticleTypes();
+
+  const [deleteTargetId, setDeleteTargetId] = useState<string | null>(null);
+  const [deleting, setDeleting] = useState(false);
+
+  useEffect(() => {
+    void refresh();
+    void refreshTypes();
+  }, [refresh, refreshTypes]);
+
+  const handleDelete = async () => {
+    if (!deleteTargetId) return;
+    setDeleting(true);
+    try {
+      await remove(deleteTargetId);
+      setDeleteTargetId(null);
+    } finally {
+      setDeleting(false);
+    }
+  };
+
+  const articleTypeLabel = (key: string | null) =>
+    key ? articleTypes.find((t) => t.key === key)?.label ?? key : '—';
+
+  return (
+    <div className="mx-auto h-full w-full max-w-4xl overflow-y-auto p-6">
+      <div className="mb-5 flex items-center justify-between">
+        <div className="flex items-center gap-2">
+          <Button variant="ghost" size="icon-sm" onClick={() => navigate(-1)}>
+            <ArrowLeft className="size-4" />
+          </Button>
+          <div>
+            <h1 className="text-lg font-semibold">研究模板配置</h1>
+            <p className="text-xs text-muted-foreground">
+              配置课题研究左栏的研究模板；点击模板时一键预填写作表单。
+            </p>
+          </div>
+        </div>
+        <Button size="sm" onClick={() => navigate(NEW_PATH)}>
+          <Plus className="size-4" />
+          新建模板
+        </Button>
+      </div>
+
+      {error && <p className="mb-3 text-sm text-destructive">{error}</p>}
+
+      {loading ? (
+        <div className="flex justify-center py-16">
+          <Loader2 className="size-6 animate-spin text-muted-foreground" />
+        </div>
+      ) : templates.length === 0 ? (
+        <p className="py-16 text-center text-sm text-muted-foreground">
+          暂无研究模板，点击右上角「新建模板」创建。
+        </p>
+      ) : (
+        <div className="flex flex-col gap-2">
+          {templates.map((t) => (
+            <div
+              key={t.id}
+              className="flex items-start gap-3 rounded-xl border border-border bg-card p-4"
+            >
+              <div className="bg-primary/10 text-primary mt-0.5 flex size-9 shrink-0 items-center justify-center rounded-lg">
+                <FileText className="size-4" />
+              </div>
+              <div className="min-w-0 flex-1">
+                <div className="flex items-center gap-2">
+                  <span className="truncate text-sm font-semibold">{t.name}</span>
+                  <span className="shrink-0 rounded bg-muted px-1.5 py-0.5 text-[11px] text-muted-foreground">
+                    {articleTypeLabel(t.article_type_key)}
+                  </span>
+                </div>
+                {t.description && (
+                  <p className="mt-0.5 line-clamp-1 text-xs text-muted-foreground">{t.description}</p>
+                )}
+                <div className="mt-1 flex flex-wrap gap-x-3 gap-y-0.5 text-[11px] text-muted-foreground/80">
+                  <span>来源：{t.material_source === 'notebook' ? '我的空间' : '通用检索'}</span>
+                  <span>字数：{t.word_count}</span>
+                  {t.audience && <span>受众：{t.audience}</span>}
+                  <span>{t.strict_mode ? '严格模式' : '宽松模式'}</span>
+                </div>
+              </div>
+              <div className="flex shrink-0 items-center gap-1">
+                <Button
+                  variant="ghost"
+                  size="icon-sm"
+                  onClick={() => navigate(editPath(t.id))}
+                  title="编辑"
+                >
+                  <Pencil className="size-4" />
+                </Button>
+                <Button
+                  variant="ghost"
+                  size="icon-sm"
+                  onClick={() => setDeleteTargetId(t.id)}
+                  title="删除"
+                  className="text-muted-foreground hover:text-destructive"
+                >
+                  <Trash2 className="size-4" />
+                </Button>
+              </div>
+            </div>
+          ))}
+        </div>
+      )}
+
+      {/* 删除确认 */}
+      <Dialog open={!!deleteTargetId} onOpenChange={(v) => !deleting && !v && setDeleteTargetId(null)}>
+        <DialogContent className="sm:max-w-sm">
+          <DialogHeader>
+            <DialogTitle>删除研究模板</DialogTitle>
+          </DialogHeader>
+          <p className="py-2 text-sm text-muted-foreground">确定删除该研究模板吗？此操作不可恢复。</p>
+          <DialogFooter>
+            <Button variant="outline" onClick={() => setDeleteTargetId(null)} disabled={deleting}>
+              取消
+            </Button>
+            <Button variant="destructive" onClick={() => void handleDelete()} disabled={deleting}>
+              {deleting && <Loader2 className="size-4 animate-spin" />}
+              删除
+            </Button>
+          </DialogFooter>
+        </DialogContent>
+      </Dialog>
+    </div>
+  );
+}
diff --git a/frontend-web/src/open-canvas/routes/OpenCanvasRoutes.tsx b/frontend-web/src/open-canvas/routes/OpenCanvasRoutes.tsx
index d1685e6e..3de029ba 100644
--- a/frontend-web/src/open-canvas/routes/OpenCanvasRoutes.tsx
+++ b/frontend-web/src/open-canvas/routes/OpenCanvasRoutes.tsx
@@ -4,6 +4,8 @@ import CanvasHomePage from "../pages/CanvasHomePage";
 import CanvasThreadPage from "../pages/CanvasThreadPage";
 import AIWritingPage from "../pages/AIWritingPage";
 import ArticleTypesConfigPage from "../pages/ArticleTypesConfigPage";
+import ResearchTemplatesConfigPage from "../pages/ResearchTemplatesConfigPage";
+import ResearchTemplateEditorPage from "../pages/ResearchTemplateEditorPage";
 
 export default function OpenCanvasRoutes() {
   return (
@@ -11,6 +13,9 @@ export default function OpenCanvasRoutes() {
       <Route index element={<CanvasHomePage />} />
       <Route path="ai-writing" element={<AIWritingPage hideAdvancedFields />} />
       <Route path="article-types" element={<ArticleTypesConfigPage />} />
+      <Route path="research-templates" element={<ResearchTemplatesConfigPage />} />
+      <Route path="research-templates/new" element={<ResearchTemplateEditorPage />} />
+      <Route path="research-templates/edit/:id" element={<ResearchTemplateEditorPage />} />
       <Route path=":thread_id" element={<CanvasThreadPage />} />
     </Routes>
   );
diff --git a/frontend-web/src/pages/AgentChatPage.tsx b/frontend-web/src/pages/AgentChatPage.tsx
index 33626595..fc485f41 100644
--- a/frontend-web/src/pages/AgentChatPage.tsx
+++ b/frontend-web/src/pages/AgentChatPage.tsx
@@ -31,15 +31,31 @@ import { textOfMessage } from "@/core/threads/utils";
 import { env } from "@/env";
 import { desensitize } from "@/lib/desensitize";
 
-export default function AgentChatPage() {
+export default function AgentChatPage({
+  embeddedAgentId,
+  chatBasePath,
+}: {
+  /** 当作为「通用问答右侧嵌入」渲染时由父组件传入的 agent id（替代路由参数）。
+   *  传入即视为嵌入态：路由仍停留在 /chats/new，仅右侧问答区换成该智能体的对话。 */
+  embeddedAgentId?: string;
+  /** 智能体对话所在的路由前缀（到 `/chats` 为止，不含 thread_id）。
+   *  默认 `${routeBase}/agents/${agent_id}/chats`；任务深链把对话嵌进「任务管理」布局时，
+   *  传入如 `/page/strategy/task-management/plan-task-sentiment/agent/${id}/chats`，
+   *  使开新会话 / 首条消息后的 URL 改写都停留在任务管理上下文（侧边栏保持对应业务高亮）。 */
+  chatBasePath?: string;
+} = {}) {
   const routeBase = useRouteBase();
   const { t } = useI18n();
   const [showFollowups, setShowFollowups] = useState(false);
   const router = useRouter();
 
-  const { agent_id } = useParams<{
+  const { agent_id: agentIdFromPath } = useParams<{
     agent_id: string;
   }>();
+  const agent_id = embeddedAgentId ?? agentIdFromPath;
+  const isEmbedded = !!embeddedAgentId;
+  // 对话路由前缀：调用方可覆盖（任务深链嵌入态），否则用当前 routeBase 下的独立智能体对话路径。
+  const agentChatsBase = chatBasePath ?? `${routeBase}/agents/${agent_id}/chats`;
 
   // If ChatPage stashed an opening message for this agent, skip the landing
   // screen on first render so the user doesn't see a flash of the welcome UI
@@ -92,7 +108,7 @@ export default function AgentChatPage() {
       history.replaceState(
         null,
         "",
-        `${window.location.pathname}${window.location.search}#${routeBase}/agents/${agent_id}/chats/${createdThreadId}`,
+        `${window.location.pathname}${window.location.search}#${agentChatsBase}/${createdThreadId}`,
       );
     },
     onFinish: (state) => {
@@ -208,8 +224,12 @@ export default function AgentChatPage() {
                       size="sm"
                       variant="secondary"
                       onClick={() => {
+                        // 嵌入态：保持停留在通用问答右侧，开同一智能体的新对话；
+                        // 非嵌入态：走该智能体的独立新对话页（按当前路由 base）。
                         router.push(
-                          `/page/workspace/agents/${agent_id}/chats/new`,
+                          isEmbedded
+                            ? `/page/workspace/chats/new?agent=${agent_id}`
+                            : `${agentChatsBase}/new`,
                         );
                       }}
                     >
@@ -285,6 +305,8 @@ export default function AgentChatPage() {
                     thread={thread}
                     paddingBottom={showFollowups ? 96 : 24}
                     tokenUsageEnabled={tokenUsageEnabled}
+                    // 输入框在流内（MessageList 下方），按钮贴对话区底部即可。
+                    scrollButtonClassName="bottom-2"
                     onClarificationSubmit={handleClarificationSubmit}
                     onEditResubmit={(text) => handleSubmit({ text, files: [] })}
                   />
diff --git a/frontend-web/src/pages/AgentModulePage.tsx b/frontend-web/src/pages/AgentModulePage.tsx
new file mode 100644
index 00000000..11f371a5
--- /dev/null
+++ b/frontend-web/src/pages/AgentModulePage.tsx
@@ -0,0 +1,99 @@
+"use client";
+
+import { Loader2 } from "lucide-react";
+import { useState } from "react";
+import { Navigate, Route, Routes, useParams } from "react-router-dom";
+
+import {
+  AgentModuleSidebar,
+  type AgentModuleConfig,
+} from "@/components/workspace/agents/agent-module-sidebar";
+import { AgentModuleSettingsDialog } from "@/components/workspace/agents/agent-module-settings-dialog";
+import { useAgents } from "@/core/agents";
+import { getStoredAuth } from "@/core/auth";
+import { RouteBaseProvider } from "@/core/route-base";
+import { useAgentModuleConfig } from "@/core/system-settings/hooks";
+
+import AgentChatPage from "./AgentChatPage";
+import { ChatRuntime } from "./WorkspaceRoutes";
+
+/**
+ * 「按类型聚合」模块页（需求 0616 #3 zzJC / #4 zzXD，重做）。
+ *
+ * 一个**常驻外壳**（左栏：按类型分组智能体 + 聊天记录 + admin 设置）包裹右侧对话；
+ * 点击智能体只在模块内切换对话（左栏不卸载，无整页跳转）。模块自带独立路由
+ * ``/page/strategy/qa/{moduleKey}``，并以该路径作为 RouteBase，使内嵌的 AgentChatPage
+ * 所有跳转（新建对话 / onStart 改写 hash）都留在模块子树内。
+ */
+export function AgentModuleRoutes({ module }: { module: AgentModuleConfig }) {
+  const base = `/page/strategy/qa/${module.key}`;
+  return (
+    <RouteBaseProvider value={base}>
+      <Routes>
+        <Route index element={<ModuleDefaultRedirect module={module} />} />
+        <Route
+          path="agents/:agent_id/chats/:thread_id"
+          element={<ModuleChatShell module={module} />}
+        />
+      </Routes>
+    </RouteBaseProvider>
+  );
+}
+
+/** 进入模块默认落地：配置类型下的第一个智能体；无配置回退到模块默认体。 */
+function ModuleDefaultRedirect({ module }: { module: AgentModuleConfig }) {
+  const { agents, isLoading: agentsLoading } = useAgents();
+  const { data: config, isLoading: configLoading } = useAgentModuleConfig(
+    module.key,
+  );
+
+  if (agentsLoading || configLoading) {
+    return (
+      <div className="flex size-full items-center justify-center">
+        <Loader2 className="text-muted-foreground size-5 animate-spin" />
+      </div>
+    );
+  }
+
+  const tagIds = config?.tag_ids ?? [];
+  let targetId = module.fallbackAgentId;
+  if (tagIds.length > 0) {
+    const allowed = new Set(tagIds);
+    const first = agents.find((a) =>
+      (a.tags ?? []).some((t) => allowed.has(t.id)),
+    );
+    if (first) targetId = first.id;
+  }
+
+  return <Navigate to={`agents/${targetId}/chats/new`} replace />;
+}
+
+/** 模块外壳：常驻左栏 + 内嵌当前智能体对话（按 agent_id 重建对话区，左栏保持挂载）。 */
+function ModuleChatShell({ module }: { module: AgentModuleConfig }) {
+  const { agent_id } = useParams<{ agent_id: string }>();
+  const isAdmin = getStoredAuth()?.system_role === "admin";
+  const [settingsOpen, setSettingsOpen] = useState(false);
+
+  return (
+    <div className="flex size-full min-h-0 overflow-hidden">
+      <AgentModuleSidebar
+        moduleKey={module.key}
+        activeAgentId={agent_id}
+        onOpenSettings={isAdmin ? () => setSettingsOpen(true) : undefined}
+      />
+      <div className="min-h-0 min-w-0 flex-1">
+        <ChatRuntime>
+          <AgentChatPage key={agent_id} />
+        </ChatRuntime>
+      </div>
+      {isAdmin && (
+        <AgentModuleSettingsDialog
+          open={settingsOpen}
+          onOpenChange={setSettingsOpen}
+          moduleKey={module.key}
+          moduleTitle={module.title}
+        />
+      )}
+    </div>
+  );
+}
diff --git a/frontend-web/src/pages/BusinessChainMappingPage.tsx b/frontend-web/src/pages/BusinessChainMappingPage.tsx
index 291f3cdd..7efffe08 100644
--- a/frontend-web/src/pages/BusinessChainMappingPage.tsx
+++ b/frontend-web/src/pages/BusinessChainMappingPage.tsx
@@ -31,10 +31,13 @@ import {
   type BusinessMapping,
 } from "@/strategy-components/api/business-mapping";
 import { cn } from "@/lib/utils";
+import { useDesensitizeSubtree } from "@/roundtable-planning/lib/use-desensitize-subtree";
 
 export default function BusinessChainMappingPage() {
   const { theme } = useTheme();
   const isDark = theme === "dark";
+  // 整片脱敏：业务链条映射页展示文本（链条名/描述/席位名/标题）。输入框 Hook 自动跳过。
+  const dzRef = useDesensitizeSubtree<HTMLDivElement>();
 
   const [rows, setRows] = useState<BusinessMapping[]>([]);
   const [agents, setAgents] = useState<Agent[]>([]);
@@ -160,7 +163,7 @@ export default function BusinessChainMappingPage() {
   ];
 
   return (
-    <div className={cn(shellClass, "[&_.t-input]:rounded-[5px] [&_.t-input__wrap]:rounded-[5px]")}>
+    <div ref={dzRef} className={cn(shellClass, "[&_.t-input]:rounded-[5px] [&_.t-input__wrap]:rounded-[5px]")}>
       <div className="mx-auto flex min-h-full max-w-7xl flex-col gap-6 px-6 py-7 lg:px-8">
 
         {/* ── 页头 banner ── */}
diff --git a/frontend-web/src/pages/ChatPage.tsx b/frontend-web/src/pages/ChatPage.tsx
index 930dac06..81f21b66 100644
--- a/frontend-web/src/pages/ChatPage.tsx
+++ b/frontend-web/src/pages/ChatPage.tsx
@@ -34,7 +34,7 @@ import { isChatsIframeWorkspacePath } from "@/core/page-layout/chats-iframe-rout
 import { useSidebar } from "@/components/ui/sidebar";
 import { Button } from "@/components/ui/button";
 import { PanelLeftOpen } from "lucide-react";
-import { useAgentSelector } from "@/core/agents";
+import { useAgentSelector, useAgentsPaginated } from "@/core/agents";
 import { useI18n } from "@/core/i18n/hooks";
 import { useModels } from "@/core/models/hooks";
 import { useNotification } from "@/core/notification/hooks";
@@ -130,6 +130,10 @@ export default function ChatPage() {
   const browserContextAutoSubmitRef = useRef<string | null>(null);
   const { agents: selectorAgents, isLoading: selectorLoading } =
     useAgentSelector();
+  // 已选 agent 的存在性校验需要更宽的来源：侧边栏可选未配置进管理弹窗的 agent
+  // （用户自己的智能体），这些也算有效，故把「我的 / 广场」一并纳入校验集（缓存共享）。
+  const { agents: mineAgents } = useAgentsPaginated({ scope: "mine", page: 1, pageSize: 20 });
+  const { agents: squareAgents } = useAgentsPaginated({ scope: "square", page: 1, pageSize: 20 });
   const { selectedArtifact } = useArtifacts();
   useSpecificChatMode();
 
@@ -142,7 +146,7 @@ export default function ChatPage() {
       setHasStartedConversation(false);
       setPromptPrefixOn(false);
       // Pre-select the agent the user last started a conversation with so a
-      // fresh chat defaults to it. Validated against the selector list below.
+      // fresh chat defaults to it. Validated against the known agent list below.
       setSelectedAgentId(getLastAgentId());
     }
   }, [isNewThread, threadId]);
@@ -152,10 +156,14 @@ export default function ChatPage() {
   // dead agent. Wait for the list to load before deciding.
   useEffect(() => {
     if (selectorLoading || !selectedAgentId) return;
-    if (!selectorAgents.some((a) => a.id === selectedAgentId)) {
+    // 校验集 = 管理弹窗 ∪ 我的 ∪ 广场；任一存在即视为有效（含用户未配置进管理弹窗
+    // 的自有智能体）。三者都还没加载出来时不做清除，避免把恢复值误删。
+    const known = [...selectorAgents, ...mineAgents, ...squareAgents];
+    if (known.length === 0) return;
+    if (!known.some((a) => a.id === selectedAgentId)) {
       setSelectedAgentId(null);
     }
-  }, [selectedAgentId, selectorAgents, selectorLoading]);
+  }, [selectedAgentId, selectorAgents, mineAgents, squareAgents, selectorLoading]);
 
   const { showNotification } = useNotification();
 
@@ -493,6 +501,8 @@ export default function ChatPage() {
                   thread={thread}
                   paddingBottom={messageListPaddingBottom}
                   tokenUsageEnabled={tokenUsageEnabled}
+                  // 输入框为底部绝对浮层（约 160px 高），按钮上移到其上方居中。
+                  scrollButtonClassName="bottom-40"
                   onClarificationSubmit={handleClarificationSubmit}
                   onEditResubmit={(text) =>
                     handleSubmit({ text, files: [] })
diff --git a/frontend-web/src/pages/IframeChatPage.tsx b/frontend-web/src/pages/IframeChatPage.tsx
index f9983498..21422426 100644
--- a/frontend-web/src/pages/IframeChatPage.tsx
+++ b/frontend-web/src/pages/IframeChatPage.tsx
@@ -3,7 +3,6 @@
 import {
   BookMarkedIcon,
   BrainCircuitIcon,
-  CheckIcon,
   FileTextIcon,
   GraduationCapIcon,
 } from "lucide-react";
@@ -23,7 +22,9 @@ import {
 } from "@/components/workspace/messages";
 import { ThreadContext } from "@/components/workspace/messages/context";
 import { TodoList } from "@/components/workspace/todo-list";
+import { Tooltip } from "@/components/workspace/tooltip";
 import { Button } from "@/components/ui/button";
+import { TooltipProvider } from "@/components/ui/tooltip";
 import { useAgentSelector } from "@/core/agents";
 import { getAccessToken } from "@/core/auth";
 import { useI18n } from "@/core/i18n/hooks";
@@ -369,7 +370,7 @@ export default function IframeChatPage() {
   }, [handleInterrupt, handleSubmit]);
 
   const isLandingState = isNewThread && !hasStartedConversation;
-  const bottomBarHeightClass = "pb-28";
+  const bottomBarHeightClass = "pb-24";
   const modeButtons = [
     {
       key: "writing",
@@ -377,8 +378,7 @@ export default function IframeChatPage() {
       icon: FileTextIcon,
       active: !!settings.context.writing_mode,
       onClick: toggleWritingMode,
-      activeClassName:
-        "border-amber-300 bg-amber-50 text-amber-700 hover:bg-amber-100",
+      activeClassName: "text-amber-600 hover:text-amber-700",
     },
     {
       key: "memory",
@@ -386,8 +386,7 @@ export default function IframeChatPage() {
       icon: BrainCircuitIcon,
       active: settings.context.memory_injection_enabled ?? true,
       onClick: toggleMemoryInjection,
-      activeClassName:
-        "border-blue-300 bg-blue-50 text-blue-700 hover:bg-blue-100",
+      activeClassName: "text-blue-600 hover:text-blue-700",
     },
     {
       key: "references",
@@ -396,8 +395,7 @@ export default function IframeChatPage() {
       active: ragModeEnabled,
       onClick: toggleRagMode,
       disabled: ragModeLocked,
-      activeClassName:
-        "border-emerald-300 bg-emerald-50 text-emerald-700 hover:bg-emerald-100",
+      activeClassName: "text-emerald-600 hover:text-emerald-700",
     },
     {
       key: "pro",
@@ -405,8 +403,7 @@ export default function IframeChatPage() {
       icon: GraduationCapIcon,
       active: proModeEnabled,
       onClick: toggleProMode,
-      activeClassName:
-        "border-violet-300 bg-violet-50 text-violet-700 hover:bg-violet-100",
+      activeClassName: "text-violet-600 hover:text-violet-700",
     },
   ] as const;
 
@@ -496,29 +493,32 @@ export default function IframeChatPage() {
               </div>
             </div>
             <div className="absolute right-0 bottom-0 left-0 z-40 border-t bg-background/95 backdrop-blur">
-              <div className="mx-auto flex max-w-(--container-width-lg) flex-wrap items-center justify-center gap-2 px-4 py-4">
-                {modeButtons.map((button) => {
-                  const Icon = button.icon;
-                  return (
-                    <Button
-                      key={button.key}
-                      type="button"
-                      variant="outline"
-                      disabled={button.disabled || isUploading}
-                      onClick={button.onClick}
-                      className={cn(
-                        "gap-1.5 rounded-full border border-transparent bg-muted/40 text-muted-foreground shadow-none hover:bg-muted hover:text-foreground",
-                        button.active && button.activeClassName,
-                        button.disabled && "cursor-not-allowed opacity-80",
-                      )}
-                    >
-                      {button.active && <CheckIcon className="size-3.5" />}
-                      <Icon className="size-3.5" />
-                      <span className="text-xs font-normal">{button.label}</span>
-                    </Button>
-                  );
-                })}
-              </div>
+              <TooltipProvider delayDuration={250}>
+                <div className="mx-auto flex max-w-(--container-width-lg) flex-wrap items-center justify-center gap-0 px-0 py-1">
+                  {modeButtons.map((button) => {
+                    const Icon = button.icon;
+                    return (
+                      <Tooltip key={button.key} content={button.label}>
+                        <Button
+                          type="button"
+                          variant="ghost"
+                          size="icon"
+                          aria-label={button.label}
+                          disabled={button.disabled || isUploading}
+                          onClick={button.onClick}
+                          className={cn(
+                            "size-9 rounded-full text-muted-foreground shadow-none hover:bg-muted/60 hover:text-foreground",
+                            button.active && button.activeClassName,
+                            button.disabled && "cursor-not-allowed opacity-80",
+                          )}
+                        >
+                          <Icon className="size-4" />
+                        </Button>
+                      </Tooltip>
+                    );
+                  })}
+                </div>
+              </TooltipProvider>
             </div>
           </main>
         </div>
diff --git a/frontend-web/src/pages/KnowledgeBasePage.tsx b/frontend-web/src/pages/KnowledgeBasePage.tsx
index b12cfb5f..5380610c 100644
--- a/frontend-web/src/pages/KnowledgeBasePage.tsx
+++ b/frontend-web/src/pages/KnowledgeBasePage.tsx
@@ -16,6 +16,12 @@ import { Input } from "@/components/ui/input";
 import { ScrollArea } from "@/components/ui/scroll-area";
 import { Textarea } from "@/components/ui/textarea";
 import { EmptyDocumentState } from "@/components/wiki/EmptyDocumentState";
+import {
+  ExtractTemplateManageDialog,
+  FolderManageDialog,
+  FolderSelect,
+  TemplateSelect,
+} from "@/components/wiki/KnowledgeManage";
 import { KnowledgeSidebar } from "@/components/wiki/KnowledgeSidebar";
 import { WikiContent } from "@/components/wiki/WikiContent";
 import { WikiHeader } from "@/components/wiki/WikiHeader";
@@ -28,6 +34,7 @@ import { ModelNameSelect } from "@/components/workspace/model-name-select";
 import {
   useKnowledgeActions,
   useKnowledgeDuplicates,
+  useKnowledgeFolders,
   useKnowledgeNote,
   useKnowledgeNotes,
   useNoteWikilinks,
@@ -70,11 +77,13 @@ function CreateNoteDialog({
   const [title, setTitle] = useState("");
   const [tags, setTags] = useState("");
   const [content, setContent] = useState("");
+  const [folder, setFolder] = useState<string | null>(null);
 
   const reset = () => {
     setTitle("");
     setTags("");
     setContent("");
+    setFolder(null);
   };
 
   const handleSubmit = async () => {
@@ -91,6 +100,7 @@ function CreateNoteDialog({
           .map((t) => t.trim())
           .filter(Boolean),
         status: "approved",
+        folder,
       });
       toast.success("知识已创建");
       onOpenChange(false);
@@ -119,6 +129,10 @@ function CreateNoteDialog({
             value={tags}
             onChange={(e) => setTags(e.target.value)}
           />
+          <label className="flex items-center justify-between gap-2 text-sm">
+            目录
+            <FolderSelect value={folder} onChange={setFolder} triggerClassName="w-[14rem]" />
+          </label>
           <Textarea
             placeholder="Markdown 正文"
             rows={12}
@@ -152,6 +166,8 @@ function BatchImportDialog({
   const [skipExisting, setSkipExisting] = useState(true);
   // Optional model override for extraction (null = configured default).
   const [model, setModel] = useState<string | null>(null);
+  const [template, setTemplate] = useState<string | null>(null);
+  const [folder, setFolder] = useState<string | null>(null);
 
   const handleRun = async () => {
     try {
@@ -160,8 +176,21 @@ function BatchImportDialog({
         skip_existing: skipExisting,
         status: "approved",
         include_sources: true,
+        // Run in the background so the request returns immediately — a large
+        // batch otherwise blocks past the gateway timeout and shows "request
+        // failed". Threads stream into the list as they finish.
+        background: true,
         model_name: model,
+        template_id: template,
+        folder,
       });
+      if (result.queued) {
+        toast.success(
+          `已转入后台导入 ${result.processed} 个线程，完成后会陆续出现在左侧列表（可稍后刷新查看）`,
+        );
+        onOpenChange(false);
+        return;
+      }
       const base = `批量导入完成：新增 ${result.created}，跳过 ${result.skipped}，失败 ${result.failed}`;
       if (result.llm_fallback && result.llm_fallback > 0) {
         // The model failed on some threads — they were saved via rule-based
@@ -217,6 +246,25 @@ function BatchImportDialog({
               triggerClassName="h-9 w-[11rem]"
             />
           </label>
+          <label className="flex items-center justify-between gap-2 text-sm">
+            抽取模板
+            <TemplateSelect
+              value={template}
+              onChange={setTemplate}
+              scope="thread"
+              disabled={batchIngest.isPending}
+              triggerClassName="w-[11rem]"
+            />
+          </label>
+          <label className="flex items-center justify-between gap-2 text-sm">
+            归入目录
+            <FolderSelect
+              value={folder}
+              onChange={setFolder}
+              disabled={batchIngest.isPending}
+              triggerClassName="w-[11rem]"
+            />
+          </label>
         </div>
         <DialogFooter>
           <Button variant="outline" onClick={() => onOpenChange(false)}>
@@ -244,12 +292,16 @@ function ImportFileDialog({
   const [title, setTitle] = useState("");
   const [tags, setTags] = useState("");
   const [model, setModel] = useState<string | null>(null);
+  const [template, setTemplate] = useState<string | null>(null);
+  const [folder, setFolder] = useState<string | null>(null);
 
   const reset = () => {
     setFile(null);
     setTitle("");
     setTags("");
     setModel(null);
+    setTemplate(null);
+    setFolder(null);
   };
 
   const handleImport = async () => {
@@ -265,6 +317,8 @@ function ImportFileDialog({
           tags: tags.trim() || undefined,
           status: "approved",
           model_name: model,
+          template_id: template,
+          folder,
         },
       });
       const count = result.notes?.length ?? 1;
@@ -330,6 +384,25 @@ function ImportFileDialog({
               triggerClassName="h-9 w-[11rem]"
             />
           </label>
+          <label className="flex items-center justify-between gap-2 text-sm">
+            抽取模板
+            <TemplateSelect
+              value={template}
+              onChange={setTemplate}
+              scope="document"
+              disabled={ingestFile.isPending}
+              triggerClassName="w-[11rem]"
+            />
+          </label>
+          <label className="flex items-center justify-between gap-2 text-sm">
+            归入目录
+            <FolderSelect
+              value={folder}
+              onChange={setFolder}
+              disabled={ingestFile.isPending}
+              triggerClassName="w-[11rem]"
+            />
+          </label>
         </div>
         <DialogFooter>
           <Button variant="outline" onClick={() => onOpenChange(false)}>
@@ -478,6 +551,8 @@ export default function KnowledgeBasePage() {
   const [batchOpen, setBatchOpen] = useState(false);
   const [importOpen, setImportOpen] = useState(false);
   const [dedupOpen, setDedupOpen] = useState(false);
+  const [foldersOpen, setFoldersOpen] = useState(false);
+  const [templatesOpen, setTemplatesOpen] = useState(false);
   const [vaultPath, setVaultPath] = useState<string | null>(null);
 
   const params = useMemo(
@@ -485,14 +560,19 @@ export default function KnowledgeBasePage() {
       keyword: debouncedKeyword || undefined,
       source_type: sourceType === "all" ? undefined : sourceType,
       status: status === "all" ? undefined : status,
-      limit: 500,
+      limit: 1000,
     }),
     [debouncedKeyword, sourceType, status],
   );
 
   const { data, isLoading } = useKnowledgeNotes(params);
+  const { data: folders = [] } = useKnowledgeFolders();
   const notes = data?.items ?? [];
-  const tree = useMemo(() => buildKnowledgeTree(notes), [notes]);
+  const folderPaths = useMemo(() => folders.map((f) => f.path), [folders]);
+  const tree = useMemo(
+    () => buildKnowledgeTree(notes, folderPaths),
+    [notes, folderPaths],
+  );
 
   const { data: note, isLoading: noteLoading } = useKnowledgeNote(noteId);
   const { data: wikiLinks } = useNoteWikilinks(noteId);
@@ -571,6 +651,8 @@ export default function KnowledgeBasePage() {
         onDedup={() => setDedupOpen(true)}
         onGraph={() => navigate("/page/workspace/knowledge/graph")}
         onCreate={() => setCreateOpen(true)}
+        onManageFolders={() => setFoldersOpen(true)}
+        onManageTemplates={() => setTemplatesOpen(true)}
       />
 
       <div
@@ -636,6 +718,8 @@ export default function KnowledgeBasePage() {
       <ImportFileDialog open={importOpen} onOpenChange={setImportOpen} />
       <BatchImportDialog open={batchOpen} onOpenChange={setBatchOpen} />
       <DedupDialog open={dedupOpen} onOpenChange={setDedupOpen} />
+      <FolderManageDialog open={foldersOpen} onOpenChange={setFoldersOpen} />
+      <ExtractTemplateManageDialog open={templatesOpen} onOpenChange={setTemplatesOpen} />
       <VaultPageDialog
         path={vaultPath}
         onOpenChange={(open) => {
diff --git a/frontend-web/src/pages/LoginPage.tsx b/frontend-web/src/pages/LoginPage.tsx
index 8f33b1da..5b495ea0 100644
--- a/frontend-web/src/pages/LoginPage.tsx
+++ b/frontend-web/src/pages/LoginPage.tsx
@@ -1,6 +1,7 @@
 'use client';
 
 import { useParams, useRouter, useSearchParams } from 'next/navigation';
+import { useNavigate } from 'react-router-dom';
 import { useEffect, useMemo, useState } from 'react';
 
 import { useTheme } from '@/contexts/ThemeContext';
@@ -11,6 +12,11 @@ import {
   readExternalLoginParams,
   resolveThemeFromQuery,
 } from '@/core/auth/external-login';
+import {
+  fetchCopTaskDetail,
+  persistCopTaskDetail,
+  resolveTaskDeeplinkTarget,
+} from '@/core/auth/task-deeplink';
 import { resolveUsernameByOnlineTicket } from '@/core/auth/wujie-ticket-login';
 import { isWujieEnv, useWujieProps } from '@/core/embed/wujie';
 import {
@@ -34,6 +40,7 @@ type LoginStatus = 'loading' | 'success' | 'error';
 export default function LoginPage() {
   const { username } = useParams<{ username?: string }>();
   const router = useRouter();
+  const navigate = useNavigate();
   const searchParams = useSearchParams();
   const { setTheme } = useTheme();
   const sidebarEnabled = useMemo(
@@ -60,6 +67,9 @@ export default function LoginPage() {
   const gatePassword = searchParams.get('password') ?? '';
   // token 换登录:?authToken= 传上游 token(与 tools_token 的 ?token= 区分开)
   const authToken = searchParams.get('authToken')?.trim() ?? '';
+  // 任务深链:外部系统带 ?taskId=&goPath= 时,登录后拉任务详情并直送对应问答/会商页
+  const taskId = searchParams.get('taskId')?.trim() ?? '';
+  const goPath = searchParams.get('goPath')?.trim() ?? '';
   const externalParams = useMemo(
     () => ({
       theme: searchParams.get('theme') ?? '',
@@ -145,6 +155,27 @@ export default function LoginPage() {
         persistAccessToken(accessToken);
         persistSidebarEnabled(sidebarEnabled);
         setStatus('success');
+
+        // 任务深链：带 taskId + goPath 时拉任务详情并直送对应问答/会商页；
+        // 任一环节失败（无地址/拉取失败/未配置智能体/goPath 无法识别）都回退默认首页。
+        if (taskId && goPath) {
+          try {
+            const detail = await fetchCopTaskDetail(taskId);
+            if (cancelled) return;
+            if (detail) {
+              persistCopTaskDetail(detail);
+              const target = await resolveTaskDeeplinkTarget(goPath, detail);
+              if (cancelled) return;
+              if (target) {
+                navigate(target.path, { replace: true, state: target.state });
+                return;
+              }
+            }
+          } catch {
+            /* 深链失败不阻断登录，落默认首页 */
+          }
+        }
+
         router.replace(yUserId ? GENERAL_QA_ROUTE : redirectPath);
       } catch (error) {
         if (cancelled) {
@@ -164,6 +195,7 @@ export default function LoginPage() {
   }, [
     resolvedUsername,
     router,
+    navigate,
     sidebarEnabled,
     yUserId,
     toolToken,
@@ -171,6 +203,8 @@ export default function LoginPage() {
     onlineTicket,
     authToken,
     gatePassword,
+    taskId,
+    goPath,
   ]);
 
   return (
diff --git a/frontend-web/src/pages/PositionsPage.tsx b/frontend-web/src/pages/PositionsPage.tsx
new file mode 100644
index 00000000..9dee2046
--- /dev/null
+++ b/frontend-web/src/pages/PositionsPage.tsx
@@ -0,0 +1,5 @@
+import { PositionManagementPage } from "@/components/workspace/admin/position-management-page";
+
+export default function PositionsPage() {
+  return <PositionManagementPage />;
+}
diff --git a/frontend-web/src/pages/RecommendedQuestionsPage.tsx b/frontend-web/src/pages/RecommendedQuestionsPage.tsx
index 3877fd1f..c8d76bdf 100644
--- a/frontend-web/src/pages/RecommendedQuestionsPage.tsx
+++ b/frontend-web/src/pages/RecommendedQuestionsPage.tsx
@@ -8,11 +8,15 @@ import {
   Loader2Icon,
   PencilIcon,
   PlusIcon,
+  TagIcon,
   Trash2Icon,
 } from "lucide-react";
 import { toast } from "sonner";
 
 import { Button } from "@/components/ui/button";
+import { ResourceTagDialog } from "@/components/workspace/tags/resource-tag-dialog";
+import { TagChipList } from "@/components/workspace/tags/tag-chip";
+import { TagSearchBar } from "@/components/workspace/tags/tag-search-bar";
 import { useTheme } from "@/contexts/ThemeContext";
 import {
   Dialog,
@@ -24,9 +28,9 @@ import {
 import { Input } from "@/components/ui/input";
 import { Textarea } from "@/components/ui/textarea";
 import {
+  useAdminRecommendedQuestions,
   useCreateRecommendedQuestion,
   useDeleteRecommendedQuestion,
-  useRecommendedQuestions,
   useReorderRecommendedQuestions,
   useUpdateRecommendedQuestion,
 } from "@/core/recommended-questions/hooks";
@@ -65,7 +69,7 @@ export default function RecommendedQuestionsPage({
   embedded?: boolean;
 } = {}) {
   const { theme } = useTheme();
-  const { questions, isLoading, error } = useRecommendedQuestions();
+  const { questions, isLoading, error } = useAdminRecommendedQuestions();
   const createMutation = useCreateRecommendedQuestion();
   const updateMutation = useUpdateRecommendedQuestion();
   const deleteMutation = useDeleteRecommendedQuestion();
@@ -75,11 +79,31 @@ export default function RecommendedQuestionsPage({
   const [formTitle, setFormTitle] = useState("");
   const [formContent, setFormContent] = useState("");
   const [deleteTarget, setDeleteTarget] = useState<RecommendedQuestion | null>(null);
+  const [tagTarget, setTagTarget] = useState<RecommendedQuestion | null>(null);
 
   // drag state
   const [dragIndex, setDragIndex] = useState<number | null>(null);
   const [dragOverIndex, setDragOverIndex] = useState<number | null>(null);
 
+  // 名称 / 标签搜索
+  const [search, setSearch] = useState("");
+  const [tagIds, setTagIds] = useState<string[]>([]);
+
+  const isFiltering = search.trim() !== "" || tagIds.length > 0;
+  const needle = search.trim().toLowerCase();
+  const visibleQuestions = isFiltering
+    ? questions.filter((q) => {
+        const matchesText =
+          needle === "" ||
+          q.title.toLowerCase().includes(needle) ||
+          q.content.toLowerCase().includes(needle);
+        const matchesTags =
+          tagIds.length === 0 ||
+          (q.tags ?? []).some((t) => tagIds.includes(t.id));
+        return matchesText && matchesTags;
+      })
+    : questions;
+
   function openAdd() {
     setFormTitle("");
     setFormContent("");
@@ -200,6 +224,16 @@ export default function RecommendedQuestionsPage({
         </Button>
       </div>
 
+      {!isLoading && !error && questions.length > 0 && (
+        <TagSearchBar
+          targetType="recommended_question"
+          searchPlaceholder="搜索标题 / 内容"
+          onSearchChange={setSearch}
+          tagIds={tagIds}
+          onTagIdsChange={setTagIds}
+        />
+      )}
+
       {isLoading ? (
         <div className="flex items-center justify-center py-16">
           <Loader2Icon className="text-muted-foreground size-6 animate-spin" />
@@ -216,15 +250,19 @@ export default function RecommendedQuestionsPage({
             添加第一个问题
           </Button>
         </div>
+      ) : visibleQuestions.length === 0 ? (
+        <div className="text-muted-foreground flex flex-col items-center gap-3 rounded-xl border border-dashed py-16 text-sm dark:text-[#7FACF2] dark:!border-[#213F62]">
+          <p>没有匹配的推荐问题</p>
+        </div>
       ) : (
         <div className="flex flex-col gap-2 ">
-          {questions.map((q, index) => (
+          {visibleQuestions.map((q, index) => (
             <div
               key={q.id}
-              draggable
-              onDragStart={() => handleDragStart(index)}
-              onDragOver={(e) => handleDragOver(e, index)}
-              onDrop={() => handleDrop(index)}
+              draggable={!isFiltering}
+              onDragStart={() => !isFiltering && handleDragStart(index)}
+              onDragOver={(e) => !isFiltering && handleDragOver(e, index)}
+              onDrop={() => !isFiltering && handleDrop(index)}
               onDragEnd={() => {
                 setDragIndex(null);
                 setDragOverIndex(null);
@@ -235,7 +273,13 @@ export default function RecommendedQuestionsPage({
                   : "border-[#e4e4e7]"
               }`}
             >
-              <GripVerticalIcon className="text-muted-foreground/40 mt-0.5 size-4 shrink-0 cursor-grab active:cursor-grabbing dark:text-[#566B7F]" />
+              <GripVerticalIcon
+                className={`text-muted-foreground/40 mt-0.5 size-4 shrink-0 dark:text-[#566B7F] ${
+                  isFiltering
+                    ? "cursor-not-allowed opacity-40"
+                    : "cursor-grab active:cursor-grabbing"
+                }`}
+              />
               <div className="min-w-0 flex-1">
                 <div className="truncate text-sm font-medium text-[#18181b] dark:text-white">
                   {q.title}
@@ -243,8 +287,21 @@ export default function RecommendedQuestionsPage({
                 <div className="text-muted-foreground mt-0.5 line-clamp-2 text-xs dark:text-[#7FACF2]">
                   {q.content}
                 </div>
+                {(q.tags?.length ?? 0) > 0 && (
+                  <TagChipList tags={q.tags!} className="mt-2" />
+                )}
               </div>
               <div className="flex shrink-0 items-center gap-1">
+                <Button
+                  variant="ghost"
+                  size="icon"
+                  className="size-7"
+                  onClick={() => setTagTarget(q)}
+                  disabled={isSaving}
+                  title="打标签（按岗位分发）"
+                >
+                  <TagIcon className="size-3.5" />
+                </Button>
                 <Button
                   variant="ghost"
                   size="icon"
@@ -308,6 +365,17 @@ export default function RecommendedQuestionsPage({
         </DialogContent>
       </Dialog>
 
+      {/* 打标签弹窗（按岗位分发推荐问题） */}
+      {tagTarget && (
+        <ResourceTagDialog
+          open={tagTarget !== null}
+          onOpenChange={(o) => !o && setTagTarget(null)}
+          targetType="recommended_question"
+          targetId={tagTarget.id}
+          title={`为「${tagTarget.title}」打标签`}
+        />
+      )}
+
       {/* 删除确认弹窗 */}
       <Dialog
         open={deleteTarget !== null}
diff --git a/frontend-web/src/pages/SkillCompressionSettingsPage.tsx b/frontend-web/src/pages/SkillCompressionSettingsPage.tsx
new file mode 100644
index 00000000..4a3a4378
--- /dev/null
+++ b/frontend-web/src/pages/SkillCompressionSettingsPage.tsx
@@ -0,0 +1,141 @@
+"use client";
+
+import { Loader2Icon } from "lucide-react";
+import { useEffect, useRef, useState } from "react";
+import { toast } from "sonner";
+
+import { Button } from "@/components/ui/button";
+import { Input } from "@/components/ui/input";
+import { Switch } from "@/components/ui/switch";
+import {
+  useSkillCompressionSettings,
+  useUpdateSkillCompressionSettings,
+} from "@/core/system-settings/hooks";
+
+/**
+ * 管理员「技能压缩」配置页。
+ *
+ * 开启后,只有被标记为「常驻(免压缩)」的技能会全文注入系统提示词;其余已启用
+ * 技能会被压缩(从提示词正文剔除),由智能体在需要时通过 `search_skills` 工具
+ * 按关键词检索调出。关闭时行为不变 —— 所有已启用技能全文注入。
+ */
+export default function SkillCompressionSettingsPage({
+  embedded = false,
+}: {
+  embedded?: boolean;
+} = {}) {
+  const { data, isLoading, error } = useSkillCompressionSettings();
+  const update = useUpdateSkillCompressionSettings();
+
+  const [enabled, setEnabled] = useState(false);
+  const [keepIndex, setKeepIndex] = useState(true);
+  const [searchTopN, setSearchTopN] = useState(5);
+  const initializedRef = useRef(false);
+
+  useEffect(() => {
+    if (!data || initializedRef.current) return;
+    setEnabled(data.enabled);
+    setKeepIndex(data.keep_index_in_prompt);
+    setSearchTopN(data.search_top_n);
+    initializedRef.current = true;
+  }, [data]);
+
+  const handleSave = async () => {
+    const topN = Math.max(1, Math.min(Math.round(searchTopN) || 5, 50));
+    try {
+      await update.mutateAsync({
+        enabled,
+        keep_index_in_prompt: keepIndex,
+        search_top_n: topN,
+      });
+      setSearchTopN(topN);
+      toast.success("已保存");
+    } catch (e) {
+      toast.error(e instanceof Error ? e.message : "保存失败");
+    }
+  };
+
+  return (
+    <div
+      className={
+        embedded
+          ? "flex w-full flex-col gap-6"
+          : "mx-auto flex w-full max-w-2xl flex-col gap-6 px-6 py-10"
+      }
+    >
+      <div>
+        <h1 className={embedded ? "text-base font-semibold" : "text-xl font-semibold"}>
+          技能压缩配置
+        </h1>
+        <p className="text-muted-foreground mt-1 text-sm">
+          技能过多会让系统提示词变得臃肿、增加 token 消耗。开启「技能压缩」后,只有被标记为
+          「常驻(免压缩)」的技能会全文注入系统提示词;其余已启用技能会被压缩,智能体在需要时
+          通过 <code>search_skills</code> 工具按关键词检索调出。关闭时行为不变 ——
+          所有已启用技能全文注入。某个技能是否常驻,在该技能的详情页单独开启。
+        </p>
+      </div>
+
+      {isLoading ? (
+        <div className="flex items-center justify-center py-16">
+          <Loader2Icon className="text-muted-foreground size-6 animate-spin" />
+        </div>
+      ) : error ? (
+        <div className="text-destructive rounded-xl border border-dashed py-16 text-center text-sm">
+          加载失败：{error.message}
+        </div>
+      ) : (
+        <div className="flex flex-col gap-5 rounded-xl border bg-white px-5 py-5 dark:border-[#213D5E] dark:bg-[#02244C] dark:text-white">
+          <div className="flex items-center justify-between gap-4">
+            <div>
+              <div className="text-sm font-medium">启用技能压缩</div>
+              <div className="text-muted-foreground mt-0.5 text-xs">
+                关闭时,所有已启用技能照旧全文注入系统提示词(默认行为)。
+              </div>
+            </div>
+            <Switch checked={enabled} onCheckedChange={setEnabled} />
+          </div>
+
+          <div className="flex items-center justify-between gap-4">
+            <div>
+              <div className="text-sm font-medium">保留技能名索引</div>
+              <div className="text-muted-foreground mt-0.5 text-xs">
+                开启后,被压缩的技能仍以「仅名字」的紧凑列表出现在提示词里,帮助智能体知道存在
+                哪些技能;关闭则完全隐藏,只能靠 search_skills 检索发现。
+              </div>
+            </div>
+            <Switch
+              checked={keepIndex}
+              disabled={!enabled}
+              onCheckedChange={setKeepIndex}
+            />
+          </div>
+
+          <div className="flex flex-col gap-1.5">
+            <p className="text-sm font-medium">检索返回条数</p>
+            <p className="text-muted-foreground text-xs">
+              search_skills 工具每次最多返回的匹配技能数量(1–50)。
+            </p>
+            <Input
+              type="number"
+              min={1}
+              max={50}
+              value={searchTopN}
+              disabled={!enabled}
+              onChange={(e) => setSearchTopN(Number(e.target.value))}
+              className="w-28"
+            />
+          </div>
+
+          <div className="flex justify-end">
+            <Button onClick={handleSave} disabled={update.isPending}>
+              {update.isPending && (
+                <Loader2Icon className="mr-1.5 size-4 animate-spin" />
+              )}
+              保存
+            </Button>
+          </div>
+        </div>
+      )}
+    </div>
+  );
+}
diff --git a/frontend-web/src/pages/StrategyRoutes.tsx b/frontend-web/src/pages/StrategyRoutes.tsx
index af149f3a..2e17bcdf 100644
--- a/frontend-web/src/pages/StrategyRoutes.tsx
+++ b/frontend-web/src/pages/StrategyRoutes.tsx
@@ -1,4 +1,4 @@
-import { Navigate, Route, Routes } from "react-router-dom";
+import { Navigate, Route, Routes, useParams } from "react-router-dom";
 import { Toaster } from "sonner";
 
 import { SidebarProvider } from "@/components/ui/sidebar";
@@ -19,6 +19,8 @@ import BusinessChainListPage from "@/roundtable-planning/pages/BusinessChainList
 import BusinessChainEditorPage from "@/roundtable-planning/pages/BusinessChainEditorPage";
 import "@/roundtable-planning/styles/roundtable-planning.css";
 import AgentChatPage from "./AgentChatPage";
+import { AgentModuleRoutes } from "./AgentModulePage";
+import { AGENT_MODULES } from "@/components/workspace/agents/agent-module-sidebar";
 import AgentsPage from "./AgentsPage";
 import BusinessChainMappingPage from "./BusinessChainMappingPage";
 import ChatPage from "./ChatPage";
@@ -41,6 +43,7 @@ import ScheduledTaskRunDetailPage from "./ScheduledTaskRunDetailPage";
 import ScheduledTasksPage from "./ScheduledTasksPage";
 import SkillEvolutionPage from "./SkillEvolutionPage";
 import SkillsPage from "./SkillsPage";
+import PositionsPage from "./PositionsPage";
 import TagsPage from "./TagsPage";
 import UsersPage from "./UsersPage";
 import LlmMetricsPage from "./LlmMetricsPage";
@@ -48,6 +51,20 @@ import AdminScheduledTasksPage from "./AdminScheduledTasksPage";
 import AdminUserActivityPage from "./AdminUserActivityPage";
 import { ChatRuntime } from "./WorkspaceRoutes";
 
+/**
+ * 任务深链：把智能体问答嵌进「任务管理」布局。读路由参数 agent_id，给 AgentChatPage
+ * 传入以 taskBase 为前缀的对话路由（开新会话 / URL 改写都停留在任务管理上下文，
+ * 侧边栏保持对应业务高亮）。
+ */
+function TaskEmbeddedAgentChat({ taskBase }: { taskBase: string }) {
+  const { agent_id } = useParams<{ agent_id: string }>();
+  return (
+    <ChatRuntime>
+      <AgentChatPage chatBasePath={`${taskBase}/agent/${agent_id}/chats`} />
+    </ChatRuntime>
+  );
+}
+
 export default function StrategyRoutes() {
   return (
     <RouteBaseProvider value="/page/strategy">
@@ -75,6 +92,10 @@ export default function StrategyRoutes() {
               }
             />
 
+            {/* 需求 0616 #3/#4：zzJC / zzXD 「按类型聚合」模块页（重做，独立路由）。 */}
+            <Route path="qa/zzjc/*" element={<AgentModuleRoutes module={AGENT_MODULES.zzjc} />} />
+            <Route path="qa/zzxd/*" element={<AgentModuleRoutes module={AGENT_MODULES.zzxd} />} />
+
             <Route path="task-management" element={<TaskManagement />}>
               <Route index element={<Navigate to="xk-task-jc" replace />} />
               <Route
@@ -107,6 +128,21 @@ export default function StrategyRoutes() {
                   />
                 }
               />
+              {/* 任务深链（3qfx / 3qfx-A）：把智能体问答 / 会商嵌进 3QFX 业务上下文。 */}
+              <Route
+                path="plan-task-sentiment/roundtable"
+                element={
+                  <ChatRuntime>
+                    <RoundtablePlanningPage />
+                  </ChatRuntime>
+                }
+              />
+              <Route
+                path="plan-task-sentiment/agent/:agent_id/chats/:thread_id"
+                element={
+                  <TaskEmbeddedAgentChat taskBase="/page/strategy/task-management/plan-task-sentiment" />
+                }
+              />
               <Route
                 path="plan-task/plan-task-fx"
                 element={
@@ -117,6 +153,21 @@ export default function StrategyRoutes() {
                   />
                 }
               />
+              {/* 任务深链（rwfx / rwfx-A）：嵌进 任务FX 业务上下文。 */}
+              <Route
+                path="plan-task/plan-task-fx/roundtable"
+                element={
+                  <ChatRuntime>
+                    <RoundtablePlanningPage />
+                  </ChatRuntime>
+                }
+              />
+              <Route
+                path="plan-task/plan-task-fx/agent/:agent_id/chats/:thread_id"
+                element={
+                  <TaskEmbeddedAgentChat taskBase="/page/strategy/task-management/plan-task/plan-task-fx" />
+                }
+              />
               <Route
                 path="plan-task/plan-task-action"
                 element={
@@ -127,6 +178,21 @@ export default function StrategyRoutes() {
                   />
                 }
               />
+              {/* 任务深链（xdfx / xdfx-A）：嵌进 XDFX 业务上下文。 */}
+              <Route
+                path="plan-task/plan-task-action/roundtable"
+                element={
+                  <ChatRuntime>
+                    <RoundtablePlanningPage />
+                  </ChatRuntime>
+                }
+              />
+              <Route
+                path="plan-task/plan-task-action/agent/:agent_id/chats/:thread_id"
+                element={
+                  <TaskEmbeddedAgentChat taskBase="/page/strategy/task-management/plan-task/plan-task-action" />
+                }
+              />
               <Route path="mnty-task-mechanism" element={<TaskTyFrom defaultPage="iframeXdty" title="XDTY任务" />} />
               <Route path="mnty-task-plan" element={<TaskTyFrom defaultPage="iframeFaty" title="方案TY任务" />} />
               <Route path="mnty-task-zsty" element={<TaskTyFrom defaultPage="iframeSysystem" title="XXTY任务" />} />
@@ -289,6 +355,16 @@ export default function StrategyRoutes() {
               path="admin/tags"
               element={<TagsPage />}
             />
+            <Route
+              path="admin/positions"
+              element={
+                getStoredAuth()?.system_role === "admin" ? (
+                  <PositionsPage />
+                ) : (
+                  <Navigate to="/page/workspace/chats/new" replace />
+                )
+              }
+            />
           </Routes>
         </div>
       </SidebarProvider>
diff --git a/frontend-web/src/pages/StudioNotebookDetailPage.tsx b/frontend-web/src/pages/StudioNotebookDetailPage.tsx
index 763989bb..83d367f4 100644
--- a/frontend-web/src/pages/StudioNotebookDetailPage.tsx
+++ b/frontend-web/src/pages/StudioNotebookDetailPage.tsx
@@ -210,8 +210,15 @@ function SourcesDialog({
 }
 
 // ── 主页面 ──────────────────────────────────────────────
-export default function StudioNotebookDetailPage() {
-  const { notebookId } = useParams<{ notebookId: string }>();
+export default function StudioNotebookDetailPage({
+  notebookId: notebookIdProp,
+}: {
+  /** 内嵌使用（如课题研究右侧）时直接传入；缺省则从路由参数取。 */
+  notebookId?: number;
+} = {}) {
+  const params = useParams<{ notebookId: string }>();
+  const notebookId =
+    notebookIdProp != null ? String(notebookIdProp) : params.notebookId;
   const { theme } = useTheme(); // 跳转「我的空间」时带上当前主题 key
   const { notebooks, isLoading, token: studioToken } = useStudioNotebooks(); // token 用于跳转 callback
   const { models } = useModels();
diff --git a/frontend-web/src/pages/StudioNotebooksPage.tsx b/frontend-web/src/pages/StudioNotebooksPage.tsx
index 6f43a561..40f444a5 100644
--- a/frontend-web/src/pages/StudioNotebooksPage.tsx
+++ b/frontend-web/src/pages/StudioNotebooksPage.tsx
@@ -44,21 +44,17 @@ function NotebookCard({
   isDark,
   onRename,
   onDelete,
+  onOpen,
 }: {
   nb: StudioNotebook;
   isDark: boolean;
   onRename: (nb: StudioNotebook) => void;
   onDelete: (nb: StudioNotebook) => void;
+  /** 传入则点击卡片回调（内嵌打开），不再路由跳转。 */
+  onOpen?: (id: number) => void;
 }) {
-  return (
-    <div
-      className={`group relative flex flex-col rounded-[20px] transition-colors ${isDark ? "bg-[#1e2330] hover:bg-[#262c3b]" : "bg-[#f2f2f2] hover:bg-[#e8e8e8]"}`}
-      style={{ minHeight: 180 }}
-    >
-      <Link
-        to={`/page/workspace/studio/notebooks/${nb.id}`}
-        className="flex flex-1 flex-col p-5"
-      >
+  const inner = (
+    <>
         <div className="flex items-start justify-between">
           <div className="flex size-11 items-center justify-center text-3xl leading-none select-none">
             {nb.avatar_url ? (
@@ -76,7 +72,27 @@ function NotebookCard({
           </span>
           <span className={`text-[13px] ${isDark ? "text-[#8b93a7]" : "text-[#666]"}`}>{nb.source_count} 个来源</span>
         </div>
-      </Link>
+    </>
+  );
+
+  return (
+    <div className={`group relative flex flex-col rounded-[20px] transition-colors ${isDark ? "bg-[#1e2330] hover:bg-[#262c3b]" : "bg-[#f2f2f2] hover:bg-[#e8e8e8]"}`} style={{ minHeight: 180 }}>
+      {onOpen ? (
+        <button
+          type="button"
+          onClick={() => onOpen(nb.id)}
+          className="flex flex-1 flex-col p-5 text-left"
+        >
+          {inner}
+        </button>
+      ) : (
+        <Link
+          to={`/page/workspace/studio/notebooks/${nb.id}`}
+          className="flex flex-1 flex-col p-5"
+        >
+          {inner}
+        </Link>
+      )}
 
       {/* three-dot menu — absolute so it doesn't shift layout */}
       <div className="absolute right-4 top-4">
@@ -108,7 +124,12 @@ function NotebookCard({
   );
 }
 
-export default function StudioNotebooksPage() {
+export default function StudioNotebooksPage({
+  onOpenNotebook,
+}: {
+  /** 内嵌使用（课题研究右侧）时传入：点击空间卡片回调，不跳路由。 */
+  onOpenNotebook?: (id: number) => void;
+} = {}) {
   const { notebooks, isLoading, isError } = useStudioNotebooks();
   const { theme } = useTheme();
   const isDark = theme === "dark";
@@ -216,6 +237,7 @@ export default function StudioNotebooksPage() {
               isDark={isDark}
               onRename={(nb) => { setRenameTarget(nb); setRenameName(nb.name); setRenameError(""); }}
               onDelete={(nb) => { setDeleteTarget(nb); setDeleteError(""); }}
+              onOpen={onOpenNotebook}
             />
           ))}
       </div>
diff --git a/frontend-web/src/pages/WorkspaceRoutes.tsx b/frontend-web/src/pages/WorkspaceRoutes.tsx
index 30ea1f35..cfaac448 100644
--- a/frontend-web/src/pages/WorkspaceRoutes.tsx
+++ b/frontend-web/src/pages/WorkspaceRoutes.tsx
@@ -1,5 +1,5 @@
 import type { CSSProperties, ReactNode } from 'react';
-import { Navigate, Route, Routes, useParams } from 'react-router-dom';
+import { Navigate, Route, Routes, useLocation, useParams } from 'react-router-dom';
 
 import { getStoredAuth } from '@/core/auth';
 
@@ -32,6 +32,7 @@ import MemoryPage from './MemoryPage';
 import PromptPrefixSettingsPage from './PromptPrefixSettingsPage';
 import PrivateSquareSettingsPage from './PrivateSquareSettingsPage';
 import PublishSquaresSettingsPage from './PublishSquaresSettingsPage';
+import SkillCompressionSettingsPage from './SkillCompressionSettingsPage';
 import SkillEvolutionPage from './SkillEvolutionPage';
 import SkillsPage from './SkillsPage';
 import UsersPage from './UsersPage';
@@ -39,6 +40,7 @@ import AdminUserActivityPage from './AdminUserActivityPage';
 import AdminMemoryPage from './AdminMemoryPage';
 import AdminUserHubPage from './AdminUserHubPage';
 import TagsPage from './TagsPage';
+import PositionsPage from './PositionsPage';
 import RecommendedQuestionsPage from './RecommendedQuestionsPage';
 import AIWritingPage from '@/open-canvas/pages/AIWritingPage';
 import StudioNotebooksPage from './StudioNotebooksPage';
@@ -87,6 +89,10 @@ export function WorkspaceLayout({
   insetOverflowHidden?: boolean;
   pageSidebarDark?: boolean;
 }) {
+  // 全量模式下通用问答（/page/workspace/chats*）的历史侧栏默认展开；其余紧凑页
+  // （如智能体对话，为聊天区更宽而默认收起）维持原默认。
+  const isGeneralQa = useLocation().pathname.startsWith("/page/workspace/chats");
+
   const pageSidebarWidth = pageSidebarHidden
     ? '0rem'
     : pageSidebarDark
@@ -113,7 +119,9 @@ export function WorkspaceLayout({
       }
       key={String(compactWorkspaceSidebar)}
       data-page-sidebar-collapsed={pageSidebarCollapsed}
-      defaultOpen={defaultHistorySidebarOpen ?? !compactWorkspaceSidebar}
+      defaultOpen={
+        defaultHistorySidebarOpen ?? (isGeneralQa || !compactWorkspaceSidebar)
+      }
     >
       {!workspaceSidebarHidden && (
         <WorkspaceSidebar compactChrome={compactWorkspaceSidebar} />
@@ -139,6 +147,21 @@ export function ChatRuntime({ children }: { children: ReactNode }) {
   );
 }
 
+/**
+ * 通用问答新对话内容区：默认渲染 ChatPage；当 URL 带 ``?agent=<id>``（点击左栏
+ * 「常用智能体」时由侧边栏写入）时，**在右侧问答区原地嵌入**该智能体的对话页
+ * （AgentChatPage），路由仍停留在 /chats/new，不跳转出去、不更换左侧栏。
+ * key 绑定 location.key —— 每次点击（即便同一 agent）都重开一个干净的会话。
+ */
+function NewChatContent() {
+  const location = useLocation();
+  const agentId = new URLSearchParams(location.search).get('agent');
+  if (agentId) {
+    return <AgentChatPage key={`${agentId}:${location.key}`} embeddedAgentId={agentId} />;
+  }
+  return <ChatPage />;
+}
+
 function NewChatEntryRoute({
   layoutProps,
 }: {
@@ -157,7 +180,7 @@ function NewChatEntryRoute({
   return (
     <WorkspaceLayout {...layoutProps} defaultHistorySidebarOpen>
       <ChatRuntime>
-        <ChatPage />
+        <NewChatContent />
       </ChatRuntime>
     </WorkspaceLayout>
   );
@@ -475,6 +498,18 @@ export default function WorkspaceRoutes({
             </WorkspaceLayout>
           }
         />
+        <Route
+          path="admin/positions"
+          element={
+            getStoredAuth()?.system_role === 'admin' ? (
+              <WorkspaceLayout {...layoutProps}>
+                <PositionsPage />
+              </WorkspaceLayout>
+            ) : (
+              <Navigate to="/page/workspace/chats/new" replace />
+            )
+          }
+        />
         <Route
           path="settings/recommended-questions"
           element={
@@ -523,6 +558,18 @@ export default function WorkspaceRoutes({
             )
           }
         />
+        <Route
+          path="settings/skill-compression"
+          element={
+            getStoredAuth()?.system_role === 'admin' ? (
+              <WorkspaceLayout {...layoutProps}>
+                <SkillCompressionSettingsPage />
+              </WorkspaceLayout>
+            ) : (
+              <Navigate to="/page/workspace/chats/new" replace />
+            )
+          }
+        />
         <Route
           path="settings/llm-metrics"
           element={
diff --git a/frontend-web/src/roundtable-planning/components/BusinessChainListDialog.tsx b/frontend-web/src/roundtable-planning/components/BusinessChainListDialog.tsx
index 5a7faa6f..1b813bda 100644
--- a/frontend-web/src/roundtable-planning/components/BusinessChainListDialog.tsx
+++ b/frontend-web/src/roundtable-planning/components/BusinessChainListDialog.tsx
@@ -29,6 +29,7 @@ import {
   DialogTitle,
 } from "@/components/ui/dialog";
 import { cn } from "@/lib/utils";
+import { useDesensitizeSubtree } from "@/roundtable-planning/lib/use-desensitize-subtree";
 
 import { listAgents } from "@/core/agents";
 
@@ -212,9 +213,11 @@ export function BusinessChainListDialog({
   onOpenChange,
   onConfigure,
 }: BusinessChainListDialogProps) {
+  // 整片脱敏：弹窗内链条列表（链条名/描述/席位名/标题）。
+  const dzRef = useDesensitizeSubtree<HTMLDivElement>();
   return (
     <Dialog open={open} onOpenChange={onOpenChange}>
-      <DialogContent className="flex max-h-[85vh] flex-col gap-0 sm:max-w-2xl">
+      <DialogContent ref={dzRef} className="flex max-h-[85vh] flex-col gap-0 sm:max-w-2xl">
         <DialogHeader className="shrink-0">
           <DialogTitle className="flex items-center gap-2">
             <LayersIcon className="size-5 text-primary" />
diff --git a/frontend-web/src/roundtable-planning/components/BusinessChainPicker.tsx b/frontend-web/src/roundtable-planning/components/BusinessChainPicker.tsx
index f484cc6f..60dda10c 100644
--- a/frontend-web/src/roundtable-planning/components/BusinessChainPicker.tsx
+++ b/frontend-web/src/roundtable-planning/components/BusinessChainPicker.tsx
@@ -26,6 +26,7 @@ import {
 import { Button } from "@/components/ui/button";
 import { DialogFooter } from "@/components/ui/dialog";
 import { cn } from "@/lib/utils";
+import { useDesensitizeSubtree } from "@/roundtable-planning/lib/use-desensitize-subtree";
 import { listAgents } from "@/core/agents";
 
 import { filterRecommendCandidates } from "../api/recommend";
@@ -161,9 +162,12 @@ export function BusinessChainPicker({
     onConfirm({ agents, mode: "chain", chain: chainMeta });
   };
 
+  // 整片脱敏：链条卡片区（链条名/描述/席位智能体名）。输入框不受影响（Hook 跳过）。
+  const dzRef = useDesensitizeSubtree<HTMLElement>();
+
   return (
     <>
-      <section className="flex-1 min-h-0 flex flex-col overflow-hidden">
+      <section ref={dzRef} className="flex-1 min-h-0 flex flex-col overflow-hidden">
         {/* 筛选：我的 / 公共 */}
         <div className="mb-2 flex shrink-0 items-center gap-1 self-start rounded-lg border border-border bg-muted/40 p-0.5">
           {(["mine", "public"] as const).map((s) => (
diff --git a/frontend-web/src/roundtable-planning/components/DagFlowChart.tsx b/frontend-web/src/roundtable-planning/components/DagFlowChart.tsx
index 5570d03b..59ca95f3 100644
--- a/frontend-web/src/roundtable-planning/components/DagFlowChart.tsx
+++ b/frontend-web/src/roundtable-planning/components/DagFlowChart.tsx
@@ -21,6 +21,7 @@
  */
 import { CheckIcon, ChevronRightIcon, Loader2Icon } from "lucide-react";
 
+import { desensitize } from "@/lib/desensitize";
 import { cn } from "@/lib/utils";
 
 import { getAvatar } from "./Avatars";
@@ -137,7 +138,7 @@ export function DagFlowChart({
                   return (
                     <FlowNode
                       key={agentId}
-                      label={meta.name}
+                      label={desensitize(meta.name)}
                       avatarType={meta.avatarType}
                       state={state}
                       compact={compact}
diff --git a/frontend-web/src/roundtable-planning/components/DagFlowMiniPanel.tsx b/frontend-web/src/roundtable-planning/components/DagFlowMiniPanel.tsx
index 7cebe2a6..83392a38 100644
--- a/frontend-web/src/roundtable-planning/components/DagFlowMiniPanel.tsx
+++ b/frontend-web/src/roundtable-planning/components/DagFlowMiniPanel.tsx
@@ -20,6 +20,7 @@
 import { useCallback, useEffect, useRef, useState } from "react";
 import { CheckIcon, Loader2Icon, Network, X } from "lucide-react";
 
+import { desensitize } from "@/lib/desensitize";
 import { cn } from "@/lib/utils";
 
 import { getAvatar } from "./Avatars";
@@ -400,7 +401,7 @@ export function DagFlowMiniPanel({
                     return (
                       <MiniNode
                         key={agentId}
-                        label={meta.name}
+                        label={desensitize(meta.name)}
                         avatarType={meta.avatarType}
                         state={state}
                       />
diff --git a/frontend-web/src/roundtable-planning/components/DispatchChainModal.tsx b/frontend-web/src/roundtable-planning/components/DispatchChainModal.tsx
index c95e82a6..c3d988d6 100644
--- a/frontend-web/src/roundtable-planning/components/DispatchChainModal.tsx
+++ b/frontend-web/src/roundtable-planning/components/DispatchChainModal.tsx
@@ -26,6 +26,7 @@ import {
   DialogHeader,
   DialogTitle,
 } from "@/components/ui/dialog";
+import { desensitize } from "@/lib/desensitize";
 import { cn } from "@/lib/utils";
 
 import { getAvatar } from "./Avatars";
@@ -236,7 +237,7 @@ export function DispatchChainModal({
                     done={node.state === "done"}
                   />
                   <FlowNode
-                    label={node.name}
+                    label={desensitize(node.name)}
                     avatarType={node.avatarType}
                     state={node.state}
                     order={node.order}
@@ -277,7 +278,7 @@ export function DispatchChainModal({
                 <div key={d.id} className="flex gap-2.5">
                   <div className="shrink-0">{getAvatar(avatarForDialogue(d), "h-7 w-7")}</div>
                   <div className="min-w-0 flex-1">
-                    <div className="mb-0.5 text-xs font-semibold text-foreground/80">{d.name}</div>
+                    <div className="mb-0.5 text-xs font-semibold text-foreground/80">{desensitize(d.name)}</div>
                     <div className="prose prose-sm max-w-none break-words text-sm leading-relaxed text-foreground/80 dark:prose-invert">
                       <MarkdownContent>{d.content || "…"}</MarkdownContent>
                     </div>
diff --git a/frontend-web/src/roundtable-planning/components/HighFidelityReport.tsx b/frontend-web/src/roundtable-planning/components/HighFidelityReport.tsx
index f12a0bb5..90903387 100644
--- a/frontend-web/src/roundtable-planning/components/HighFidelityReport.tsx
+++ b/frontend-web/src/roundtable-planning/components/HighFidelityReport.tsx
@@ -32,6 +32,7 @@ import {
   type ThreadIdMap,
 } from "@/roundtable-planning/api/multi-agent";
 import { ConsensusDocxDownloadButton } from "@/roundtable-planning/components/ConsensusDocxDownloadButton";
+import { desensitize } from "@/lib/desensitize";
 import { markdownComponents } from "@/roundtable-planning/components/markdown-components";
 import {
   Step3ReportMode,
@@ -707,7 +708,7 @@ export const HighFidelityReport: React.FC<HighFidelityReportProps> = ({
                         lit ? "text-slate-800 dark:text-foreground" : "text-slate-400 dark:text-muted-foreground",
                       )}
                     >
-                      {seat.displayName}
+                      {desensitize(seat.displayName)}
                     </span>
                     <span className="block truncate text-[9.5px] font-mono text-slate-400 dark:text-muted-foreground">
                       {lit ? `${seat.delivery!.length} 字` : "无交付"}
@@ -771,7 +772,7 @@ export const HighFidelityReport: React.FC<HighFidelityReportProps> = ({
                   {consensusText ? (
                     <div className="text-xs text-slate-700 leading-relaxed dark:text-foreground/80">
                       <ReactMarkdown remarkPlugins={[remarkGfm]} components={markdownComponents}>
-                        {consensusText}
+                        {desensitize(consensusText)}
                       </ReactMarkdown>
                     </div>
                   ) : (
@@ -830,7 +831,7 @@ export const HighFidelityReport: React.FC<HighFidelityReportProps> = ({
                         {getAvatar(seat.avatarType, "h-5.5 w-5.5")}
                       </span>
                       <span className="text-xs font-black text-slate-800 truncate dark:text-foreground">
-                        {seat.displayName}
+                        {desensitize(seat.displayName)}
                       </span>
                     </div>
                     <span className="text-[10px] text-slate-400 font-mono shrink-0 dark:text-muted-foreground">
@@ -858,7 +859,7 @@ export const HighFidelityReport: React.FC<HighFidelityReportProps> = ({
                             remarkPlugins={[remarkGfm]}
                             components={markdownComponents}
                           >
-                            {seat.delivery!}
+                            {desensitize(seat.delivery!)}
                           </ReactMarkdown>
                         ) : (
                           <p className="text-[11px] text-slate-400 italic dark:text-muted-foreground">该席位暂无交付内容。</p>
diff --git a/frontend-web/src/roundtable-planning/components/MessageBubble.tsx b/frontend-web/src/roundtable-planning/components/MessageBubble.tsx
index 06e90626..89748b3a 100644
--- a/frontend-web/src/roundtable-planning/components/MessageBubble.tsx
+++ b/frontend-web/src/roundtable-planning/components/MessageBubble.tsx
@@ -18,6 +18,7 @@ import ReactMarkdown from "react-markdown";
 import remarkGfm from "remark-gfm";
 import { BrainIcon, ChevronDownIcon, Loader2 } from "lucide-react";
 
+import { desensitize } from "@/lib/desensitize";
 import { cn } from "@/lib/utils";
 
 export interface MessageBubbleProps {
@@ -58,7 +59,8 @@ export function MessageBubble({
                 senderNameClassName,
               )}
             >
-              {senderName}
+              {/* 智能体名称需脱敏（消息列表里唯一要脱敏的就是名称）。 */}
+              {desensitize(senderName)}
             </span>
             {badge && (
               <span className="roundtable-chat-meta font-mono font-black uppercase text-[var(--brand-900)] bg-[var(--brand-200)] border border-red-200 px-1.5 py-0.2 rounded">
@@ -190,6 +192,7 @@ export const sharedMarkdownComponents = {
  * 便捷封装:用 sharedMarkdownComponents + remarkGfm 渲染 markdown。
  * 大多数 caller 都不需要再覆盖任何 components,直接用这个就好。
  */
+// 消息列表正文**不脱敏**（仅名称脱敏）；MarkdownContent 是会商对话气泡正文渲染器，故原文渲染。
 export function MarkdownContent({ children }: { children: string }) {
   return (
     <ReactMarkdown remarkPlugins={[remarkGfm]} components={sharedMarkdownComponents}>
diff --git a/frontend-web/src/roundtable-planning/components/ParallelSeatStrip.tsx b/frontend-web/src/roundtable-planning/components/ParallelSeatStrip.tsx
index 5e03c3e1..d5b63bce 100644
--- a/frontend-web/src/roundtable-planning/components/ParallelSeatStrip.tsx
+++ b/frontend-web/src/roundtable-planning/components/ParallelSeatStrip.tsx
@@ -11,6 +11,7 @@
  */
 import { CheckIcon, Loader2Icon, XIcon } from "lucide-react";
 
+import { desensitize } from "@/lib/desensitize";
 import { cn } from "@/lib/utils";
 import { getAvatar } from "@/roundtable-planning/components/Avatars";
 import type { ParallelSeatView } from "@/roundtable-planning/hooks/useStep2Orchestration";
@@ -42,7 +43,7 @@ export function ParallelSeatStrip({ batch, selectedSeats, onToggleSeat }: Parall
               key={seat.agentId}
               type="button"
               onClick={() => onToggleSeat(seat.agentId)}
-              title={`${seat.displayName}${isSelected ? "（主区显示中）" : "（点击在主区显示）"}`}
+              title={`${desensitize(seat.displayName)}${isSelected ? "（主区显示中）" : "（点击在主区显示）"}`}
               className={cn(
                 "relative shrink-0 rounded-full transition",
                 isSelected ? "ring-2 ring-primary ring-offset-1" : "ring-1 ring-transparent hover:ring-slate-300",
diff --git a/frontend-web/src/roundtable-planning/components/RecommendAgentsDialog.tsx b/frontend-web/src/roundtable-planning/components/RecommendAgentsDialog.tsx
index 61bb1de5..6619cc2f 100644
--- a/frontend-web/src/roundtable-planning/components/RecommendAgentsDialog.tsx
+++ b/frontend-web/src/roundtable-planning/components/RecommendAgentsDialog.tsx
@@ -23,6 +23,7 @@ import {
 import { Button } from "@/components/ui/button";
 import { listAgents, type Agent } from "@/core/agents";
 import { useTheme } from "@/contexts/ThemeContext";
+import { desensitize } from "@/lib/desensitize";
 import { cn } from "@/lib/utils";
 
 import {
@@ -1122,7 +1123,7 @@ function AgentCard({
       </span>
       <div className="flex-1 min-w-0">
         <div className="flex items-center gap-2 flex-wrap">
-          <div className="font-medium text-sm truncate">{candidate.name || candidate.agent_id}</div>
+          <div className="font-medium text-sm truncate">{desensitize(candidate.name || candidate.agent_id)}</div>
           {recommended && (
             <span className="text-[10px] px-1.5 py-0.5 rounded bg-purple-500/10 text-purple-700 dark:text-purple-300 font-medium">
               推荐
@@ -1131,12 +1132,12 @@ function AgentCard({
         </div>
         {candidate.description && (
           <div className="text-xs text-muted-foreground mt-0.5 line-clamp-2">
-            {candidate.description}
+            {desensitize(candidate.description)}
           </div>
         )}
         {reason ? (
           <div className="text-xs text-primary/80 mt-1 italic">
-            <span className="font-medium not-italic">推荐理由:</span> {reason}
+            <span className="font-medium not-italic">推荐理由:</span> {desensitize(reason)}
           </div>
         ) : streaming ? (
           <div className="text-xs text-muted-foreground mt-1 italic">正在生成推荐理由...</div>
diff --git a/frontend-web/src/roundtable-planning/components/Step1Panel.tsx b/frontend-web/src/roundtable-planning/components/Step1Panel.tsx
index 695bae14..202dbaee 100644
--- a/frontend-web/src/roundtable-planning/components/Step1Panel.tsx
+++ b/frontend-web/src/roundtable-planning/components/Step1Panel.tsx
@@ -56,6 +56,7 @@ import { ScrollToBottomButton } from "@/roundtable-planning/components/ScrollToB
 import { useAutoScroll } from "@/roundtable-planning/hooks/useAutoScroll";
 import { useDialogueSpeedMetrics } from "@/roundtable-planning/hooks/useDialogueSpeedMetrics";
 import type { InteractiveMessage } from "@/roundtable-planning/hooks/useStep1Intent";
+import { useDesensitizeSubtree } from "@/roundtable-planning/lib/use-desensitize-subtree";
 import { splitInlineReasoning } from "@/roundtable-planning/lib/reasoning";
 import { splitIntentReadyBlock } from "@/roundtable-planning/lib/roundtable-constants";
 import type {
@@ -149,10 +150,13 @@ export function Step1Panel(props: Step1PanelProps) {
     setAttachedFiles([]);
   };
 
+  // 顶部标题栏整片脱敏（消息列表正文在下方，不在此容器内，保持原文）。
+  const headerDzRef = useDesensitizeSubtree<HTMLDivElement>();
+
   return (
     <div className="flex-1 flex flex-col min-h-0">
       {/* Dialogue Header */}
-      <div className="shrink-0 px-6 py-4.5 border-b border-slate-100 flex items-center justify-between bg-slate-50/50">
+      <div ref={headerDzRef} className="shrink-0 px-6 py-4.5 border-b border-slate-100 flex items-center justify-between bg-slate-50/50">
         <div className="flex items-center gap-2">
           <MessageSquare className="w-4.5 h-4.5 text-primary" strokeWidth={2.5} />
           <h3 className="text-sm font-extrabold text-slate-900">
diff --git a/frontend-web/src/roundtable-planning/components/Step2Panel.tsx b/frontend-web/src/roundtable-planning/components/Step2Panel.tsx
index 7835841a..6d3e2d90 100644
--- a/frontend-web/src/roundtable-planning/components/Step2Panel.tsx
+++ b/frontend-web/src/roundtable-planning/components/Step2Panel.tsx
@@ -81,6 +81,8 @@ import {
   getAvatarTypeFor,
   type RoleEntry,
 } from "@/roundtable-planning/lib/roundtable-constants";
+import { desensitize } from "@/lib/desensitize";
+import { useDesensitizeSubtree } from "@/roundtable-planning/lib/use-desensitize-subtree";
 import { splitInlineReasoning } from "@/roundtable-planning/lib/reasoning";
 import type { ClarificationPayload } from "@/roundtable-planning/api/multi-agent";
 
@@ -405,10 +407,13 @@ export function Step2Panel(props: Step2PanelProps) {
     isOrchestrating ||
     step2RoundtableDialogues.some((dialogue) => dialogue.streaming === true);
 
+  // 顶部标题栏整片脱敏（消息列表正文不在此容器，保持原文）。
+  const headerDzRef = useDesensitizeSubtree<HTMLDivElement>();
+
   const chatBody = (
     <>
       {/* Headline */}
-      <div className="shrink-0 px-6 py-4.5 border-b border-slate-100 flex items-center justify-between bg-white shadow-sm">
+      <div ref={headerDzRef} className="shrink-0 px-6 py-4.5 border-b border-slate-100 flex items-center justify-between bg-white shadow-sm">
         <div className="flex items-center gap-2">
           <Cpu className="w-5 h-5 text-primary" strokeWidth={2.5} />
           <div>
@@ -584,7 +589,7 @@ export function Step2Panel(props: Step2PanelProps) {
                       />
                     ) : (
                       <div className="py-6 text-center text-xs text-slate-400">
-                        {seat.displayName} · 等待输出…
+                        {desensitize(seat.displayName)} · 等待输出…
                       </div>
                     )}
                   </div>
@@ -599,7 +604,7 @@ export function Step2Panel(props: Step2PanelProps) {
                   <span className="roundtable-chat-meta font-medium text-slate-600">
                     仅显示{" "}
                     <span className="font-bold text-primary">
-                      {seatFilterLabel(seatFilter)}
+                      {desensitize(seatFilterLabel(seatFilter))}
                     </span>{" "}
                     的发言
                   </span>
@@ -615,7 +620,7 @@ export function Step2Panel(props: Step2PanelProps) {
               )}
               {seatFilter && visibleDialogues.length === 0 ? (
                 <div className="py-10 text-center text-xs text-slate-400">
-                  没有匹配「{seatFilterLabel(seatFilter)}」的消息
+                  没有匹配「{desensitize(seatFilterLabel(seatFilter))}」的消息
                 </div>
               ) : (
                 visibleDialogues.map((msg) => (
@@ -878,7 +883,7 @@ export function Step2Message({
       title={
         isAvatarSelected
           ? "取消筛选,显示全部发言"
-          : `只看 ${msg.sender} 的发言`
+          : `只看 ${desensitize(msg.sender)} 的发言`
       }
       className={cn(
         "shrink-0 cursor-pointer rounded-full transition hover:opacity-80 focus:outline-none focus-visible:ring-2 focus-visible:ring-primary/60",
diff --git a/frontend-web/src/roundtable-planning/components/Step2SandboxLayout.tsx b/frontend-web/src/roundtable-planning/components/Step2SandboxLayout.tsx
index 59585fa9..6c6bec18 100644
--- a/frontend-web/src/roundtable-planning/components/Step2SandboxLayout.tsx
+++ b/frontend-web/src/roundtable-planning/components/Step2SandboxLayout.tsx
@@ -116,6 +116,8 @@ export function Step2SandboxLayout({
           threadId={previewThreadId}
           onClose={artifactsFill && onArtifactsClose ? handleArtifactsClose : undefined}
           allowVirtualMarkdownEdit
+          // 会商「消息列表沙箱」按需求不脱敏产物内容（用户要拿原始文件）。
+          disableDesensitize
         />
       ) : (
         <div className="relative flex size-full justify-center">
diff --git a/frontend-web/src/roundtable-planning/components/Step3DashboardMode.tsx b/frontend-web/src/roundtable-planning/components/Step3DashboardMode.tsx
index 9fce3033..09fd6470 100644
--- a/frontend-web/src/roundtable-planning/components/Step3DashboardMode.tsx
+++ b/frontend-web/src/roundtable-planning/components/Step3DashboardMode.tsx
@@ -35,6 +35,7 @@ import {
 } from "lucide-react";
 
 import { useLocalSettings } from "@/core/settings";
+import { desensitizeDeep } from "@/lib/desensitize";
 import type { DraftStep3Snapshot } from "@/roundtable-planning/api/drafts";
 import { DashboardApp } from "@/roundtable-planning/dashboard/DashboardApp";
 import { adaptReportData } from "@/roundtable-planning/dashboard/data-context";
@@ -101,10 +102,12 @@ export function Step3DashboardMode({
   } = dashboard;
 
   // 真实数据（report-json）：优先本轮实时生成，其次草稿快照。应用渲染/网页渲染共用。
-  const report = useMemo(
-    () => dashboard.data ?? parseReportJson(savedReport?.dashboardJson ?? null),
-    [dashboard.data, savedReport?.dashboardJson],
-  );
+  // 整体深度脱敏后再渲染——大屏（DashboardApp）与自包含 HTML（iframe/新窗口/分享/下载）
+  // 都从这份 report 派生，故在此一处脱敏即覆盖大屏的所有展示文本（会商页非沙箱内容需脱敏）。
+  const report = useMemo(() => {
+    const raw = dashboard.data ?? parseReportJson(savedReport?.dashboardJson ?? null);
+    return raw ? desensitizeDeep(raw) : raw;
+  }, [dashboard.data, savedReport?.dashboardJson]);
 
   // 应用渲染用的大屏数据（不传则 DashboardApp 回落 DEMO）。
   const dashboardData = useMemo(
diff --git a/frontend-web/src/roundtable-planning/components/Step3ReportMode.tsx b/frontend-web/src/roundtable-planning/components/Step3ReportMode.tsx
index c9a49fed..f743d794 100644
--- a/frontend-web/src/roundtable-planning/components/Step3ReportMode.tsx
+++ b/frontend-web/src/roundtable-planning/components/Step3ReportMode.tsx
@@ -44,6 +44,7 @@ import { ConsensusDocxDownloadButton } from "@/roundtable-planning/components/Co
 import { ScrollToBottomButton } from "@/roundtable-planning/components/ScrollToBottomButton";
 import { Step2Message } from "@/roundtable-planning/components/Step2Panel";
 import { Step2SandboxLayout } from "@/roundtable-planning/components/Step2SandboxLayout";
+import { desensitize } from "@/lib/desensitize";
 import { markdownComponents } from "@/roundtable-planning/components/markdown-components";
 import type { DraftStep3Snapshot } from "@/roundtable-planning/api/drafts";
 import { Step3DashboardMode } from "@/roundtable-planning/components/Step3DashboardMode";
@@ -407,7 +408,7 @@ export function Step3ReportMode({
         >
           {consensusText ? (
             <ReactMarkdown remarkPlugins={[remarkGfm]} components={markdownComponents}>
-              {consensusText}
+              {desensitize(consensusText)}
             </ReactMarkdown>
           ) : (
             <p className="text-xs text-slate-400 italic">暂无总控结论。</p>
@@ -773,14 +774,14 @@ function SeatDeliveriesView({ seatDeliveries }: { seatDeliveries: SeatDelivery[]
             >
               <div className="flex items-center gap-2 min-w-0">
                 <ListOrdered className="w-3.5 h-3.5 text-slate-400 shrink-0" />
-                <span className="text-[11.5px] font-bold text-slate-800 truncate">{seat.sender}</span>
+                <span className="text-[11.5px] font-bold text-slate-800 truncate">{desensitize(seat.sender)}</span>
               </div>
               <span className="text-[10px] text-slate-400 font-mono shrink-0">{seat.content.length} 字</span>
             </button>
             {open && (
               <div className="px-4 py-3 text-xs text-slate-700 leading-relaxed border-t border-slate-100 bg-white">
                 <ReactMarkdown remarkPlugins={[remarkGfm]} components={markdownComponents}>
-                  {seat.content || "(空)"}
+                  {desensitize(seat.content || "(空)")}
                 </ReactMarkdown>
               </div>
             )}
diff --git a/frontend-web/src/roundtable-planning/components/Step3SummaryMode.tsx b/frontend-web/src/roundtable-planning/components/Step3SummaryMode.tsx
index a8dce3a7..4a317329 100644
--- a/frontend-web/src/roundtable-planning/components/Step3SummaryMode.tsx
+++ b/frontend-web/src/roundtable-planning/components/Step3SummaryMode.tsx
@@ -51,6 +51,7 @@ import {
 import { ScrollToBottomButton } from "@/roundtable-planning/components/ScrollToBottomButton";
 import { Step2Message } from "@/roundtable-planning/components/Step2Panel";
 import { Step2SandboxLayout } from "@/roundtable-planning/components/Step2SandboxLayout";
+import { desensitize } from "@/lib/desensitize";
 import { markdownComponents } from "@/roundtable-planning/components/markdown-components";
 import type { ModelOption } from "@/roundtable-planning/components/Step3ReportMode";
 import type { DraftStep3SummarySnapshot } from "@/roundtable-planning/api/drafts";
@@ -449,7 +450,7 @@ export function Step3SummaryMode({
         >
           {consensusText ? (
             <ReactMarkdown remarkPlugins={[remarkGfm]} components={markdownComponents}>
-              {consensusText}
+              {desensitize(consensusText)}
             </ReactMarkdown>
           ) : (
             <p className="text-xs text-slate-400 italic">暂无总控结论。</p>
@@ -846,14 +847,14 @@ function SeatDeliveriesView({ seatDeliveries }: { seatDeliveries: SeatDelivery[]
             >
               <div className="flex items-center gap-2 min-w-0">
                 <FileText className="w-3.5 h-3.5 text-slate-400 shrink-0" />
-                <span className="text-[11.5px] font-bold text-slate-800 truncate">{seat.sender}</span>
+                <span className="text-[11.5px] font-bold text-slate-800 truncate">{desensitize(seat.sender)}</span>
               </div>
               <span className="text-[10px] text-slate-400 font-mono shrink-0">{seat.content.length} 字</span>
             </button>
             {open && (
               <div className="px-4 py-3 text-xs text-slate-700 leading-relaxed border-t border-slate-100 bg-white">
                 <ReactMarkdown remarkPlugins={[remarkGfm]} components={markdownComponents}>
-                  {seat.content || "(空)"}
+                  {desensitize(seat.content || "(空)")}
                 </ReactMarkdown>
               </div>
             )}
diff --git a/frontend-web/src/roundtable-planning/lib/step-display.tsx b/frontend-web/src/roundtable-planning/lib/step-display.tsx
index e7887c64..36ffa9dd 100644
--- a/frontend-web/src/roundtable-planning/lib/step-display.tsx
+++ b/frontend-web/src/roundtable-planning/lib/step-display.tsx
@@ -24,6 +24,7 @@ import {
 
 import { ChainOfThoughtSearchResult } from "@/components/ai-elements/chain-of-thought";
 import { CodeBlock } from "@/components/ai-elements/code-block";
+import { desensitize } from "@/lib/desensitize";
 import { MarkdownContent } from "@/roundtable-planning/components/MessageBubble";
 
 /**
@@ -52,7 +53,7 @@ export function renderStepArgs(
       return (
         <>
           {displayName && (
-            <ChainOfThoughtSearchResult>{displayName}</ChainOfThoughtSearchResult>
+            <ChainOfThoughtSearchResult>{desensitize(displayName)}</ChainOfThoughtSearchResult>
           )}
           {task && (
             <div className="roundtable-chat-message mt-1 max-h-64 overflow-y-auto rounded-md border border-slate-200 bg-white px-2.5 py-1 text-slate-600">
diff --git a/frontend-web/src/roundtable-planning/lib/use-desensitize-subtree.ts b/frontend-web/src/roundtable-planning/lib/use-desensitize-subtree.ts
new file mode 100644
index 00000000..c5185621
--- /dev/null
+++ b/frontend-web/src/roundtable-planning/lib/use-desensitize-subtree.ts
@@ -0,0 +1,67 @@
+/**
+ * 区域级脱敏 Hook —— 把某个容器（DOM 子树）里的**所有文本节点**做脱敏，
+ * 用于「左/右侧面板 + 顶部标题」这类需要整片脱敏的区域：静态标题、动态内容、
+ * 嵌套子组件文本一并覆盖，无需逐串包 `desensitize()`（那样易漏、难维护）。
+ *
+ * 用法：`const ref = useDesensitizeSubtree<HTMLDivElement>(); <section ref={ref}>…</section>`。
+ *
+ * 设计要点：
+ *  - 只读改文本节点（display-only），不碰 `<input>/<textarea>` 的 value、不改 React 状态/回传值；
+ *  - `desensitize` 幂等（替换词不再命中词表），MutationObserver 重入安全；改写前先 disconnect 防自触发循环；
+ *  - 想在区域内**保留原文**（如内嵌沙箱/消息列表），给该子树根加 `data-no-desensitize` 属性即可跳过。
+ *
+ * 注意：**不要**用它包「消息列表正文」或「沙箱产物预览」——这两处按需求保持原文，
+ * 应作为独立容器、不挂本 Hook（智能体名称的脱敏单独在 MessageBubble 的 senderName 处理）。
+ */
+import { useEffect, useRef } from "react";
+
+import { desensitize } from "@/lib/desensitize";
+
+const _SKIP_TAGS = new Set(["SCRIPT", "STYLE", "TEXTAREA", "INPUT", "IFRAME", "CODE", "PRE"]);
+
+export function useDesensitizeSubtree<T extends HTMLElement>() {
+  const ref = useRef<T | null>(null);
+
+  useEffect(() => {
+    const root = ref.current;
+    if (!root || typeof MutationObserver === "undefined") return;
+
+    let observer: MutationObserver | null = null;
+
+    const apply = () => {
+      const walker = document.createTreeWalker(root, NodeFilter.SHOW_TEXT, {
+        acceptNode(node) {
+          const parent = node.parentElement;
+          if (!parent) return NodeFilter.FILTER_REJECT;
+          if (_SKIP_TAGS.has(parent.tagName)) return NodeFilter.FILTER_REJECT;
+          // 显式标记 data-no-desensitize 的子树保留原文（如内嵌沙箱）。
+          if (parent.closest("[data-no-desensitize]")) return NodeFilter.FILTER_REJECT;
+          return NodeFilter.FILTER_ACCEPT;
+        },
+      });
+      const edits: Array<[Text, string]> = [];
+      let current = walker.nextNode();
+      while (current) {
+        const textNode = current as Text;
+        const original = textNode.nodeValue ?? "";
+        if (original.trim()) {
+          const masked = desensitize(original);
+          if (masked !== original) edits.push([textNode, masked]);
+        }
+        current = walker.nextNode();
+      }
+      if (edits.length === 0) return;
+      // 改写期间停掉观察，避免自身改动再次触发，改完恢复。
+      observer?.disconnect();
+      for (const [textNode, masked] of edits) textNode.nodeValue = masked;
+      observer?.observe(root, { subtree: true, childList: true, characterData: true });
+    };
+
+    apply();
+    observer = new MutationObserver(() => apply());
+    observer.observe(root, { subtree: true, childList: true, characterData: true });
+    return () => observer?.disconnect();
+  });
+
+  return ref;
+}
diff --git a/frontend-web/src/roundtable-planning/pages/BusinessChainEditorPage.tsx b/frontend-web/src/roundtable-planning/pages/BusinessChainEditorPage.tsx
index 4eabfe46..dfa096f0 100644
--- a/frontend-web/src/roundtable-planning/pages/BusinessChainEditorPage.tsx
+++ b/frontend-web/src/roundtable-planning/pages/BusinessChainEditorPage.tsx
@@ -63,6 +63,7 @@ import {
 import { CSS } from "@dnd-kit/utilities";
 
 import { cn } from "@/lib/utils";
+import { useDesensitizeSubtree } from "@/roundtable-planning/lib/use-desensitize-subtree";
 import { TagSearchBar } from "@/components/workspace/tags/tag-search-bar";
 import { listAgents } from "@/core/agents";
 
@@ -134,6 +135,8 @@ type ActiveDrag =
 export default function BusinessChainEditorPage() {
   const navigate = useNavigate();
   const [searchParams] = useSearchParams();
+  // 整片脱敏：链条配置页展示文本（链条名/描述/席位名/标题）。可编辑的 input/textarea 不脱敏（Hook 跳过），保证编辑用原文。
+  const dzRef = useDesensitizeSubtree<HTMLDivElement>();
 
   // ─── 数据 ────────────────────────────────────────────────────────────
   const [pool, setPool] = useState<RecommendCandidate[]>([]);
@@ -804,7 +807,7 @@ export default function BusinessChainEditorPage() {
   );
 
   return (
-    <div className="flex h-full min-h-0 flex-col bg-muted/40 text-foreground">
+    <div ref={dzRef} className="flex h-full min-h-0 flex-col bg-muted/40 text-foreground">
       {/* 顶栏 */}
       <header className="flex shrink-0 flex-wrap items-center gap-3 border-b border-border bg-card px-6 py-3">
         <button
diff --git a/frontend-web/src/roundtable-planning/pages/BusinessChainListPage.tsx b/frontend-web/src/roundtable-planning/pages/BusinessChainListPage.tsx
index 2a7e1882..aab9004e 100644
--- a/frontend-web/src/roundtable-planning/pages/BusinessChainListPage.tsx
+++ b/frontend-web/src/roundtable-planning/pages/BusinessChainListPage.tsx
@@ -62,6 +62,7 @@ import {
 } from "@/components/ui/dialog";
 import { useRouteNavigate } from "@/core/route-base";
 import { cn } from "@/lib/utils";
+import { useDesensitizeSubtree } from "@/roundtable-planning/lib/use-desensitize-subtree";
 
 import { deleteChain, listChains, setChainPublic, type ChainScope } from "../api/chains";
 import { chainTagClasses, type BusinessChain } from "../lib/business-chain";
@@ -79,6 +80,8 @@ const BRAND =
 
 export default function BusinessChainListPage() {
   const navigate = useRouteNavigate();
+  // 整片脱敏：链条列表页展示文本（链条名/描述/席位名/标题）。
+  const dzRef = useDesensitizeSubtree<HTMLDivElement>();
 
   // 进入编辑器：带 chainId = 打开并切到该链条；不带 = 空白新建。
   // 路径写 workspace base，`useRouteNavigate` 会按当前路由 base 改写。
@@ -233,7 +236,7 @@ export default function BusinessChainListPage() {
   }, [chains, search, sortBy]);
 
   return (
-    <div className="flex h-full min-h-0 w-full flex-col bg-zinc-50/60 font-sans text-zinc-900 antialiased dark:bg-[#0C1836] dark:text-[#C5CFE0]">
+    <div ref={dzRef} className="flex h-full min-h-0 w-full flex-col bg-zinc-50/60 font-sans text-zinc-900 antialiased dark:bg-[#0C1836] dark:text-[#C5CFE0]">
       {/* 顶栏：标题 + 描述 + 新建链条 */}
       <header className="shrink-0 border-b border-zinc-200/80 bg-white/85 px-6 py-4 backdrop-blur-md dark:border-[#1A2F5A] dark:bg-[#122347]/80">
         <div className="mx-auto flex w-full max-w-7xl flex-col gap-4 md:flex-row md:items-center md:justify-between">
diff --git a/frontend-web/src/roundtable-planning/pages/RoundtablePlanningPage.tsx b/frontend-web/src/roundtable-planning/pages/RoundtablePlanningPage.tsx
index 7ad4ee81..beabc8f1 100644
--- a/frontend-web/src/roundtable-planning/pages/RoundtablePlanningPage.tsx
+++ b/frontend-web/src/roundtable-planning/pages/RoundtablePlanningPage.tsx
@@ -38,6 +38,7 @@ import {
   Cpu,
   Download,
   History,
+  Layers,
   ListOrdered,
   Loader2,
   Pencil,
@@ -54,7 +55,9 @@ import {
 import { useTheme } from "@/contexts/ThemeContext";
 import { useArtifacts } from "@/components/workspace/artifacts";
 import { useSidebarSafe } from "@/components/ui/sidebar";
+import { desensitize } from "@/lib/desensitize";
 import { cn } from "@/lib/utils";
+import { useDesensitizeSubtree } from "@/roundtable-planning/lib/use-desensitize-subtree";
 import { useModels } from "@/core/models/hooks";
 
 import { getAvatar, Man1Avatar, RobotAvatar } from "@/roundtable-planning/components/Avatars";
@@ -98,6 +101,7 @@ import {
 } from "@/roundtable-planning/lib/seat-filter";
 import {
   buildChainSeatSelection,
+  type BusinessChain,
   type SeatSelectionResult,
 } from "@/roundtable-planning/lib/business-chain";
 import { DEFAULT_SEAT_MODE, type SeatMode } from "@/roundtable-planning/lib/seat-mode";
@@ -110,7 +114,8 @@ import {
   startJobKeepalive,
 } from "@/roundtable-planning/api/roundtable-jobs";
 import { useRoundtableJobNotifier } from "@/roundtable-planning/components/RoundtableJobNotifier";
-import { getChain } from "@/roundtable-planning/api/chains";
+import { getChain, listChains } from "@/roundtable-planning/api/chains";
+import { useRoundtableDrafts } from "@/roundtable-planning/hooks/useRoundtableDrafts";
 import { filterRecommendCandidates } from "@/roundtable-planning/api/recommend";
 import { listAgents } from "@/core/agents";
 import { listBusinessMappings } from "@/strategy-components/api/business-mapping";
@@ -776,6 +781,34 @@ export default function RoundtablePlanningPage() {
     [buildCurrentStep2Snapshot, drafts, mergeActiveRun, step2, step3Report, triggerToast],
   );
 
+  // 需求 0616 #2：Step1 左栏点击某条业务链 → 仅**勾选**该链（不直接进 Step 2）。
+  // 复用 autoChainResult 机制：用户在第一步完成意图澄清后，点「进入多智能体研讨」时
+  // onEnterRoundtable 的 `if (autoChainResult)` 分支会据此跳过选链弹窗、按该链进研讨。
+  // 再次点击同一条链 = 取消勾选。
+  const handleSelectChainFromSidebar = useCallback(
+    async (chain: BusinessChain) => {
+      if (autoChainResult?.chain?.id === chain.id) {
+        setAutoChainResult(null);
+        triggerToast(`已取消业务链「${chain.title}」`);
+        return;
+      }
+      try {
+        const agents = await listAgents().catch(() => []);
+        const candidates = filterRecommendCandidates(agents);
+        const result = buildChainSeatSelection(chain, candidates);
+        if (!result) {
+          triggerToast("⚠️ 该业务链条暂无法选用（匹配到的席位不足）");
+          return;
+        }
+        setAutoChainResult(result);
+        triggerToast(`✅ 已选择业务链「${chain.title}」，完成第一步澄清后点「进入多智能体研讨」`);
+      } catch {
+        triggerToast("❌ 选择业务链失败，请重试");
+      }
+    },
+    [autoChainResult, triggerToast],
+  );
+
   // ─── 会商自动化串联（模块 C） ──────────────────────────────────
   // 消费 prefill 后清掉 history state，避免刷新 / 后退重放自动开跑与自动选链。
   useEffect(() => {
@@ -1147,6 +1180,8 @@ export default function RoundtablePlanningPage() {
   // 真正建好的窗口里，用它让历史记录显示「挂起中」（作业建好后转 queued/running，由后端状态接管）。
   // resetToStep1 不卸载本页，故该 state 跨「跳转」存活。
   const [pendingSuspendIds, setPendingSuspendIds] = useState<Set<string>>(() => new Set());
+  // 整片脱敏：顶部页头（步骤导航 / 历史记录 / 标题等）的所有文本。
+  const headerDzRef = useDesensitizeSubtree<HTMLElement>();
   const markPendingSuspend = useCallback((draftId: string) => {
     setPendingSuspendIds((prev) => {
       const next = new Set(prev);
@@ -1446,7 +1481,7 @@ export default function RoundtablePlanningPage() {
       )}
 
       {/* HEADER */}
-      <header className="relative z-[40] flex shrink-0 items-center justify-between gap-4 border-b border-slate-100 bg-white px-6 py-3 shadow-[0_1px_3px_rgba(0,0,0,0.02)] dark:border-border dark:bg-card dark:shadow-none">
+      <header ref={headerDzRef} className="relative z-[40] flex shrink-0 items-center justify-between gap-4 border-b border-slate-100 bg-white px-6 py-3 shadow-[0_1px_3px_rgba(0,0,0,0.02)] dark:border-border dark:bg-card dark:shadow-none">
         <StepNavigation
           currentStep={currentStep}
           intentReady={step1.intentReady}
@@ -1557,6 +1592,9 @@ export default function RoundtablePlanningPage() {
           onSelectStep2Run={handleSelectStep2Run}
           seatFilter={seatFilter}
           setSeatFilter={setSeatFilter}
+          selectedChainId={autoChainResult?.chain?.id ?? null}
+          onSelectChain={handleSelectChainFromSidebar}
+          onOpenDraft={drafts.handleLoadDraft}
         />
 
         {/* MIDDLE VIEWPORT */}
@@ -2056,6 +2094,12 @@ interface LeftSidebarProps {
   /** 第二步「按席位筛选会商记录」—— 点角色卡片筛选,与中间消息头像共用。 */
   seatFilter: SeatFilter | null;
   setSeatFilter: React.Dispatch<React.SetStateAction<SeatFilter | null>>;
+  /** 需求 0616 #2：当前已勾选的业务链 id（高亮用），来自 autoChainResult.chain.id。 */
+  selectedChainId: string | null;
+  /** 需求 0616 #2：Step1 落地左栏「业务链」点击 → 勾选该链（不直接进研讨）。 */
+  onSelectChain: (chain: BusinessChain) => void;
+  /** 需求 0616 #2：Step1 落地左栏「分析记录」点击 → 打开历史会商分析（加载草稿）。 */
+  onOpenDraft: (draftId: string) => void;
 }
 
 function Step2SandboxSidebarSync({
@@ -2097,7 +2141,13 @@ function LeftSidebar({
   onSelectStep2Run,
   seatFilter,
   setSeatFilter,
+  selectedChainId,
+  onSelectChain,
+  onOpenDraft,
 }: LeftSidebarProps) {
+  // 整片脱敏：左侧面板（任务概览/参与角色/人工参与等）的所有文本——标题、动态内容、
+  // 嵌套子组件文本一并脱敏。消息列表与沙箱是另外的容器，不挂此 Hook，保持原文。
+  const dzRef = useDesensitizeSubtree<HTMLElement>();
   const isStep2 = currentStep === 2;
   // 仅第二步可按席位筛选会商记录;点中已选则取消(回到显示全部)。
   const toggleSeatFilter = (filter: SeatFilter) =>
@@ -2109,6 +2159,7 @@ function LeftSidebar({
 
   return (
     <section
+      ref={dzRef}
       id="left-sidebar"
       className={cn(
         "shrink-0 flex flex-col gap-3 overflow-y-auto pr-1 transition-[width] duration-200 ease-out",
@@ -2223,28 +2274,147 @@ function LeftSidebar({
           </>
         )
       ) : (
-        <>
-          <TaskOverviewCard currentStep={currentStep} />
-
-          <div className="bg-white border border-slate-200/50 rounded-2xl p-4 shadow-[0_2px_8px_rgba(0,0,0,0.01)] flex-1 flex flex-col min-h-[300px]">
-            <div className="flex items-center gap-2 border-b border-slate-100 pb-2.5 mb-3">
-              <Users className="w-4 h-4 text-primary" strokeWidth={2.5} />
-              <h2 className="text-xs font-bold text-slate-900 uppercase tracking-widest">参与角色</h2>
-            </div>
-            <div className="space-y-2.5 overflow-y-auto flex-1 pr-1 custom-scrollbar">
-              {roles.map((role) => (
-                <RoleRow key={role.id} role={role} agentNameToRole={agentNameToRole} />
-              ))}
-            </div>
-          </div>
-
-          <HumanParticipationCard />
-        </>
+        // Step 1（落地，未开始研讨）：左栏改为业务链 + 分析记录入口（需求 0616 #2）。
+        <RoundtableEntryRail
+          selectedChainId={selectedChainId}
+          onSelectChain={onSelectChain}
+          onOpenDraft={onOpenDraft}
+        />
       )}
     </section>
   );
 }
 
+/**
+ * 需求 0616 #2：多智能体会商 Step1（落地）左栏入口栏。
+ *  - 业务链：仅公共链条（``listChains("public")``）前 3 个；点击以该链席位启动会商。
+ *  - 分析记录：圆桌历史草稿（``useRoundtableDrafts``）；点击打开对应历史会商分析。
+ * 进入研讨（Step 2/3）后左栏恢复为「参与角色 / 任务概览」工作面板。
+ */
+function RoundtableEntryRail({
+  selectedChainId,
+  onSelectChain,
+  onOpenDraft,
+}: {
+  selectedChainId: string | null;
+  onSelectChain: (chain: BusinessChain) => void;
+  onOpenDraft: (draftId: string) => void;
+}) {
+  const [chains, setChains] = useState<BusinessChain[]>([]);
+  const { drafts: records, refresh } = useRoundtableDrafts();
+
+  useEffect(() => {
+    let cancelled = false;
+    void listChains("public")
+      .then((all) => {
+        if (!cancelled) setChains(all.slice(0, 3));
+      })
+      .catch(() => {
+        /* 业务链非关键，静默忽略 */
+      });
+    void refresh();
+    return () => {
+      cancelled = true;
+    };
+  }, [refresh]);
+
+  const formatTime = (value: string) => {
+    const date = new Date(value);
+    if (Number.isNaN(date.getTime())) return "";
+    return date.toLocaleString("zh-CN", {
+      month: "2-digit",
+      day: "2-digit",
+      hour: "2-digit",
+      minute: "2-digit",
+      hour12: false,
+    });
+  };
+
+  return (
+    <>
+      {/* 业务链（公共，前 3 个） */}
+      <div className="bg-white border border-slate-200/50 rounded-2xl p-4 shadow-[0_2px_8px_rgba(0,0,0,0.01)]">
+        <div className="flex items-center gap-2 border-b border-slate-100 pb-2.5 mb-3">
+          <Layers className="w-4 h-4 text-primary" strokeWidth={2.5} />
+          <h2 className="text-xs font-bold text-slate-900 uppercase tracking-widest">业务链</h2>
+        </div>
+        {chains.length === 0 ? (
+          <p className="text-xs text-slate-400 py-1.5">暂无公共业务链</p>
+        ) : (
+          <div className="space-y-2">
+            {chains.map((chain) => {
+              const selected = selectedChainId === chain.id;
+              return (
+                <button
+                  key={chain.id}
+                  type="button"
+                  onClick={() => onSelectChain(chain)}
+                  title={selected ? "再次点击取消勾选" : `勾选业务链「${chain.title}」`}
+                  aria-pressed={selected}
+                  className={cn(
+                    "w-full rounded-xl border p-2.5 text-left transition-colors",
+                    selected
+                      ? "border-primary/40 bg-primary/5"
+                      : "border-slate-100 bg-slate-50/60 hover:border-primary/30 hover:bg-white",
+                  )}
+                >
+                  <div className="flex items-center justify-between gap-2">
+                    <span
+                      className={cn(
+                        "truncate text-xs font-extrabold",
+                        selected ? "text-primary" : "text-slate-800",
+                      )}
+                    >
+                      {chain.title}
+                    </span>
+                    {selected ? (
+                      <Check className="h-3.5 w-3.5 shrink-0 text-primary" />
+                    ) : (
+                      <ChevronRight className="h-3.5 w-3.5 shrink-0 text-slate-400" />
+                    )}
+                  </div>
+                  <div className="mt-1 text-[10px] font-semibold text-slate-400">
+                    {chain.seats.length} 个席位
+                    {selected ? " · 已选" : ""}
+                  </div>
+                </button>
+              );
+            })}
+          </div>
+        )}
+      </div>
+
+      {/* 分析记录（圆桌历史草稿） */}
+      <div className="bg-white border border-slate-200/50 rounded-2xl p-4 shadow-[0_2px_8px_rgba(0,0,0,0.01)] flex-1 flex flex-col min-h-[200px]">
+        <div className="flex items-center gap-2 border-b border-slate-100 pb-2.5 mb-3">
+          <History className="w-4 h-4 text-primary" strokeWidth={2.5} />
+          <h2 className="text-xs font-bold text-slate-900 uppercase tracking-widest">分析记录</h2>
+        </div>
+        {records.length === 0 ? (
+          <p className="text-xs text-slate-400 py-1.5">暂无分析记录</p>
+        ) : (
+          <div className="space-y-2 overflow-y-auto flex-1 pr-1 custom-scrollbar">
+            {records.map((rec) => (
+              <button
+                key={rec.id}
+                type="button"
+                onClick={() => onOpenDraft(rec.id)}
+                title={rec.title}
+                className="w-full rounded-xl border border-slate-100 bg-slate-50/60 p-2.5 text-left transition-colors hover:border-primary/20 hover:bg-white"
+              >
+                <div className="truncate text-xs font-bold text-slate-700">{rec.title}</div>
+                <div className="mt-1 text-[10px] font-semibold text-slate-400">
+                  {formatTime(rec.updatedAt)}
+                </div>
+              </button>
+            ))}
+          </div>
+        )}
+      </div>
+    </>
+  );
+}
+
 function Step2RunSwitcher({
   runs,
   activeRunId,
@@ -2869,14 +3039,14 @@ function RoleAvatarBadge({
   );
 
   if (!clickable) {
-    return <div title={`${role.name} · ${role.status}`}>{inner}</div>;
+    return <div title={`${desensitize(role.name)} · ${role.status}`}>{inner}</div>;
   }
   return (
     <button
       type="button"
       onClick={onSelect}
       aria-pressed={active}
-      title={active ? "取消筛选,显示全部发言" : `只看 ${role.name} 的发言`}
+      title={active ? "取消筛选,显示全部发言" : `只看 ${desensitize(role.name)} 的发言`}
       className="shrink-0 cursor-pointer rounded-full transition hover:opacity-80 focus:outline-none focus-visible:ring-2 focus-visible:ring-primary/60"
     >
       {inner}
@@ -2916,7 +3086,7 @@ function RoleRow({
         clickable
           ? active
             ? "取消筛选,显示全部发言"
-            : `只看 ${role.name} 的发言`
+            : `只看 ${desensitize(role.name)} 的发言`
           : undefined
       }
       onClick={onSelect}
@@ -2946,7 +3116,7 @@ function RoleRow({
         <RoleAvatarBadge role={role} agentNameToRole={agentNameToRole} size="h-10 w-10" />
         <div className="flex min-w-0 flex-col gap-0.5">
           <span className="truncate text-sm font-bold leading-5 tracking-tight text-slate-800">
-            {role.name}
+            {desensitize(role.name)}
           </span>
           {role.type === "agent" && (
             <span className="text-[11px] text-primary font-medium leading-4">AI 智能体</span>
@@ -3016,10 +3186,13 @@ function RightSidebar({
   triggerToast,
   onSaveIntent,
 }: RightSidebarProps) {
+  // 整片脱敏：右侧面板（推荐模型 / 意图摘要 / 澄清记录）所有文本。
+  const dzRef = useDesensitizeSubtree<HTMLElement>();
   // Step 2 侧栏内容已迁至左侧;Step 3 沿用原逻辑不展示右侧栏。
   if (currentStep !== 1) return null;
   return (
     <section
+      ref={dzRef}
       id="right-sidebar"
       className="w-[300px] shrink-0 flex flex-col gap-4 overflow-y-auto pl-1"
     >
diff --git a/frontend-web/src/strategy-components/Header.tsx b/frontend-web/src/strategy-components/Header.tsx
index fae3d1a1..27c617ba 100644
--- a/frontend-web/src/strategy-components/Header.tsx
+++ b/frontend-web/src/strategy-components/Header.tsx
@@ -8,6 +8,7 @@ import { NotificationDialog } from "@/components/workspace/notification-dialog";
 import { getStoredAuth } from "@/core/auth";
 import { isWujieEnv } from "@/core/embed/wujie";
 import { useUnreadCount } from "@/core/notifications";
+import { useMyPosition } from "@/core/positions";
 import {
   getActiveThirdLevelItems,
   type MenuItem,
@@ -131,6 +132,8 @@ const Header = ({ collapsed }: HeaderProps) => {
   const auth = getStoredAuth();
   const isAdmin = auth?.system_role === "admin";
   const visibleMenuItems = useSidebarMenuItems(!!isAdmin);
+  const { data: myPosition } = useMyPosition();
+  const positionName = myPosition?.name?.trim() || "";
 
   // 登录时写入的真实用户名（外部登录存于 localStorage.userInfo.userName）；
   // 缺省回退到 stored auth 的邮箱前缀。
@@ -278,6 +281,11 @@ const Header = ({ collapsed }: HeaderProps) => {
           <p className="text-sm leading-tight font-medium text-white">
             {displayName}
           </p>
+          {positionName && (
+            <p className="text-[11px] leading-tight text-white/70">
+              {positionName}
+            </p>
+          )}
         </div>
         <ChevronDown
           className={cn(
diff --git a/frontend-web/src/strategy-components/components/task-components/task-mock-data.ts b/frontend-web/src/strategy-components/components/task-components/task-mock-data.ts
index dfad039b..da45f0b0 100644
--- a/frontend-web/src/strategy-components/components/task-components/task-mock-data.ts
+++ b/frontend-web/src/strategy-components/components/task-components/task-mock-data.ts
@@ -399,6 +399,102 @@ const MOCK_ROUTES: Array<{ match: string; dataset: MockDataset }> = [
   { match: '/task/cop-tasks', dataset: 'task' },
 ];
 
+// ───────────── 任务详情假数据（cop-task-detail，给登录任务深链用） ─────────────
+//
+// consumer 接口 `{VITE_AIPI_CONSUMER_URL}/taskAnalyseSearch/cop-task-detail` 外网可能不可达，
+// 可用假数据模拟登录深链（?taskId=&goPath=）效果。返回结构与真实接口一致：
+//   `{ code, message, data: { id, overview, taskDirection, taskName, taskContent } }`
+//
+// 启停由 `VITE_MOCK_COP_TASK_DETAIL` 控制（默认真实数据，=true 时走假数据），
+// 见 `utils/request.ts` 的请求拦截器。
+
+/** cop-task-detail 命中路径。 */
+const COP_TASK_DETAIL_ROUTE = '/taskAnalyseSearch/cop-task-detail';
+
+interface CopTaskDetailSeed {
+  taskName: string;
+  taskContent: string;
+  overview: string;
+  taskDirection: string;
+}
+
+/** 任务详情假数据池（按 taskId 取模选一条，让不同 taskId 呈现不同任务）。 */
+const COP_TASK_DETAIL_SEEDS: CopTaskDetailSeed[] = [
+  {
+    taskName: '暑期文旅消费季舆情倾向分析',
+    taskContent:
+      '围绕"暑期文旅消费季"全网讨论，分析公众情绪走向、核心关切与潜在争议点，识别需重点引导的群体，并给出传播策略建议。',
+    overview: '暑期文旅消费季相关讨论的情绪与倾向分析。',
+    taskDirection: '王磊',
+  },
+  {
+    taskName: '居民阶梯电价调整公众情绪研判',
+    taskContent:
+      '针对居民阶梯电价调整方案，梳理公众主要担忧（用电成本、公平性），判断情绪峰值节点与高敏群体，形成分群体沟通要点。',
+    overview: '阶梯电价调整的公众情绪研判。',
+    taskDirection: '李婷',
+  },
+  {
+    taskName: '城市更新片区舆论引导任务',
+    taskContent:
+      '对老城区改造涉及的拆迁、安置等敏感议题进行情绪与诉求拆解，识别居民、商户等不同立场，制定差异化回应策略。',
+    overview: '城市更新片区敏感议题的舆论引导分析。',
+    taskDirection: '张伟',
+  },
+];
+
+/** 从 taskId 反推一个稳定下标（同一 taskId 每次返回同一条）。 */
+function pickSeedIndex(taskId: string): number {
+  const digits = (taskId.match(/\d+/g) ?? []).join('');
+  const n = digits ? Number(digits.slice(-6)) : 0;
+  return Number.isFinite(n) ? n % COP_TASK_DETAIL_SEEDS.length : 0;
+}
+
+/** 模拟 cop-task-detail 的 body。 */
+function buildCopTaskDetailBody(taskId: string) {
+  const seed = COP_TASK_DETAIL_SEEDS[pickSeedIndex(taskId)];
+  return {
+    code: 0,
+    message: 'ok',
+    data: {
+      id: taskId,
+      overview: seed.overview,
+      taskDirection: seed.taskDirection,
+      taskName: seed.taskName,
+      taskContent: seed.taskContent,
+    },
+  };
+}
+
+/** 判断某请求 url 是否是 cop-task-detail。 */
+export function matchCopTaskDetailMock(url?: string): boolean {
+  return !!url && url.includes(COP_TASK_DETAIL_ROUTE);
+}
+
+/** 从 axios config 取 taskId（优先 params，其次 url query）。 */
+function readTaskIdFromConfig(config: { url?: string; params?: Record<string, unknown> }): string {
+  const fromParams = config.params?.taskId;
+  if (fromParams != null && fromParams !== '') return String(fromParams);
+  if (config.url) {
+    const m = config.url.match(/[?&]taskId=([^&]+)/);
+    if (m) return decodeURIComponent(m[1]);
+  }
+  return '';
+}
+
+/** 临时假数据 axios adapter：命中 cop-task-detail 时直接返回 mock body，不发真实请求。 */
+export function copTaskDetailMockAdapter(config: {
+  url?: string;
+  params?: Record<string, unknown>;
+}): Promise<{ data: unknown; status: number; statusText: string; headers: Record<string, unknown>; config: unknown }> {
+  const body = buildCopTaskDetailBody(readTaskIdFromConfig(config));
+  return new Promise((resolve) => {
+    setTimeout(() => {
+      resolve({ data: body, status: 200, statusText: 'OK', headers: {}, config });
+    }, 200);
+  });
+}
+
 /** 判断某个请求 url 是否应走假数据。 */
 export function matchTaskBoardMock(url?: string): MockDataset | null {
   if (!url) return null;
diff --git a/frontend-web/src/strategy-components/lib/light-app.ts b/frontend-web/src/strategy-components/lib/light-app.ts
index ab796bf6..afda0d7c 100644
--- a/frontend-web/src/strategy-components/lib/light-app.ts
+++ b/frontend-web/src/strategy-components/lib/light-app.ts
@@ -9,6 +9,7 @@ import type { LightApp, LightAppRouteParam } from "@/strategy-components/api/lig
 /** iframe 可挂载的父菜单（取自 core/page-layout/sidebar-menu.ts 的现有菜单 id）。 */
 export const MOUNT_OPTIONS: { label: string; value: string }[] = [
   { label: "XK中任务", value: "xk-center-task" },
+  { label: "COH中任务", value: "coh-center-task" },
   { label: "PG中任务", value: "pg-center-task" },
   { label: "TY中任务", value: "ty-center-task" },
   { label: "资源管理", value: "resource-management" },
diff --git a/frontend-web/src/utils/request.ts b/frontend-web/src/utils/request.ts
index 6666aa95..9162dae1 100644
--- a/frontend-web/src/utils/request.ts
+++ b/frontend-web/src/utils/request.ts
@@ -1,5 +1,10 @@
 import axios from 'axios'
-import { matchTaskBoardMock, taskBoardMockAdapter } from '@/strategy-components/components/task-components/task-mock-data'
+import {
+    matchTaskBoardMock,
+    taskBoardMockAdapter,
+    matchCopTaskDetailMock,
+    copTaskDetailMockAdapter,
+} from '@/strategy-components/components/task-components/task-mock-data'
 
 // 任务看板（任务分析 / 三情分析）数据源开关：
 //   VITE_USE_REAL_DATA=true → 走真实接口；其它（含未设置）→ 这两个接口走本地假数据。
@@ -7,6 +12,11 @@ import { matchTaskBoardMock, taskBoardMockAdapter } from '@/strategy-components/
 // 仅本机 .env 默认 false（假数据）。改 env 后需重启 dev server 生效。
 const USE_REAL_DATA = import.meta.env.VITE_USE_REAL_DATA === 'true'
 
+// 任务详情（cop-task-detail，登录深链用）数据源开关：
+//   默认真实数据；仅当 VITE_MOCK_COP_TASK_DETAIL=true 时走本地假数据（consumer 外网不可达时调试用）。
+//   与 USE_REAL_DATA 独立。改 env 后需重启 dev server 生效。
+const MOCK_COP_TASK_DETAIL = import.meta.env.VITE_MOCK_COP_TASK_DETAIL === 'true'
+
 // axios.interceptors
 const service = axios.create({
     // baseURL: '', // url = base url + request url
@@ -43,6 +53,10 @@ service.interceptors.request.use(
         if (!USE_REAL_DATA && matchTaskBoardMock(config.url)) {
             config.adapter = taskBoardMockAdapter as any;
         }
+        // 任务详情（登录深链）：默认真实接口；仅 VITE_MOCK_COP_TASK_DETAIL=true 时走本地假数据。
+        if (MOCK_COP_TASK_DETAIL && matchCopTaskDetailMock(config.url)) {
+            config.adapter = copTaskDetailMockAdapter as any;
+        }
         // let token: any = localStorage.getItem('token');
         config.headers['Authorization'] = `Bearer admin`;
         // if (token) {
diff --git a/offline-backend-20260512/backend/CLAUDE.md b/offline-backend-20260512/backend/CLAUDE.md
index 0e623e64..d75ca744 100644
--- a/offline-backend-20260512/backend/CLAUDE.md
+++ b/offline-backend-20260512/backend/CLAUDE.md
@@ -1,759 +1,899 @@
-# CLAUDE.md
-??
-This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
-
-## Project Overview
-
-DeerFlow is a LangGraph-based AI super agent system with a full-stack architecture. The backend provides a "super agent" with sandbox execution, persistent memory, subagent delegation, and extensible tool integration - all operating in per-thread isolated environments.
-
-**Architecture**:
-- **Gateway API** (port 8001): REST API plus embedded LangGraph-compatible agent runtime
-- **Frontend** (port 3000): Next.js web interface
-- **Nginx** (port 2026): Unified reverse proxy entry point
-- **Provisioner** (port 8002, optional in Docker dev): Started only when sandbox is configured for provisioner/Kubernetes mode
-
-**Runtime**:
-- `make dev`, Docker dev, and production all run the agent runtime in Gateway via `RunManager` + `run_agent()` + `StreamBridge` (`packages/harness/deerflow/runtime/`). Nginx exposes that runtime at `/api/langgraph/*` and rewrites it to Gateway's native `/api/*` routers.
-
-**Project Structure**:
-```
-deer-flow/
-??? Makefile                    # Root commands (check, install, dev, stop)
-??? config.yaml                 # Main application configuration
-??? extensions_config.json      # MCP servers and skills configuration
-??? backend/                    # Backend application (this directory)
-?   ??? Makefile               # Backend-only commands (dev, gateway, lint)
-?   ??? langgraph.json         # LangGraph Studio graph configuration
-?   ??? packages/
-?   ?   ??? harness/           # deerflow-harness package (import: deerflow.*)
-?   ?       ??? pyproject.toml
-?   ?       ??? deerflow/
-?   ?           ??? agents/            # LangGraph agent system
-?   ?           ?   ??? lead_agent/    # Main agent (factory + system prompt)
-?   ?           ?   ??? middlewares/   # 10 middleware components
-?   ?           ?   ??? memory/        # Memory V2: provider/manager/builtin/hindsight
-?   ?           ?   ??? thread_state.py # ThreadState schema
-?   ?           ??? sandbox/           # Sandbox execution system
-?   ?           ?   ??? local/         # Local filesystem provider
-?   ?           ?   ??? sandbox.py     # Abstract Sandbox interface
-?   ?           ?   ??? tools.py       # bash, ls, read/write/str_replace
-?   ?           ?   ??? middleware.py  # Sandbox lifecycle management
-?   ?           ??? subagents/         # Subagent delegation system
-?   ?           ?   ??? builtins/      # general-purpose, bash agents
-?   ?           ?   ??? executor.py    # Background execution engine
-?   ?           ?   ??? registry.py    # Agent registry
-?   ?           ??? tools/builtins/    # Built-in tools (present_files, ask_clarification, view_image)
-?   ?           ??? mcp/               # MCP integration (tools, cache, client)
-?   ?           ??? models/            # Model factory with thinking/vision support
-?   ?           ??? skills/            # Skills discovery, loading, parsing
-?   ?           ??? config/            # Configuration system (app, model, sandbox, tool, etc.)
-?   ?           ??? community/         # Community tools (tavily, jina_ai, firecrawl, image_search, aio_sandbox)
-?   ?           ??? reflection/        # Dynamic module loading (resolve_variable, resolve_class)
-?   ?           ??? utils/             # Utilities (network, readability)
-?   ?           ??? client.py          # Embedded Python client (DeerFlowClient)
-?   ??? app/                   # Application layer (import: app.*)
-?   ?   ??? gateway/           # FastAPI Gateway API
-?   ?   ?   ??? app.py         # FastAPI application
-?   ?   ?   ??? routers/       # FastAPI route modules (models, mcp, memory, skills, uploads, threads, artifacts, agents, suggestions, channels)
-?   ?   ??? channels/          # IM platform integrations
-?   ??? tests/                 # Test suite
-?   ??? docs/                  # Documentation
-??? frontend/                   # Next.js frontend application
-??? skills/                     # Agent skills directory
-    ??? public/                # Public skills (committed)
-    ??? custom/                # Custom skills (gitignored)
-```
-
-## Important Development Guidelines
-
-### Documentation Update Policy
-**CRITICAL: Always update README.md and CLAUDE.md after every code change**
-
-When making code changes, you MUST update the relevant documentation:
-- Update `README.md` for user-facing changes (features, setup, usage instructions)
-- Update `CLAUDE.md` for development changes (architecture, commands, workflows, internal systems)
-- Keep documentation synchronized with the codebase at all times
-- Ensure accuracy and timeliness of all documentation
-
-## Commands
-
-**Root directory** (for full application):
-```bash
-make check      # Check system requirements
-make install    # Install all dependencies (frontend + backend)
-make dev        # Start all services (Gateway + Frontend + Nginx), with config.yaml preflight
-make start      # Start production services locally
-make stop       # Stop all services
-```
-
-**Backend directory** (for backend development only):
-```bash
-make install    # Install backend dependencies
-make dev        # Run Gateway API with reload (port 8001)
-make gateway    # Run Gateway API only (port 8001)
-make test       # Run all backend tests
-make lint       # Lint with ruff
-make format     # Format code with ruff
-```
-
-Regression tests related to Docker/provisioner behavior:
-- `tests/test_docker_sandbox_mode_detection.py` (mode detection from `config.yaml`)
-- `tests/test_provisioner_kubeconfig.py` (kubeconfig file/directory handling)
-
-Boundary check (harness ? app import firewall):
-- `tests/test_harness_boundary.py` ? ensures `packages/harness/deerflow/` never imports from `app.*`
-
-CI runs these regression tests for every pull request via [.github/workflows/backend-unit-tests.yml](../.github/workflows/backend-unit-tests.yml).
-
-## Architecture
-
-### Harness / App Split
-
-The backend is split into two layers with a strict dependency direction:
-
-- **Harness** (`packages/harness/deerflow/`): Publishable agent framework package (`deerflow-harness`). Import prefix: `deerflow.*`. Contains agent orchestration, tools, sandbox, models, MCP, skills, config ? everything needed to build and run agents.
-- **App** (`app/`): Unpublished application code. Import prefix: `app.*`. Contains the FastAPI Gateway API and IM channel integrations (Feishu, Slack, Telegram, DingTalk).
-
-**Dependency rule**: App imports deerflow, but deerflow never imports app. This boundary is enforced by `tests/test_harness_boundary.py` which runs in CI.
-
-**Import conventions**:
-```python
-# Harness internal
-from deerflow.agents import make_lead_agent
-from deerflow.models import create_chat_model
-
-# App internal
-from app.gateway.app import app
-from app.channels.service import start_channel_service
-
-# App ? Harness (allowed)
-from deerflow.config import get_app_config
-
-# Harness ? App (FORBIDDEN ? enforced by test_harness_boundary.py)
-# from app.gateway.routers.uploads import ...  # ? will fail CI
-```
-
-### Agent System
-
-**Lead Agent** (`packages/harness/deerflow/agents/lead_agent/agent.py`):
-- Entry point: `make_lead_agent(config: RunnableConfig)` registered in `langgraph.json`
-- Dynamic model selection via `create_chat_model()` with thinking/vision support
-- Tools loaded via `get_available_tools()` - combines sandbox, built-in, MCP, community, and subagent tools
-- System prompt generated by `apply_prompt_template()` with skills, memory, and subagent instructions
-
-**ThreadState** (`packages/harness/deerflow/agents/thread_state.py`):
-- Extends `AgentState` with: `sandbox`, `thread_data`, `title`, `artifacts`, `todos`, `uploaded_files`, `viewed_images`
-- Uses custom reducers: `merge_artifacts` (deduplicate), `merge_viewed_images` (merge/clear)
-
-**Runtime Configuration** (via `config.configurable`):
-- `thinking_enabled` - Enable model's extended thinking
-- `model_name` - Select specific LLM model
-- `is_plan_mode` - Enable TodoList middleware
-- `subagent_enabled` - Enable task delegation tool
-
-### Middleware Chain
-
-Lead-agent middlewares are assembled in strict append order across `packages/harness/deerflow/agents/middlewares/tool_error_handling_middleware.py` (`build_lead_runtime_middlewares`) and `packages/harness/deerflow/agents/lead_agent/agent.py` (`_build_middlewares`):
-
-1. **ThreadDataMiddleware** - Creates per-thread directories under the user's isolation scope (`backend/.deer-flow/users/{user_id}/threads/{thread_id}/user-data/{workspace,uploads,outputs}`); resolves `user_id` via `get_effective_user_id()` (falls back to `"default"` in no-auth mode); Web UI thread deletion now follows LangGraph thread removal with Gateway cleanup of the local thread directory
-2. **UploadsMiddleware** - Tracks and injects newly uploaded files into conversation
-3. **SandboxMiddleware** - Acquires sandbox, stores `sandbox_id` in state
-4. **DanglingToolCallMiddleware** - Injects placeholder ToolMessages for AIMessage tool_calls that lack responses (e.g., due to user interruption), including raw provider tool-call payloads preserved only in `additional_kwargs["tool_calls"]`
-5. **LLMErrorHandlingMiddleware** - Normalizes provider/model invocation failures into recoverable assistant-facing errors before later middleware/tool stages run
-6. **GuardrailMiddleware** - Pre-tool-call authorization via pluggable `GuardrailProvider` protocol (optional, if `guardrails.enabled` in config). Evaluates each tool call and returns error ToolMessage on deny. Three provider options: built-in `AllowlistProvider` (zero deps), OAP policy providers (e.g. `aport-agent-guardrails`), or custom providers. See [docs/GUARDRAILS.md](docs/GUARDRAILS.md) for setup, usage, and how to implement a provider.
-7. **SandboxAuditMiddleware** - Audits sandboxed shell/file operations for security logging before tool execution continues
-8. **ToolErrorHandlingMiddleware** - Converts tool exceptions into error `ToolMessage`s so the run can continue instead of aborting
-9. **SummarizationMiddleware** (`DeerFlowSummarizationMiddleware`) - Context reduction when approaching token limits (optional, if enabled). **Non-destructive**: overrides the stock `before_model` (which persists a `RemoveMessage(REMOVE_ALL_MESSAGES)`) to a no-op and instead compresses **transiently** in `wrap_model_call` ? the model request is replaced with `[summary, *recent]` while the full original conversation stays in state (so the frontend keeps rendering the user's real Q&A). The `summary` message is `hide_from_ui` and never persisted; summaries are cached per-thread (bounded, sticky boundary) to avoid an LLM call on every model invocation. Tests: `tests/test_summarization_middleware.py`
-10. **TodoListMiddleware** - Task tracking with `write_todos` tool (optional, if plan_mode)
-11. **TokenUsageMiddleware** - Records token usage metrics when token tracking is enabled (optional)
-12. **TitleMiddleware** - Two-phase auto-title (never shows "untitled"): `before_model` sets an instant question-derived *provisional* title the moment the first user message arrives (no LLM, zero latency), then `after_model` asynchronously generates the final LLM title and replaces it (tracked via the `title_provisional` state flag). Normalizes structured message content before prompting the title model
-13. **MemoryMiddleware** - Memory V2: `wrap_model_call` injects Hindsight recall transiently; `after_agent` auto-retains the turn to Hindsight
-14. **ViewImageMiddleware** - Injects base64 image data before LLM call (conditional on vision support)
-15. **DeferredToolFilterMiddleware** - Hides deferred tool schemas from the bound model until tool search is enabled (optional)
-16. **SubagentLimitMiddleware** - Truncates excess `task` tool calls from model response to enforce `MAX_CONCURRENT_SUBAGENTS` limit (optional, if `subagent_enabled`)
-17. **LoopDetectionMiddleware** - Detects repeated tool-call loops; hard-stop responses clear both structured `tool_calls` and raw provider tool-call metadata before forcing a final text answer
-18. **ClarificationMiddleware** - Intercepts `ask_clarification` tool calls, interrupts via `Command(goto=END)` (must be last)
-8a. **ToolMetricsMiddleware** - Appended immediately after `ToolErrorHandlingMiddleware` in `_build_runtime_middlewares`, so it sits *deepest* in the `wrap_tool_call` chain. Records one `tool_call_metrics` row per tool/skill/MCP invocation ? timing duration and bucketing failures into a coarse `error_category` (rate_limited / ip_blocked / auth_error / timeout / network_error / not_found / empty_result / tool_error). It catches both raised exceptions (before #8 converts them) and *self-handled* errors returned as error `ToolMessage`s (search tools that swallow HTTP 429/403). It also captures the targeted Agent Skill into `tool_call_metrics.skill_name`: the lead agent invokes a skill by `read_file`-ing its `/mnt/skills/<category>/<name>/SKILL.md` path (per the `<skill_system>` prompt block ? there is no dedicated "use skill" tool), so `skill_name` is recovered by scanning every tool call's string arguments for the `skills/{public,custom}/{name}/` path shape (`_SKILL_PATH_RE`), plus the explicit `name` argument of `skill_view`/`skill_manage`/`skill_archive`. `skill_view`-specific success detection handles its swallowed failures (`Skill 'x' not found.` returned as a plain string rather than raised). Regression coverage: `tests/test_tool_metrics_middleware.py`. Writes are best-effort fire-and-forget. **Per-run skill merging**: `SqlToolMetricsStore.record` upserts skill rows by `(run_id, skill_name)` instead of inserting one row per call ? a skill touched by several tool calls / retries inside one run collapses to a single row. Any success in the run wins (a skill that failed then finally succeeded is recorded as one `success`); a skill that only ever failed yields one `error` row whose `error_message` accumulates every attempt's message; durations are summed. Non-skill tool calls, and skill calls with no `run_id`, keep one row each. The read?merge?write is guarded by a per-event-loop `asyncio.Lock`. Surfaced by the admin `GET /api/admin/tool-metrics` router (which has a `kind=skill` filter and a per-skill `/summary` breakdown).
-9. **PromptPrefixMiddleware** - Injects the admin-configured prompt prefix into the model request only. The prefix is tagged onto the first human message's `additional_kwargs['prompt_prefix']` by the Gateway (`app.gateway.services`) and prepended to content here at call time, so the persisted/displayed message stays prefix-free
-10. **SummarizationMiddleware** (`DeerFlowSummarizationMiddleware`) - Context reduction when approaching token limits (optional, if enabled). **Non-destructive**: overrides the stock `before_model` (which persists a `RemoveMessage(REMOVE_ALL_MESSAGES)`) to a no-op and instead compresses **transiently** in `wrap_model_call` ? the model request is replaced with `[summary, *recent]` while the full original conversation stays in state (so the frontend keeps rendering the user's real Q&A). The `summary` message is `hide_from_ui` and never persisted; summaries are cached per-thread (bounded, sticky boundary) to avoid an LLM call on every model invocation. Tests: `tests/test_summarization_middleware.py`
-11. **TodoListMiddleware** - Task tracking with `write_todos` tool (optional, if plan_mode)
-12. **TokenUsageMiddleware** - Records token usage metrics when token tracking is enabled (optional)
-13. **TitleMiddleware** - Two-phase auto-title (never shows "untitled"): `before_model` sets an instant question-derived *provisional* title the moment the first user message arrives (no LLM, zero latency), then `after_model` asynchronously generates the final LLM title and replaces it (tracked via the `title_provisional` state flag). Normalizes structured message content before prompting the title model
-14. **MemoryMiddleware** - Queues conversations for async memory update (filters to user + final AI responses)
-15. **ViewImageMiddleware** - Injects base64 image data before LLM call (conditional on vision support)
-16. **DeferredToolFilterMiddleware** - Hides deferred tool schemas from the bound model until tool search is enabled (optional)
-17. **SubagentLimitMiddleware** - Truncates excess `task` tool calls from model response to enforce `MAX_CONCURRENT_SUBAGENTS` limit (optional, if `subagent_enabled`)
-18. **LoopDetectionMiddleware** - Detects repeated tool-call loops; hard-stop responses clear both structured `tool_calls` and raw provider tool-call metadata before forcing a final text answer
-19. **ClarificationMiddleware** - Intercepts `ask_clarification` tool calls, interrupts via `Command(goto=END)` (must be last)
-
-### Configuration System
-
-**Main Configuration** (`config.yaml`):
-
-Setup: Copy `config.example.yaml` to `config.yaml` in the **project root** directory.
-
-**Config Versioning**: `config.example.yaml` has a `config_version` field. On startup, `AppConfig.from_file()` compares user version vs example version and emits a warning if outdated. Missing `config_version` = version 0. Run `make config-upgrade` to auto-merge missing fields. When changing the config schema, bump `config_version` in `config.example.yaml`.
-
-**Config Caching**: `get_app_config()` caches the parsed config, but automatically reloads it when the resolved config path changes or the file's mtime increases. This keeps Gateway and LangGraph reads aligned with `config.yaml` edits without requiring a manual process restart.
-
-Configuration priority:
-1. Explicit `config_path` argument
-2. `DEER_FLOW_CONFIG_PATH` environment variable
-3. `config.yaml` in current directory (backend/)
-4. `config.yaml` in parent directory (project root - **recommended location**)
-
-Config values starting with `$` are resolved as environment variables (e.g., `$OPENAI_API_KEY`).
-`ModelConfig` also declares `use_responses_api` and `output_version` so OpenAI `/v1/responses` can be enabled explicitly while still using `langchain_openai:ChatOpenAI`.
-
-**Extensions Configuration** (`extensions_config.json`):
-
-MCP servers and skills are configured together in `extensions_config.json` in project root:
-
-Configuration priority:
-1. Explicit `config_path` argument
-2. `DEER_FLOW_EXTENSIONS_CONFIG_PATH` environment variable
-3. `extensions_config.json` in current directory (backend/)
-4. `extensions_config.json` in parent directory (project root - **recommended location**)
-
-### Gateway API (`app/gateway/`)
-
-FastAPI application on port 8001 with health check at `GET /health`. Set `GATEWAY_ENABLE_DOCS=false` to disable `/docs`, `/redoc`, and `/openapi.json` in production (default: enabled).
-
-**Routers**:
-
-| Router | Endpoints |
-|--------|-----------|
-| **Models** (`/api/models`) | `GET /` - list models; `GET /{name}` - model details |
-| **MCP** (`/api/mcp`) | `GET /config` - get config; `PUT /config` - update config (saves to extensions_config.json) |
-| **Skills** (`/api/skills`) | `GET /` - list skills; `GET /{name}` - details; `PUT /{name}` - update enabled; `POST /install` - install from .skill archive (accepts standard optional frontmatter like `version`, `author`, `compatibility`) |
-| **Memory V2** (`/api/memory/v2`) | `GET /` - builtin data (USER.md/MEMORY.md); `POST/PUT/DELETE /entries` - add/replace/remove entry; `GET/PATCH/DELETE /config` - effective config + per-user overrides; `POST /search` - Hindsight recall/reflect proxy |
-| **Agents** (`/api/agents`) | `GET /` - list built-in, owned, and published agents; `POST /` - create user-owned agent; `GET/PUT/DELETE /{id}` - read/update/delete by stable agent id; `GET /check` - validate non-empty display name (duplicates allowed) |
-| **Skills** (`/api/skills`) | `GET /` - list skills (optional `search` fuzzy-name + `tag_id` filters; each item carries `tags`); `GET /{name}` - details; `PUT /{name}` - update enabled; `POST /install` - install from .skill archive (accepts standard optional frontmatter like `version`, `author`, `compatibility`) |
-| **Skills** (`/api/skills`) | `GET /` - list skills (optional `search` fuzzy-name + `tag_id` filters; each item carries `tags` + owner-authored `detail`); `GET /{name}` - details; `PUT /{name}` - update enabled; `PUT /{name}/detail` - set the long-form `detail` text (custom skill: owner or admin; built-in/public: admin only ? materializes a `skills` row on demand); `POST /install` - install from .skill archive (accepts standard optional frontmatter like `version`, `author`, `compatibility`) |
-| **Skills** (`/api/skills`) | `GET /` - list skills (optional `search` fuzzy-name + `tag_id` filters; each item carries `tags` + owner-authored `detail`); `GET /{name}` - details; `PUT /{name}` - update enabled; `PUT /{name}/detail` - set the long-form `detail` text (custom skill: owner or admin; built-in/public: admin only ? materializes a `skills` row on demand); `GET /{name}/content` - raw SKILL.md source, restricted to an admin or the publisher (custom-skill owner); other users get 403 and see only description + detail; `POST /install` - install from .skill archive (accepts standard optional frontmatter like `version`, `author`, `compatibility`) |
-| **Skills** (`/api/skills`) | `GET /` - list skills (optional `search` fuzzy-name + repeatable `tag_ids` multi-select filter, OR-matched; each item carries `tags` + owner-authored `detail`); `GET /{name}` - details; `PUT /{name}` - update enabled; `PUT /{name}/detail` - set the long-form `detail` text (custom skill: owner or admin; built-in/public: admin only ? materializes a `skills` row on demand); `GET /{name}/content` - raw SKILL.md source, restricted to an admin or the publisher (custom-skill owner); other users get 403 and see only description + detail; `POST /install` - install from .skill archive (accepts standard optional frontmatter like `version`, `author`, `compatibility`) |
-| **Memory** (`/api/memory`) | `GET /` - memory data; `POST /reload` - force reload; `GET /config` - config; `GET /status` - config + data |
-| **Agents** (`/api/agents`) | `GET /` - list built-in, owned, and published agents (optional `search` + repeatable `tag_ids` multi-select filter, OR-matched; each item carries `tags` and `featured_order`); `POST /` - create user-owned agent; `GET/PUT/DELETE /{id}` - read/update/delete by stable agent id; `GET /check` - validate non-empty display name (duplicates allowed); `GET /selector` - homepage agent picker (single admin-curated list, visibility-filtered, sorted by `featured_order` asc); `PUT /{id}/featured` - **admin** sets/clears the homepage `featured_order` (must be `published=true` to feature; passing `order: null` clears unconditionally); `PUT /featured/order` - **admin** bulk-reorders the homepage list (body `{agent_ids: [...]}` ? indices become 0..N-1) |
-| **Tags** (`/api/tags`) | `GET /` - list tags (optional `tag_type` filter; open to all); `POST /` - create tag (admin); `PUT /{id}` / `DELETE /{id}` - update/delete tag (admin); `GET /assignments/{target_type}/{target_id}` - tags on a resource; `PUT /assignments/{target_type}/{target_id}` - replace a resource's tags (admin). Tags are typed `agent`/`skill`/`scheduled_task` and may only tag resources of their own type. |
-| **Uploads** (`/api/threads/{id}/uploads`) | `POST /` - upload files (auto-converts PDF/PPT/Excel/Word); `GET /list` - list; `DELETE /{filename}` - delete |
-| **Threads** (`/api/threads/{id}`) | `DELETE /` - remove DeerFlow-managed local thread data after LangGraph thread deletion; unexpected failures are logged server-side and return a generic 500 detail |
-| **Threads search** (`POST /threads/search`, LangGraph-compat) | Conversation listing. **System threads are excluded by default**: rows whose `metadata.system` is truthy (scheduler runs, roundtable workers) never crowd the user's list; a caller whose `metadata` filter explicitly references `thread_type`/`system` (e.g. studio notebooks) opts out of the exclusion. New `query` field = case-insensitive title (`display_name`) substring search, used by the chats page's server-side pagination. `POST /threads/count` takes the same request body (limit/offset ignored) and returns `{total}` under identical filtering/exclusion semantics ? powers the chats page's ??/??? display without fetching rows (`ThreadMetaStore.count`). `ThreadMetaStore.search` gained `exclude_system`/`query` params; JSON-blob filters batch-scan until the requested `limit`/`offset` window is full (so pagination over a heavily-filtered set never under-fills), with a `thread_id` tie-breaker on the `updated_at DESC` ordering for stable OFFSET paging. Tests: `tests/test_thread_meta_search.py` |
-| **Artifacts** (`/api/threads/{id}/artifacts`) | `GET /{path}` - serve artifacts; active content types (`text/html`, `application/xhtml+xml`, `image/svg+xml`) are always forced as download attachments to reduce XSS risk; `?download=true` still forces download for other file types |
-| **Suggestions** (`/api/threads/{id}/suggestions`) | `POST /` - generate follow-up questions; rich list/block model content is normalized before JSON parsing |
-| **Thread Runs** (`/api/threads/{id}/runs`) | `POST /` - create background run; `POST /stream` - create + SSE stream; `POST /wait` - create + block; `GET /` - list runs; `GET /{rid}` - run details; `POST /{rid}/cancel` - cancel; `GET /{rid}/join` - join SSE; `GET /{rid}/messages` - paginated messages `{data, has_more}`; `GET /{rid}/events` - full event stream; `GET /../messages` - thread messages with feedback; `GET /../token-usage` - aggregate tokens |
-| **Feedback** (`/api/threads/{id}/runs/{rid}/feedback`) | `PUT /` - upsert feedback; `DELETE /` - delete user feedback; `POST /` - create feedback; `GET /` - list feedback; `GET /stats` - aggregate stats; `DELETE /{fid}` - delete specific |
-| **Runs** (`/api/runs`) | `POST /stream` - stateless run + SSE; `POST /wait` - stateless run + block; `GET /{rid}/messages` - paginated messages by run_id `{data, has_more}` (cursor: `after_seq`/`before_seq`); `GET /{rid}/feedback` - list feedback by run_id |
-| **Thread Shares** (`/api/thread-shares`) | Per-conversation sharing with two modes (`thread_shares.mode`). **`import`**: `POST /` `{thread_id, mode}` - create (or reuse) a revocable share code; `GET /{code}/preview` + `POST /{code}/import` - copy that conversation into the caller's account (fresh `thread_id` via `app/gateway/thread_copy.py` + new `threads_meta` row + `user-data` copy; dedup on `metadata.shared_origin`). **`view`**: a public, read-only static page ? see Public Shares below. Both modes: `GET /` - list my codes; `DELETE /{code}` - revoke. |
-| **Public Shares** (`/api/public/shares`) | **Unauthenticated** (the `/api/public/` prefix is whitelisted in `auth_middleware._PUBLIC_PATH_PREFIXES`). `GET /{code}` - read-only conversation snapshot (title + serialized messages + artifacts) for a `view`-mode share; messages are read straight from the latest checkpoint. `GET /{code}/files/{path}` - serves a file referenced by that conversation (images, Markdown, text, PDF, ?) so the shared page's links are openable; path resolution is confined to the owner's thread directory. Active content (`text/html`, `application/xhtml+xml`, `image/svg+xml`) is forced to download (`_public_share_disposition`); everything else is served inline. Missing/revoked/non-`view` codes all return 404. |
-| **Roundtable Drafts** (`/api/roundtable-drafts`) | Per-user persistence for the roundtable-planning page (replaces the old browser `localStorage`). `GET /` - list drafts (lightweight meta); `POST /` - create; `GET/PUT/DELETE /{id}` - full read / partial update / delete. A draft = one full session (Step 1 intent + Step 2 roundtable), step snapshots stored as JSON in `PortableLongText`. Recommendation history bound to a draft: `GET /{id}/recommendations` - list newest-first; `POST /{id}/recommendations` - append one run (objective/status/rationale/picks/candidates). Backed by `deerflow.persistence.roundtable_drafts` (tables `roundtable_drafts`, `roundtable_recommend_history`); store wired in `deps.py` as `app.state.roundtable_draft_store`. |
-| **Roundtable Draft Shares** (`/api/roundtable-draft-shares`) | Per-draft sharing for the roundtable-planning page, mirroring Thread Shares but the shared resource is a `roundtable_drafts` row. Two modes (`roundtable_draft_shares.mode`). **`import`**: `POST /` `{draft_id, mode}` - create (or reuse) a revocable code; `GET /{code}/preview` + `POST /{code}/import` - copy that draft (step1/2/3) into the caller's account under a deterministic per-(code,recipient) id (`imp_<sha1>` ? re-import is a no-op). **`view`**: a public, read-only result page ? see Public Roundtable Shares. Both: `GET /` - list my codes; `DELETE /{code}` - revoke. Backed by `deerflow.persistence.roundtable_draft_shares` (table `roundtable_draft_shares`, migration `20260606_01`); store wired in `deps.py` as `app.state.roundtable_draft_share_store`. Tests: `tests/test_roundtable_draft_shares.py`. |
-| **Public Roundtable Shares** (`/api/public/roundtable-shares`) | **Unauthenticated** (`/api/public/` whitelisted). `GET /{code}` - read-only snapshot for a `view`-mode draft share: returns the Step 3??????report HTML (`step3.html`, a self-contained document) + title/summary/`has_report`. The frontend renders it in a sandboxed `iframe` (no `allow-scripts`) at `#/roundtable-share/<code>`. Missing/revoked/non-`view` codes all return 404. |
-| **Scheduled Tasks** (`/api/scheduled-tasks`) | Per-user cron-like agent tasks. `GET/POST /` - list/create owned tasks; `PATCH/DELETE /{id}` - update/delete (owner); `POST /{id}/{pause,resume,trigger}`; `GET /{id}/runs`, `GET /runs/{rid}`, run-file preview/edit/download; `GET /runs/{rid}/messages` - agent messages for the run's underlying agent run (used to inspect a stuck/running run; authorized via `_load_run`); `GET /runs/{rid}/html` - generated HTML document for an `html_page` run; `DELETE /runs/{rid}` - owner-scoped, with an admin fallback so admins can delete any user's run; publish/subscribe under `/{id}/{publish,unpublish,subscribe}` and `GET /published`. **Admin (cross-user):** `GET /admin/all` - every task; `PATCH/DELETE /admin/{id}` - edit/delete any task; `POST /admin/{id}/trigger`; `GET /admin/{id}/runs`. Admin routes require `system_role == "admin"`. |
-| **HTML Page Favorites** (`/api/html-page-favorites`) | Bookmarked generated HTML pages, owner-scoped. `GET /` - list; `POST /` `{page_id, title?, description?, tags?}` - bookmark a `scheduled_html_pages` row (extracts a reusable style/layout summary via GLM5 at favorite time); a page can only be favorited once per user (`DuplicateFavoriteError` ? HTTP 409). `GET/DELETE /{id}` - read/delete (the create-task reference picker deletes favorites here). A favorite can be selected as a style reference when creating a new `html_page` task. |
-| **Public Scheduled Tasks** (`/api/public/scheduled-tasks`) | **Unauthenticated** mirror for embedding. Adds `GET /runs/{rid}/html` and `GET /{task_id}/latest-html` for the `html_page` preview; the run result keeps `result.html` (only file paths are redacted). |
-| **Light Apps** (`/api/light-apps`) | Light-app registry (????? / ????), **global/shared** (no user scoping; `created_by`/`updated_by` audit-only). Two kinds: `window_open` (launched in a new tab from the ???? card grid) and `iframe` (embedded under a parent sidebar menu via `mount_location` and dynamically injected into the sidebar). Both carry `route_params` ? a list of `{key, paramType: login\|theme, source, customValue, themeValue}` descriptors the **frontend** resolves at launch time (login params read from `localStorage.userInfo`; theme params use a literal value) and appends to `url` as query string. `GET /` - list (filters `app_type`/`mount_location`/`enabled_only`/`search`); `GET /{app_id}` - read one (both open to any authenticated user); `POST /`, `PUT /{app_id}`, `DELETE /{app_id}` - **admin only**; iframe apps must carry a `mount_location` (400 otherwise); the unique `name` collides as HTTP 409. Backed by `deerflow.persistence.light_apps` (table `light_apps`, auto-created by `create_all`); store wired in `deps.py` as `app.state.light_app_store`. Tests: `tests/test_light_apps.py`. |
-| **Knowledge** (`/api/knowledge`) | Product knowledge base, **global/shared** (no user isolation; writes record `created_by`/`updated_by` for audit). `POST /ingest/thread/{thread_id}` - sediment one thread into a note (reads messages from the checkpointer, `threads:read` owner-checked); `POST /ingest/threads/batch` - batch-sediment the caller's recent threads (`skip_existing` via `.manifest.json` dedup). **Optional model selection**: both ingest endpoints accept `model_name` (overrides `knowledge.extract_model_name` for that capture; threaded through `KnowledgeService.capture_thread` ? `extract_thread_llm_multi`). **Model-error surfacing**: when the chosen model produces nothing usable (timeout / API error / non-JSON) capture still **falls back to rule-based extraction and lands a draft** (`llm_fallback:true`); the single-thread (non-background) response and the batch response (`llm_fallback` count) carry this so the UI tells the user "???????". The "??????" chat button stays background/fire-and-forget (passes `model_name`, no per-save error toast); the batch-import dialog is synchronous and shows the fallback count. `POST /ingest/file` - **import an uploaded file** (multipart: `file` + optional `title`/`tags`/`status`/`model_name`) into wiki note(s): the file is converted to Markdown (md/txt read directly, pdf/word/ppt/excel via `convert_file_to_markdown`), distilled by the same wiki-capture LLM pipeline as thread capture (`source_type="file"`), and falls back to a verbatim **draft** note when the model returns nothing usable (`llm_fallback:true`); `POST /ingest/search` - sediment search/tool results as a reference note; `POST /notes` - manual note; `GET /notes` (keyword/tag/source_type/status + pagination); `GET/PUT/DELETE /notes/{id}` (DELETE is soft = archive, `?hard=true` removes the vault file); `POST /search` - `mode=keyword\|vector\|hybrid` (vector?keyword fallback when no embedding endpoint); `GET /notes/{id}/versions` - edit history; `GET /duplicates` + `POST /merge` - dedup; `POST /reindex` - rebuild vector index; `GET /graph` - product graph data (notes + entities, graph-blacklist filtered); `GET/POST /graph/blacklist` + `DELETE /graph/blacklist/{id}` - graph node blacklist (labels = entity names / note titles, case-insensitive match; blacklisted nodes and their edges are dropped from `/graph` and the wiki-export artifacts; table `knowledge_graph_blacklist`, auto-created); `POST /export/graph` + `GET /export/files/{graph.json\|graph.html\|graph.graphml\|cypher.txt}` - obsidian-wiki `wiki-export` artifacts (self-contained `graph.html` for iframe); `POST /resolve` `{targets:[?]}` + `GET /notes/{id}/wikilinks` - resolve `[[wikilink]]` targets to a note id / vault page / missing (powers clickable wikilinks); `GET /vault-page?path=` - read a raw vault meta/hub Markdown page (no DB note). Backed by the `deerflow.knowledge` module (see below). |
-
-Proxied through nginx: `/api/langgraph/*` ? LangGraph, all other `/api/*` ? Gateway.
-
-### AI Writing Built-in Functional Agents (????????)
-
-The AI-writing graph (`packages/harness/deerflow/agents/ai_writing/`) has four stage "brains" ? ?????? / ????? / ?? / ?? ? that can each run in one of two modes, switched per-stage by `config.yaml ? ai_writing.use_builtin_agents.{researcher,outliner,writer,editor}` (default **false**, mtime hot-reloaded, new sessions pick up changes immediately):
-
-- **Legacy (flag off)**: hard-coded prompts in `prompts/*.py`, direct LLM call (unchanged behavior).
-- **Builtin-agent (flag on)**: the node loads the corresponding functional agent from `.deer-flow/agents/ai-writing-<stage>/` ? `SOUL.md` becomes the system prompt and `config.yaml`'s `model` overrides the stage model (priority: `agent.model > node explicit fallback (keyword_model/rank_model) > state.model_name`). Admins edit SOUL/model from the agent management page. Runner: `nodes/_agent_runner.py` (`run_writing_agent` / `stream_writing_agent` / `parse_marker_json`). Output contracts (marker + ```json``` blocks, intent.py-style): researcher `[KEYWORDS_READY]` / `[MATERIALS_RANKED]`, outliner `[OUTLINE_READY]`, editor (per-dimension) `[REVIEW_READY]`; writer emits raw section Markdown (keeps per-section streaming) with the legacy `[[SECTION_BLOCKED]]` sentinel in strict mode. **Orchestration stays in the nodes** (search fan-out, per-section loop, 3-dimension review loop, all pause points) ? the agent only does single structured generations, and a missing/empty SOUL silently falls back to the legacy path (`WritingAgentUnavailable`).
-
-Seeding mirrors the roundtable pattern: verified `config.yaml` + `SOUL.md` ship in `app/gateway/routers/_ai_writing_seed_assets/<id>/`; `_ai_writing_seed.py::ensure_ai_writing_functional_agents()` copies them to `.deer-flow/agents/` when missing (never overwrites local edits), called from the startup lifespan (before `_sync_legacy_agents`, so the agents appear in `/api/agents` on first boot) and self-heals from `GET /api/ai-writing/sessions` + `GET /api/ai-writing/article-types`. Tests: `tests/test_ai_writing_seed.py`, `tests/test_ai_writing_agent_runner.py`.
-
-**Skill-driven retrieval (检索类型).** `material_source` has exactly two values: `general` (通用检索, default) and `notebook` (我的空间); legacy `knowledge_base`/`skill` are normalized to `general` by the researcher node. General retrieval is orchestrated from the **skills configured on the `ai-writing-researcher` agent** (`config.yaml → skills`, editable from the agent management page): the built-in `knowledge-base-search` skill (知识库检索) runs the native keyword fan-out web/intranet search (the old default path), while any other configured skill searches with a `description`-targeted query; skills run in configured order with early-stop at `max_materials`, and an empty/unresolvable skills config falls back to the built-in knowledge search so retrieval always works. The skill itself is seeded from `_ai_writing_seed_assets/skills/knowledge-base-search/SKILL.md` into `skills/public/` (never overwrites), and a pre-existing researcher `config.yaml` lacking the `skills` key is patched in place with the default (explicit `skills: []` is respected). **PAUSE-1 input append flow**: when the user types natural language into the material-confirm input (`re_search`/legacy `enable_skill_search` + `user_query`), the node matches the most relevant configured skill via `SkillPicker(candidates=…)` (skipped when only one skill is configured), searches with the user's words, and **appends** the new materials to the existing package (URL-dedup, original order preserved, no LLM re-rank/truncation, 100-item hard cap). **Skill-failure fault tolerance**: a configured skill that times out / raises / has all its keyword searches fail (vs. legitimately returning 0 results) no longer just silently moves on — the researcher node emits a `skill_search_step` of type `skill_error` (carrying the error `reason`), and if the configured skills collected **no** new materials at all it runs a `fallback` pass over other available skills (`_fallback_candidate_skills`: the built-in `knowledge-base-search` native path first, then other enabled catalog skills not yet tried) instead of erroring out. The frontend `SkillStepsCard` renders `skill_error` (rose) and `fallback` (amber) so the timeline shows "技能报错 → 改用其他技能搜集". Tests: `tests/test_ai_writing_material_source.py`.
-
-### Admin Leaderboard Snapshots (`/api/admin/users/leaderboard*`)
-
-The admin ????? is served from **precomputed per-day snapshots** in `admin_leaderboard_daily_stats` (keyed by `(stat_date, settings_hash)`), not by aggregating raw rows on every request. A background scheduler inside the Gateway (`_leaderboard_snapshot_scheduler_loop` in `app/gateway/app.py`, singleton file-locked) ticks every `ADMIN_LEADERBOARD_BACKGROUND_INTERVAL_SECONDS` (default 300s): it prewarms recent days, then builds any `queued`/`error`/stale-`running` day. `GET /leaderboard` merges the per-day snapshots for the requested range. **Today** is rebuilt whenever its snapshot is older than `ADMIN_LEADERBOARD_TODAY_TTL_SECONDS` (default 43200s = half a day, so the leaderboard refreshes ~twice daily); a past day, once `ready`, is treated as fresh forever.
-
-Two on-demand/automatic refresh paths layer on top:
-- **Manual "????"** ? `POST /leaderboard/refresh-today` force-rebuilds *today's* snapshot **synchronously** (bounded by the same semaphore/timeout the scheduler uses), then returns the merged range. Lets an admin pull the latest real-time records into the stats without waiting for the TTL. Frontend: a "????" button on `AdminUserLeaderboardPage` ? `useRefreshTodayLeaderboard` (`core/admin/hooks.ts`), which primes + invalidates the `["admin-activity","leaderboard"]` query.
-- **Daily finalize of the previous day** ? once per day, on/after `ADMIN_LEADERBOARD_FINALIZE_HOUR` (local Beijing, default `1` = ?? 1 ?), the scheduler tick force-recomputes *yesterday* a single time so runs/tool-calls that only landed near midnight are fully captured. The decision is the pure `deerflow.persistence.admin_stats.finalize.should_finalize_previous_day` (status/generated_at/now/hour) ? the rebuild stamps `generated_at` to today, which self-debounces it for the rest of the day (no extra state file). Tests: `tests/test_leaderboard_finalize.py`.
-
-### Scheduled Task Runtime
-
-`ScheduledTaskService` (`deerflow/runtime/scheduler/service.py`) polls due tasks every `poll_interval_seconds` and executes each in its own throwaway thread, serialized by a shared `_agent_run_lock`.
-
-**Stuck-run recovery**: each execution attempt is bounded by `_RUN_ATTEMPT_TIMEOUT_SECONDS` (30 min) via `asyncio.wait_for`. On timeout the run is retried up to `_MAX_RUN_RETRIES` (3) times, then marked `failed`. A watchdog (`_reconcile_stuck_runs`, run every poll) recovers runs left in `running` with no in-process owner ? e.g. orphaned by a crash or restart ? re-driving them through the same retry budget so a task can never stay perpetually "executing". The `scheduled_task_runs.retry_count` column tracks consumed retries.
-
-**Unique task names**: task names are unique per user (`uq_scheduled_tasks_user_name`). `create_task`/`update_task`/`admin_update_task` proactively check via `_assert_name_available()` and raise `DuplicateTaskNameError` on a collision; the router translates that into HTTP `409` instead of letting the raw DB `IntegrityError` surface as a `500`.
-
-**HTML page tasks (`task_kind == "html_page"`)**: a second task kind that renders a full HTML document each run instead of a Markdown report. The kind + its config (`html_config`: `page_prompt`, `negative_prompt`, `reference_favorite_id`, `model_name`) live inside `execution_context` ? no schema change to `scheduled_tasks`. Data gathering is delegated to the agent's own search skills (driven by `page_prompt`), so there is no explicit search-query/data-source field; the render model is user-selectable (`model_name`, default `zai-org/GLM-5-FP8`). `_run_task_attempt()` dispatches by kind; `_run_html_page_attempt()` runs the pipeline: gather real data via the lead agent (its search tools) ? optionally load a reference favorite's reusable style/layout ? render with the chosen model (thinking off) ? fault-tolerant JSON parse ? sanitize (BeautifulSoup, strips `<script>`/`<iframe>`/`on*`/`javascript:`) ? structural validate ? one GLM5 repair pass if invalid ? **completeness gate**: if the HTML is still not a full valid document (missing doctype/html/head/body, empty body, or residual scripts) after the repair pass, raise `HtmlGenerationError` so the run is marked **failed** (a page only counts as successful when it is a complete, previewable HTML document) ? best-effort Playwright screenshot (no-op when Playwright absent) ? archive `index.html` ? persist a `scheduled_html_pages` row ? return `result` with `format: "html"` and the full HTML inline. Pure helpers live in `deerflow/runtime/scheduler/html_page.py`. The frontend previews the HTML in a sandboxed `iframe` (no `allow-scripts`) on the run-detail and public pages; a user can bookmark a page (style summary extracted via GLM5) and later pick it as a style reference. New tables `scheduled_html_pages` + `html_page_favorites` (`deerflow/persistence/html_pages/`, wired as `app.state.html_page_store`) are created by `Base.metadata.create_all` with no migration. Tests: `tests/test_html_page_tasks.py`.
-
-**Unified output switches (`require_html` + `require_markdown`).** The task form no longer splits "???? / HTML ??" into separate kinds. `execution_context` now carries two **independent** booleans that may both be on: `require_html` (???? HTML) and `require_markdown` (???? Markdown). `_task_requires_html()` returns true for `require_html` **or** legacy `task_kind == "html_page"` (so old tasks need no data migration), and `_run_task_attempt()` dispatches to `_run_html_page_attempt` when it is true, else the regular agent attempt. When **both** switches are on, `_gather_html_data()` returns `(text, files)` ? the gather agent run is additionally told to save a `.md` file ? and `_run_html_page_attempt` validates it (`MarkdownRequirementError` on failure) and surfaces it alongside the generated page, so the run delivers both artifacts. The single main `prompt` doubles as the HTML page requirement (`html_config.page_prompt`); there is no separate page-prompt field. Tests: `tests/test_scheduled_task_require_html.py` + the markdown-coexistence cases in `tests/test_html_page_tasks.py`.
-
-### Per-user Prompts & UI Preferences
-
-**Custom prompts can be scoped to one agent.** `user_custom_prompts` gains a nullable `agent_id` column (migration `20260603_02`): `NULL` = global (the chip shows in every chat), otherwise the prompt only surfaces when that agent is selected. The visibility filter is applied in the frontend `CustomPromptBar` (it receives the current agent id from `ChatPage`'s selector or `AgentChatPage`'s `assistantId`); the backend (`/api/user-prompts` CRUD) just stores/returns/clears `agent_id` (update accepts an explicit `null` to clear). Tests: `tests/test_user_prompts_agent_id.py`.
-
-**Per-user UI preferences** (`deerflow/persistence/user_preferences/`, wired as `app.state.user_preferences_store`; migration `20260603_03`, also auto-created by `create_all`): one JSON-blob row per user so new toggles need no schema change. Router `/api/user-preferences` (`GET`/`PUT`, account-scoped). Current key: `hide_admin_prompt_prefix` ? when true the chat input hides the admin-configured prompt-prefix switch and never applies the prefix (the user uses only their own prompts; enforced in `ChatPage` for both `shouldApplyPrefix` and `promptPrefixProp`). Toggled from the "???????" dialog. Tests: `tests/test_user_preferences.py`.
-
-**HTML skills (tools the render step uses).** Two offline skills under `skills/public/` shape the generated HTML and are toggled in the skills settings page (enabled by default): `html-page-builder` (a self-contained inline-CSS design system ? colors/cards/tables/KPI/responsive) and `html-charts` (inline-SVG charts that render in the no-`allow-scripts` preview iframe; optional local ECharts documented in its `assets/README.md`). A skill opts into HTML injection via `metadata.kind: html_page` in its SKILL.md frontmatter; `_collect_html_skill_instructions()` reads the bodies of **enabled** such skills and `build_generation_prompt(skill_instructions=?)` injects them, so the (raw-model) render step honors the user's skill selection. The data-gathering step and the render/repair/style passes all use the task's selected `model_name` (no silent fallback to the default model ? and there is **no hardcoded GLM-5**; an unset selection falls back to config `models[0]`). To stop auxiliary LLM calls from leaking onto the default model, **`TitleMiddleware` is skipped when `is_scheduled_run` is set** (`_build_middlewares` in `lead_agent/agent.py`) ? scheduled/HTML runs use throwaway threads whose titles are never shown, and `config.title.model_name` is usually unset (? would call `models[0]`, not the chosen model). The completeness gate (`validate_html`) also requires a non-empty `<title>` and forbids external stylesheets, on top of the doctype/html/head/body + non-empty-body + no-script checks.
-
-### Knowledge Base (`packages/harness/deerflow/knowledge/`)
-
-Self-contained product knowledge base. **All knowledge capability lives in this one module** (harness layer, `deerflow.*` only ? no `app.*` imports). Wired as `app.state.knowledge_service` in `deps.py` via `make_knowledge_service(sf, config.knowledge)` (returns `None` on the memory backend or when `knowledge.enabled=false` ? router 503). Surfaced by `app/gateway/routers/knowledge.py` (`/api/knowledge`). Config: `config.yaml` ? `knowledge` (`deerflow/config/knowledge_config.py`).
-
-**Two stores, kept in sync on every write:**
-- **DB mirror** ? tables `knowledge_notes` / `knowledge_sources` / `knowledge_tags` / `knowledge_note_tags` (`knowledge/models.py`, registered in `persistence/models/__init__.py`, auto-created by `Base.metadata.create_all`; portable types `PortableLongText`/`BeijingDateTime`). Powers listing/filter/search/audit. Global/shared ? no `user_id` column; `created_by`/`updated_by` are audit-only (resolved via `get_effective_user_id()`).
-- **Markdown vault** ? obsidian-wiki compatible, default `{$DEER_FLOW_HOME}/data/knowledge/obsidian-vault`. `ObsidianWikiAdapter` (`obsidian_adapter.py`) writes wiki-capture frontmatter+body, and after **every** write syncs `index.md` / `log.md` (CAPTURE/INGEST/UPDATE/ARCHIVE/EXPORT) / `hot.md` / `.manifest.json` (source-hash dedup for `skip_existing`). `markdown_vault.py` handles frontmatter build/parse, slugging, wikilink extraction and **path-traversal-proof** resolution.
-
-**Module layout:** `models.py` (ORM: notes/sources/tags/note_tags + embeddings/versions/entities/relations) ? `repository.py` (DB only, no LLM) ? `schemas.py` (NoteDraft/SourceDraft DTOs) ? `skill_templates.py` ? `markdown_vault.py` ? `manifest.py` ? `extractor.py` (rule-based fallback) ? `llm_extractor.py` (**default**: distills declarative knowledge per wiki-capture, structured JSON incl. entities/relations; falls back to `extractor.py`) ? `sources.py` ? `exporter.py` (`graph.json`/`graph.graphml`/`cypher.txt` + self-contained vanilla-JS `graph.html`; entity nodes + relation edges. **`graph.html` is theme-aware** ? it reads `?theme=light|dark` from its own URL since the embedding iframe can't inherit the app's CSS; `KnowledgeGraphPage.tsx` appends the active day/night mode and re-skins on toggle. The layout cools down and **freezes** so dense graphs stay clickable, node radius scales with degree, and hovering/clicking a node focuses it + its direct neighbours while dimming the rest + a detail panel ? taming heavy node/edge clutter) ? `chunking.py` + `embeddings.py` (OpenAI-compatible embedding client, keyword fallback) + `search.py` (keyword/vector/hybrid) ? `sensitive.py` (secret/PII redaction) ? `dedup.py` (Jaccard + hash dedup) ? `auto_ingest.py` (background queue/worker) ? `obsidian_adapter.py` ? `service.py` (`KnowledgeService` orchestrator ? the only thing the router touches; capture entry points `capture_thread` / `capture_search` / `capture_file` / `create_manual_note`, all converging on `_persist_draft`). Tests: `tests/test_knowledge.py` (20 cases). Frontend: `frontend-web/src/core/knowledge/*` + `pages/KnowledgeBasePage.tsx` / `KnowledgeGraphPage.tsx` + chat `knowledge-save-trigger.tsx` / `knowledge-rag-toggle.tsx`; spec at `frontend-web/docs/product-knowledge-base-dev.md`.
-
-**Wikilinks & multi-note capture** (`knowledge.{related_notes_enabled,related_notes_limit,multi_note_enabled}`, all default-on; `project_name` defaults to `zncm`):
-- **Navigable `[[wikilinks]]`** ? `service.resolve_wikilink(s)` maps a `[[target]]` to a knowledge note (by id ? `vault_path` ? title) or a vault meta page (`read_vault_page`, path-traversal-proof), else `missing`. `markdown_vault.normalize_wikilink_target` canonicalizes targets (drops `|alias`, slashes, `.md`). The frontend `MarkdownRenderer` renders non-numeric `[[...]]` as clickable links from the resolver map (numeric `[[123]]` stay citation badges); note links navigate, vault links open a `VaultPageDialog`, missing links are greyed.
-- **Real related notes only (no structural meta-links)** ? `extractor.related_links()` now returns `[]`: notes no longer auto-link `index` / `_meta/taxonomy` / a project hub (they were noise in every `## Related`). On capture, `service._find_related_notes` cross-links **only by shared *distinctive* tags** (project/`knowledge-base` defaults excluded; free-text title/summary matching was removed because CJK tokenization linked unrelated notes via generic words like ??/????) and `_inject_related_notes` writes real `[[Title]]` links into `## Related` (plus note?note `related` graph edges); no distinctive tags ? no related notes. The renderers omit the `## Related` section entirely when there are no related notes and no sources. The vault's own `index.md`/`_meta/taxonomy.md` still exist (obsidian scaffold) and remain resolvable for manually-written links. The router enriches `GET /notes/{id}` with `created_by_name`/`updated_by_name` (auth-provider email lookup) so the UI shows a person, not a UUID. The frontend `MarkdownRenderer` renders wiki-capture provenance markers `^[inferred]`/`^[ambiguous]` as small "??"/"??" badges.
-- **LLM-extraction outcomes** ? capture distinguishes three LLM outcomes: (1) **success** ? declarative note(s); (2) **deliberate rejection** ? the model returns `should_ingest:false` with no notes (greeting / meta chatter) ? `extract_thread_llm_multi` raises `LLMRejectedIngest` and `capture_thread` raises `EmptyThreadError` (skip ? nothing persisted, so worthless chats never enter the KB); (3) **genuine failure** ? API error / non-JSON ? returns `[]`, capture falls back to the rule-based verbatim extractor and lands the note as a **`draft`** (not `approved`), logs a WARNING, and returns `llm_fallback:true` (surfaced by the router). `llm_extractor` logs an actionable WARNING/INFO for each non-success path so extraction issues are diagnosable.
-- **Multi-note split** ? `llm_extractor.extract_thread_llm_multi` lets one conversation distill into 1?4 cross-linked notes (`notes:[?]` JSON array; legacy single-object output still accepted). `capture_thread` persists each, sibling-cross-links them by title, and returns `{note, created, notes:[?]}`. Gated by `multi_note_enabled` (false ? first note only). Tests: `tests/test_knowledge_wikilinks.py`.
-
-**Phases 2?4 (all config-gated, defaults safe):**
-- **Phase 2 ? vector RAG**: `knowledge.embedding.{enabled,base_url,api_key,model,...}` enables an OpenAI-compatible embedding endpoint; notes are chunked + embedded on write (`knowledge_embeddings`), `POST /search` supports `mode=keyword|vector|hybrid` (vector degrades to keyword when unconfigured). `KnowledgeRagMiddleware` (lead-agent chain) injects a transient `<knowledge-context>` recall before the model call, gated **per-conversation** by the `knowledge_rag_enabled` runtime flag (frontend chat toggle) falling back to global `knowledge.rag_enabled`. `POST /reindex` rebuilds the index.
-- **Phase 3 ? governance**: `sensitive.py` redacts secrets/PII before persistence (`sensitive_redaction_enabled`); every update snapshots `knowledge_note_versions` (`GET /notes/{id}/versions`); `GET /duplicates` + `POST /merge` for dedup; background `AutoIngestQueue` (wired in `deps.py`, started/stopped in the gateway lifespan) auto-sediments finished conversations **off the chat path** when `auto_ingest_enabled` ? enqueued from `KnowledgeRagMiddleware.after_agent`, value-judged by min-chars + `skip_existing`, landing as `draft`.
-- **Phase 4 ? graph**: LLM extraction also yields entities/relations ? `knowledge_entities`/`knowledge_relations`; `exporter.build_graph` adds entity nodes + note?entity/entity?entity edges (reflected in `graph.html`/`graph.json`); `GET /graph` returns product graph data.
-
-### Custom Agent Persistence
-
-Custom agent metadata lives in the `agents` ORM table (`deerflow.persistence.agents`). `agents.id` is the stable runtime id and the filesystem directory name under `.deer-flow/agents/{id}/`; `agents.name` is display-only, can contain Chinese characters, and is not unique. `user_id IS NULL` marks built-in agents, user-owned agents are visible only to their owner unless `published = true`, and update/delete operations require ownership. The runtime accepts `agent_id` in context/configurable, and non-default `assistant_id` values are treated as agent ids after visibility checks.
-
-**Listing cache + pagination** ? to keep `GET /api/agents` fast, the gallery's per-card extras (skills + tool_groups) are denormalized into two cache tables instead of reading every agent's `config.yaml` on each request: `agent_skills` (relational, one row per (agent, skill) with `position`) and `agent_extras` (one row per agent: `tool_groups` JSON + `model` + `skills_synced_at`). `config.yaml` remains authoritative for the agent runtime; these tables are a read cache, write-through on create/update and backfilled from disk by `_sync_legacy_agents` (mtime-vs-`skills_synced_at` drift check). The list endpoint enriches via `AgentStore.get_extras_for([ids])` (two batched `IN` queries). Both tables are auto-created by `Base.metadata.create_all` (no migration needed). `GET /api/agents` supports **opt-in server-side pagination**: `scope` (`mine` = owned + favorited / `square` = built-in/published public), `page`, `page_size`, returning `{agents, total, page, page_size}`; omitting `scope`/`page` preserves the legacy "return everything" behavior. Name `search` matches the raw OR desensitized name (the frontend's static term?abbrev map is mirrored server-side in `deerflow/config/agent_name_desensitize.py`); `tag_ids` is an SQL OR-filter; `owner` keeps only agents published by that user id (the card's ??? ? built-in agents never match), powering the gallery's click-???-to-filter. `POST /api/agents/private-square/list` takes the same pagination params (incl. `owner`). Repository entry point: `AgentStore.list_paginated(...)`. Tests: `tests/test_agent_listing_cache.py::test_list_paginated_owner_filter_keeps_one_publisher`.
-
-**Homepage selector (agent picker on the new-chat page)** ? single admin-curated list backed by `agents.featured_order` (nullable BIGINT). NULL = not featured. Only `published = true` agents can be featured (router enforces on writes); any published agent is eligible regardless of ownership. Cleared with `order: null`. The list rendered to end users is `WHERE featured_order IS NOT NULL` + visibility-filter, sorted by `featured_order ASC`. Admins reorder by sending the desired ids to `PUT /api/agents/featured/order` (assigns indices 0..N-1). BIGINT is used rather than INT so the management UI can seed initial values from `Date.now()` (overflows INT on MySQL).
-
-### Publish Squares (????)
-
-Admin-configurable, **access-control-free** categories that published agents/skills are filed under (distinct from the password-gated ????). The square list lives in `system_settings.json` ? `publish_squares` (`PublishSquaresSettings`: ordered `squares: [{id,name,description}]` + `default_square_id`); helpers `list_publish_squares` / `resolve_default_square_id` / `normalize_square_id` always yield a valid square (built-in "????" fallback when none configured). Admin CRUD: `GET/PUT /api/system-settings/publish-squares` (GET readable by any user ? the gallery sub-tabs + publish picker need it; PUT admin-only, auto-slugs missing ids and re-anchors a dangling default).
-
-Each agent/skill carries a single `square_id` column (`AgentRow` / `SkillRow`, `String(64)` NOT NULL default `''`, migration `20260611_01`, idempotent SQL in `column_additions_mysql.sql`). One column = at most one square, so an item can never be "published twice". `''` = unassigned ? resolved to the default square at read/filter time. Writes normalize via `normalize_square_id`: agents on create/update (`square_id` field), skills on publish (`PUT /api/skills/custom/{name}/published` accepts optional `square_id`). Listing filters: `GET /api/agents?square=` and `GET /api/skills?square=` keep only that square's items; the **default** square additionally sweeps in legacy `''` rows (`square_is_default`). Tests: `tests/test_publish_squares.py`. Frontend: `core/system-settings` (`usePublishSquares`/`useUpdatePublishSquares`), admin `PublishSquaresSettingsPage` (embedded in the appearance/settings hub), square picker in the agent edit dialog + per-skill card, and a ??-tab square sub-filter in both galleries.
-
-### Portable Column Types (`deerflow/persistence/types.py`)
-
-Two `TypeDecorator`s keep ORM columns correct across the `sqlite` / `postgres` / `mysql` backends. **All ORM timestamp/JSON columns must use these, not the raw SQLAlchemy types.**
-
-- **`PortableJSON`** ? native `JSON` on SQLite/Postgres; `TEXT` (with manual `json.dumps`/`loads`) on MySQL, where older servers lack JSON support.
-- **`PortableLongText`** ? a large text column: maps to MySQL `LONGTEXT` (4 GB) and plain `TEXT` on SQLite/Postgres (already unbounded). Use for big JSON/markdown blobs that would overflow MySQL's 64 KB `TEXT` cap ? e.g. `ai_writing_sessions.transcript`.
-- **`BeijingDateTime`** ? stores timestamps as Beijing (UTC+8) wall-clock time. The app records time with `datetime.now(UTC)`, but MySQL `DATETIME` / SQLite have no timezone and persist whatever wall clock the driver serializes ? so a raw `DateTime` column would store UTC and read 8 hours behind Beijing. `BeijingDateTime` converts UTC?Beijing on bind and re-attaches `+08:00` on result, so the database holds Beijing time while Python still gets timezone-aware datetimes (instant comparisons stay correct, including in WHERE clauses). PostgreSQL `timestamptz` stores the instant unambiguously, so values pass through unchanged there. Uses a fixed UTC+8 offset (China has no DST), so no `tzdata` dependency.
-
-### Schema Migrations / Manual Column-Add SQL (`scripts/sql/`)
-
-Alembic auto-migration is **disabled** ? `_run_alembic_upgrade(url)` is commented out in `persistence/engine.py`. On startup only `Base.metadata.create_all` runs, which creates **new tables** but never patches **existing** ones (no new columns, no new indexes). So a column added to an ORM model appears automatically on a brand-new deployment, but is **missing on any existing MySQL database** ? and code reading/writing that column will error there until the column is added by hand.
-
-**Convention (required):** whenever code adds a field to an existing table (a new `mapped_column` on an ORM model), you must, in the same change:
-1. Keep the matching alembic version file under `persistence/migrations/versions/` (guarded with `_has_table`/`_has_column`), and
-2. Append the idempotent `ADD COLUMN` statement to **`scripts/sql/column_additions_mysql.sql`**, with a definition matching the ORM column exactly (type, length, nullability).
-
-Operators upgrading an existing DB run the whole file (`USE <db>;` then execute) ? it is idempotent (existing columns are skipped via `information_schema.columns`), so it is safe to re-run. Index-only changes go in the sibling `scripts/sql/leaderboard_indexes_mysql.sql` (same idempotent pattern).
-
-### Sandbox System (`packages/harness/deerflow/sandbox/`)
-
-**Interface**: Abstract `Sandbox` with `execute_command`, `read_file`, `write_file`, `list_dir`
-**Provider Pattern**: `SandboxProvider` with `acquire`, `get`, `release` lifecycle
-**Implementations**:
-- `LocalSandboxProvider` - Singleton local filesystem execution with path mappings
-- `AioSandboxProvider` (`packages/harness/deerflow/community/`) - Docker-based isolation
-
-**Virtual Path System**:
-- Agent sees: `/mnt/user-data/{workspace,uploads,outputs}`, `/mnt/skills`
-- Physical: `backend/.deer-flow/users/{user_id}/threads/{thread_id}/user-data/...`, `deer-flow/skills/`
-- Translation: `replace_virtual_path()` / `replace_virtual_paths_in_command()`
-- Detection: `is_local_sandbox()` checks `sandbox_id == "local"`
-
-**Sandbox Tools** (in `packages/harness/deerflow/sandbox/tools.py`):
-- `bash` - Execute commands with path translation and error handling
-- `ls` - Directory listing (tree format, max 2 levels)
-- `read_file` - Read file contents with optional line range
-- `write_file` - Write/append to files, creates directories
-- `str_replace` - Substring replacement (single or all occurrences); same-path serialization is scoped to `(sandbox.id, path)` so isolated sandboxes do not contend on identical virtual paths inside one process
-
-### Subagent System (`packages/harness/deerflow/subagents/`)
-
-**Built-in Agents**: `general-purpose` (all tools except `task`) and `bash` (command specialist)
-**Execution**: Dual thread pool - `_scheduler_pool` (3 workers) + `_execution_pool` (3 workers)
-**Concurrency**: `MAX_CONCURRENT_SUBAGENTS = 3` enforced by `SubagentLimitMiddleware` (truncates excess tool calls in `after_model`), 15-minute timeout
-**Flow**: `task()` tool ? `SubagentExecutor` ? background thread ? poll 5s ? SSE events ? result
-**Events**: `task_started`, `task_running`, `task_completed`/`task_failed`/`task_timed_out`
-
-### Tool System (`packages/harness/deerflow/tools/`)
-
-`get_available_tools(groups, include_mcp, model_name, subagent_enabled)` assembles:
-1. **Config-defined tools** - Resolved from `config.yaml` via `resolve_variable()`
-2. **MCP tools** - From enabled MCP servers (lazy initialized, cached with mtime invalidation)
-3. **Built-in tools**:
-   - `present_files` - Make output files visible to user (only `/mnt/user-data/outputs`)
-   - `ask_clarification` - Request clarification (intercepted by ClarificationMiddleware ? interrupts)
-   - `view_image` - Read image as base64 (added only if model supports vision)
-4. **Subagent tool** (if enabled):
-   - `task` - Delegate to subagent (description, prompt, subagent_type, max_turns)
-
-**Community tools** (`packages/harness/deerflow/community/`):
-- `tavily/` - Web search (5 results default) and web fetch (4KB limit)
-- `jina_ai/` - Web fetch via Jina reader API with readability extraction
-- `firecrawl/` - Web scraping via Firecrawl API
-
-**ACP agent tools**:
-- `invoke_acp_agent` - Invokes external ACP-compatible agents from `config.yaml`
-- ACP launchers must be real ACP adapters. The standard `codex` CLI is not ACP-compatible by itself; configure a wrapper such as `npx -y @zed-industries/codex-acp` or an installed `codex-acp` binary
-- Missing ACP executables now return an actionable error message instead of a raw `[Errno 2]`
-- Each ACP agent uses a per-thread workspace at `{base_dir}/users/{user_id}/threads/{thread_id}/acp-workspace/`. The workspace is accessible to the lead agent via the virtual path `/mnt/acp-workspace/` (read-only). In docker sandbox mode, the directory is volume-mounted into the container at `/mnt/acp-workspace` (read-only); in local sandbox mode, path translation is handled by `tools.py`
-- `image_search/` - Image search via DuckDuckGo
-
-### MCP System (`packages/harness/deerflow/mcp/`)
-
-- Uses `langchain-mcp-adapters` `MultiServerMCPClient` for multi-server management
-- **Lazy initialization**: Tools loaded on first use via `get_cached_mcp_tools()`
-- **Cache invalidation**: Detects config file changes via mtime comparison
-- **Transports**: stdio (command-based), SSE, HTTP
-- **OAuth (HTTP/SSE)**: Supports token endpoint flows (`client_credentials`, `refresh_token`) with automatic token refresh + Authorization header injection
-- **Runtime updates**: Gateway API saves to extensions_config.json; LangGraph detects via mtime
-
-### Skills System (`packages/harness/deerflow/skills/`)
-
-- **Location**: `deer-flow/skills/{public,custom}/`
-- **Format**: Directory with `SKILL.md` (YAML frontmatter: name, description, license, allowed-tools)
-- **Loading**: `load_skills()` recursively scans `skills/{public,custom}` for `SKILL.md`, parses metadata, and reads enabled state from extensions_config.json
-- **Injection**: Enabled skills listed in agent system prompt with container paths
-- **Installation**: `POST /api/skills/install` extracts .skill ZIP archive to custom/ directory
-- **Chinese name (`name_zh`)**: a Chinese display name stored in the **`skills` DB table** (`SkillRow.name_zh`, `String(128)`, migration `20260609_01`) ? never written to the on-disk SKILL.md, so it works for built-in skills too. Threaded through `SkillStore` create/ensure_legacy/update/update_admin exactly like `detail`; surfaced on `SkillResponse.name_zh` from the ownership record in `_skill_to_response`. Authorization mirrors `detail`: custom skill ? owner or admin; built-in/public ? admin only (row materialized on demand via `ensure_legacy`). Endpoints (in `app/gateway/routers/skills.py`): `PUT /api/skills/{name}/name-zh` (set/clear), `POST /api/skills/{name}/name-zh/generate` (LLM generates one), `POST /api/skills/name-zh/generate-missing` (????: **admin-only** bulk-generate for every skill that lacks one, bounded-concurrency `asyncio.gather`). **Optional model selection**: both generate endpoints accept an optional JSON body `{model_name}` ? an explicit name wins, else the curator model setting, else the main model (`_resolve_skill_name_zh_model_name`). **Model-error surfacing**: `_generate_skill_name_zh` raises `SkillNameZhModelError` when the model itself fails (build/timeout/API error) ? distinct from the model returning an unusable name (? `None`). The single endpoint maps that to `502 ?????,?????????:?`; the bulk endpoint fails fast on a model-build error (502) and otherwise carries a representative `model_error` string in `SkillNameZhBulkResponse` so the UI can say "N ?????????". The frontend (`skill-settings-page.tsx` list + `skill-detail-dialog.tsx` editor, plus the agent skill pickers/badges in `NewAgentPage.tsx` + `agent-card.tsx`) shows `name_zh || name`, offers per-skill "AI ??" plus an **admin-only** header "???????" button, and a compact model dropdown (`components/workspace/model-name-select.tsx`, default = configured model) beside each action.
-
-#### Skill Curator (`packages/harness/deerflow/skills/curator.py`)
-
-Background skill lifecycle manager. The Gateway `lifespan` starts a resident
-scheduler task (`_curator_scheduler_loop` in `app/gateway/app.py`) that ? after a
-5-minute startup delay ? calls `maybe_run_curator()` every hour. A cross-process
-file lock (`.curator_scheduler.lock`) elects a single leader worker so the
-scheduler runs once per deployment. Per run:
-
-- **Master switch** ? `skill_evolution.enabled` (default **`false`**) gates the
-  whole feature; `curator.enabled` (default **`false`**) gates the curator
-  specifically. Both must be `true` for the curator to act. `should_run_now()`
-  and `_curator_thread_target()` both enforce this ? so manual `POST
-  /api/curator/run` also respects the master switch. Config is
-  `config.yaml.skill_evolution` deep-merged with per-user
-  `users/{user_id}/skill_evolution.json`.
-- **Per-user run lock** ? `.curator_state` carries `running` / `started_at`;
-  `_try_acquire_run_lock()` prevents the same user's curator from running
-  concurrently (scheduler vs. manual trigger, or double manual clicks). A lock
-  whose `started_at` is older than `_CURATOR_STALE_LOCK_HOURS` (6h) is treated as
-  process-crash residue and can be taken over.
-- **Concurrency cap** ? `maybe_run_curator()` collects all due users and hands
-  them to `_dispatch_curator_runs()`, which runs them through a
-  `ThreadPoolExecutor` capped at `_CURATOR_MAX_CONCURRENCY` (5) ? avoids a
-  thundering herd of LLM calls when many users come due at once (e.g. first
-  deploy).
-- **Failure back-off** ? `_release_run_lock()` always stamps `last_attempt_at`.
-  When a first run keeps crashing (`last_run_at` never set), `should_run_now()`
-  retries on `_CURATOR_RETRY_HOURS` (6h) instead of every hourly tick.
-- **`should_run_now()`** is the real gate: the hourly tick only checks "is the
-  per-user `interval_hours` (default 168h) elapsed". Timestamp parsing tolerates
-  naive (tz-less) strings ? a corrupt `last_run_at` triggers a re-run rather than
-  a permanent deadlock.
-- **Published-skill filter** ? `SkillStore.list_published_names()` is a batch
-  query; the curator fetches the published-name set once per run instead of one
-  DB round-trip per candidate skill.
-- **LLM review timeout** ? `_run_llm_review()` runs `agent.invoke()` in a worker
-  thread bounded by `curator.review_timeout_seconds` (default 600s). The model's
-  per-request `request_timeout` can't bound a multi-turn React agent loop, so on
-  timeout the run is abandoned (a warning is written to `curator_log`, visible in
-  the Web UI) and the curator finishes cleanly + releases the run lock. The
-  worker context is copied via `contextvars.copy_context()` so tool calls still
-  resolve the correct `user_id`.
-- **Review model** ? `_run_llm_review()` uses `curator.model_name` (a
-  `config.yaml` `models[]` entry name) if set, else the primary model; it is
-  editable from the Web UI evolution panel. The skill security scanner
-  (`scan_skill_content` in `security_scanner.py`) reuses the same
-  `skill_evolution.curator.model_name` setting ? review and moderation share one
-  model. There is no separate moderation-model config field.
-
-### Model Factory (`packages/harness/deerflow/models/factory.py`)
-
-- `create_chat_model(name, thinking_enabled)` instantiates LLM from config via reflection
-- Supports `thinking_enabled` flag with per-model `when_thinking_enabled` overrides
-- Supports vLLM-style thinking toggles via `when_thinking_enabled.extra_body.chat_template_kwargs.enable_thinking` for Qwen reasoning models, while normalizing legacy `thinking` configs for backward compatibility
-- Supports `supports_vision` flag for image understanding models
-- Config values starting with `$` resolved as environment variables
-- Missing provider modules surface actionable install hints from reflection resolvers (for example `uv add langchain-google-genai`)
-
-### vLLM Provider (`packages/harness/deerflow/models/vllm_provider.py`)
-
-- `VllmChatModel` subclasses `langchain_openai:ChatOpenAI` for vLLM 0.19.0 OpenAI-compatible endpoints
-- Preserves vLLM's non-standard assistant `reasoning` field on full responses, streaming deltas, and follow-up tool-call turns
-- Designed for configs that enable thinking through `extra_body.chat_template_kwargs.enable_thinking` on vLLM 0.19.0 Qwen reasoning models, while accepting the older `thinking` alias
-
-### IM Channels System (`app/channels/`)
-
-Bridges external messaging platforms (Feishu, Slack, Telegram, DingTalk) to the DeerFlow agent via the LangGraph Server.
-
-
-**Architecture**: Channels communicate with Gateway through the `langgraph-sdk` HTTP client (same as the frontend), ensuring threads are created and managed server-side. The internal SDK client injects process-local internal auth plus a matching CSRF cookie/header pair so Gateway accepts state-changing thread/run requests from channel workers without relying on browser session cookies.
-
-**Components**:
-- `message_bus.py` - Async pub/sub hub (`InboundMessage` ? queue ? dispatcher; `OutboundMessage` ? callbacks ? channels)
-- `store.py` - JSON-file persistence mapping `channel_name:chat_id[:topic_id]` ? `thread_id` (keys are `channel:chat` for root conversations and `channel:chat:topic` for threaded conversations)
-- `manager.py` - Core dispatcher: creates threads via `client.threads.create()`, routes commands, keeps Slack/Telegram on `client.runs.wait()`, and uses `client.runs.stream(["messages-tuple", "values"])` for Feishu incremental outbound updates
-- `base.py` - Abstract `Channel` base class (start/stop/send lifecycle)
-- `service.py` - Manages lifecycle of all configured channels from `config.yaml`
-- `slack.py` / `feishu.py` / `telegram.py` / `dingtalk.py` - Platform-specific implementations (`feishu.py` tracks the running card `message_id` in memory and patches the same card in place; `dingtalk.py` optionally uses AI Card streaming for in-place updates when `card_template_id` is configured)
-
-**Message Flow**:
-1. External platform -> Channel impl -> `MessageBus.publish_inbound()`
-2. `ChannelManager._dispatch_loop()` consumes from queue
-3. For chat: look up/create thread through Gateway's LangGraph-compatible API
-4. Feishu chat: `runs.stream()` ? accumulate AI text ? publish multiple outbound updates (`is_final=False`) ? publish final outbound (`is_final=True`)
-5. Slack/Telegram chat: `runs.wait()` ? extract final response ? publish outbound
-6. Feishu channel sends one running reply card up front, then patches the same card for each outbound update (card JSON sets `config.update_multi=true` for Feishu's patch API requirement)
-7. DingTalk AI Card mode (when `card_template_id` configured): `runs.stream()` ? create card with initial text ? stream updates via `PUT /v1.0/card/streaming` ? finalize on `is_final=True`. Falls back to `sampleMarkdown` if card creation or streaming fails
-8. For commands (`/new`, `/status`, `/models`, `/memory`, `/help`): handle locally or query Gateway API
-9. Outbound ? channel callbacks ? platform reply
-
-**Configuration** (`config.yaml` -> `channels`):
-- `langgraph_url` - LangGraph-compatible Gateway API base URL (default: `http://localhost:8001/api`)
-- `gateway_url` - Gateway API URL for auxiliary commands (default: `http://localhost:8001`)
-- In Docker Compose, IM channels run inside the `gateway` container, so `localhost` points back to that container. Use `http://gateway:8001/api` for `langgraph_url` and `http://gateway:8001` for `gateway_url`, or set `DEER_FLOW_CHANNELS_LANGGRAPH_URL` / `DEER_FLOW_CHANNELS_GATEWAY_URL`.
-- Per-channel configs: `feishu` (app_id, app_secret), `slack` (bot_token, app_token), `telegram` (bot_token), `dingtalk` (client_id, client_secret, optional `card_template_id` for AI Card streaming)
-
-
-### Memory System ? Memory V2 (`packages/harness/deerflow/agents/memory/`)
-
-Pluggable memory architecture (Hermes-style). See `docs/MEMORY_V2_DESIGN_ZH.md`
-and `docs/HINDSIGHT_SETUP_ZH.md` for the full design.
-
-**Components**:
-- `provider.py` - `MemoryProvider` abstract base (12 lifecycle/hook methods)
-- `manager.py` - `MemoryManager` orchestrates builtin + ?1 external provider; `build_memory_manager()` is the construction entry point
-- `providers/builtin.py` - `BuiltinFileProvider`: local USER.md / MEMORY.md, ? delimiter, char quotas, file locking, atomic writes, frozen snapshot
-- `providers/hindsight.py` - `HindsightProvider`: HTTP client to a self-hosted Hindsight Docker instance, dual-bank routing
-- `security.py` - injection/exfiltration scan + invisible-unicode block for memory content
-- `message_processing.py` - `filter_messages_for_memory` (last-turn extraction for Hindsight retain)
-
-**Two stores (binary classification)**:
-- `USER.md` at `{base_dir}/users/{user_id}/USER.md` ? user profile, shared across agents
-- `MEMORY.md` at `{base_dir}/users/{user_id}/agents/{agent_id}/MEMORY.md` ? agent-private working notes
-- `user_id` resolved via `get_effective_user_id()`; `"default"` in no-auth mode
-
-**Write paths**:
-- LLM actively calls the `memory` tool (`tools/builtins/memory_tool.py`) ? writes USER.md/MEMORY.md, mirrors to Hindsight
-- LLM calls `hindsight_recall`/`reflect`/`retain` tools (`tools/builtins/hindsight_tools.py`) ? only in `tools`/`hybrid` memory_mode
-- Hindsight `auto_retain` ? every turn auto-stored to the work bank (via `MemoryMiddleware.after_agent`)
-
-**Injection**:
-- builtin frozen snapshot ? injected into the system prompt via `lead_agent/prompt.py` `_get_memory_context`
-- Hindsight dynamic recall ? injected per model call via `MemoryMiddleware.wrap_model_call` (transient ? not persisted to state, so it never leaks to the frontend)
-
-**Hindsight dual-bank isolation** (B-scheme):
-- global user bank `deerflow-user-{user_id}` ? cross-agent
-- agent work bank `deerflow-work-{user_id}-{agent_id}` ? per-agent
-- `hindsight-client` is an optional dependency; when missing, `HindsightProvider.is_available()` returns False and the system falls back to builtin-only
-
-**Migration**: `PYTHONPATH=. python scripts/migrate_memory_to_v2.py [--all|--user-id X] [--dry-run]` converts legacy `memory.json` into USER.md/MEMORY.md.
-
-**Configuration** (`config.yaml` ? `memory`, nested): `enabled`, `provider` (builtin/hindsight),
-`builtin.{memory_char_limit,user_char_limit,...}`, `hindsight.{mode,api_url,api_key,bank_id_template,work_bank_id_template,memory_mode,...}`,
-`injection.{enabled,max_tokens,include_builtin,include_hindsight,context_tag}`, `security.{scan_content,block_invisible_unicode}`.
-Per-user runtime overrides (whitelisted fields) live at `{base_dir}/users/{user_id}/memory_config.json` ? see `get_effective_memory_config()`.
-
-**Gateway API** ? see the Memory router below (`/api/memory/v2*`).
-
-### Reflection System (`packages/harness/deerflow/reflection/`)
-
-- `resolve_variable(path)` - Import module and return variable (e.g., `module.path:variable_name`)
-- `resolve_class(path, base_class)` - Import and validate class against base class
-
-### Config Schema
-
-**`config.yaml`** key sections:
-- `models[]` - LLM configs with `use` class path, `supports_thinking`, `supports_vision`, provider-specific fields
-- vLLM reasoning models should use `deerflow.models.vllm_provider:VllmChatModel`; for Qwen-style parsers prefer `when_thinking_enabled.extra_body.chat_template_kwargs.enable_thinking`, and DeerFlow will also normalize the older `thinking` alias
-- `tools[]` - Tool configs with `use` variable path and `group`
-- `tool_groups[]` - Logical groupings for tools
-- `sandbox.use` - Sandbox provider class path
-- `skills.path` / `skills.container_path` - Host and container paths to skills directory
-- `title` - Auto-title generation (enabled, max_words, max_chars, prompt_template)
-- `summarization` - Context summarization (enabled, trigger conditions, keep policy)
-- `subagents.enabled` - Master switch for subagent delegation
-- `memory` - Memory system (enabled, storage_path, debounce_seconds, model_name, max_facts, fact_confidence_threshold, injection_enabled, max_injection_tokens)
-- `auth_login` - Custom login modes layered on the base auth (`deerflow/config/login_config.py`):
-  - `username_login.{require_password, password}` - Password gate for `POST /api/v1/auth/login/username`. When `require_password=true`, the (otherwise passwordless) username login additionally requires `?password=<shared-secret>`; empty `password` falls back to `123ewq`. Gate only ? it does not change which account logs in.
-  - `token_login.{enabled, token_info_url, service_authorization, timeout_seconds, verify_ssl, ca_cert_path}` - Token-exchange login (`POST /api/v1/auth/login/token`). The gateway POSTs to `token_info_url` with the caller-supplied token placed directly in the `Authorization: Bearer <token>` header (no cookie; the static `service_authorization` field is deprecated/unused, kept only for backward compat), reads `data.username` from a `{"state":"200",...}` body, then signs the matching local user in (auto-registering if absent). `token_info_url` is HTTPS ? `verify_ssl: false` skips cert verification for self-signed/internal UAT hosts, or set `ca_cert_path` to a PEM bundle (takes precedence, keeps verification on). See [docs/AUTH_LOGIN.md](docs/AUTH_LOGIN.md) for the full request/response contract.
-
-Both endpoints are whitelisted in `auth_middleware._PUBLIC_EXACT_PATHS`. The shared `_get_or_create_user_by_username()` helper (in `routers/auth.py`) backs both `/login/username` and `/login/token`.
-
-- `log_system` - Admin-only external log-system entry (`deerflow/config/log_system_config.py`): `{enabled, url, label}`. Surfaced read-only by `GET /api/system-settings/log-system` (admin only); the frontend renders it as a "????" item in the bottom-right settings menu (`workspace-nav-menu.tsx`) that opens `url` in a new tab. Disabled by default.
-
-**`extensions_config.json`**:
-- `mcpServers` - Map of server name ? config (enabled, type, command, args, env, url, headers, oauth, description)
-- `skills` - Map of skill name ? state (enabled)
-
-Both can be modified at runtime via Gateway API endpoints or `DeerFlowClient` methods.
-
-### Embedded Client (`packages/harness/deerflow/client.py`)
-
-`DeerFlowClient` provides direct in-process access to all DeerFlow capabilities without HTTP services. All return types align with the Gateway API response schemas, so consumer code works identically in HTTP and embedded modes.
-
-**Architecture**: Imports the same `deerflow` modules that Gateway API uses. Shares the same config files and data directories. No FastAPI dependency.
-
-**Agent Conversation**:
-- `chat(message, thread_id)` ? synchronous, accumulates streaming deltas per message-id and returns the final AI text
-- `stream(message, thread_id)` ? subscribes to LangGraph `stream_mode=["values", "messages", "custom"]` and yields `StreamEvent`:
-  - `"values"` ? full state snapshot (title, messages, artifacts); AI text already delivered via `messages` mode is **not** re-synthesized here to avoid duplicate deliveries
-  - `"messages-tuple"` ? per-chunk update: for AI text this is a **delta** (concat per `id` to rebuild the full message); tool calls and tool results are emitted once each
-  - `"custom"` ? forwarded from `StreamWriter`
-  - `"end"` ? stream finished (carries cumulative `usage` counted once per message id)
-- Agent created lazily via `create_agent()` + `_build_middlewares()`, same as `make_lead_agent`
-- Supports `checkpointer` parameter for state persistence across turns
-- `reset_agent()` forces agent recreation (e.g. after memory or skill changes)
-- See [docs/STREAMING.md](docs/STREAMING.md) for the full design: why Gateway and DeerFlowClient are parallel paths, LangGraph's `stream_mode` semantics, the per-id dedup invariants, and regression testing strategy
-
-**Gateway Equivalent Methods** (replaces Gateway API):
-
-| Category | Methods | Return format |
-|----------|---------|---------------|
-| Models | `list_models()`, `get_model(name)` | `{"models": [...]}`, `{name, display_name, ...}` |
-| MCP | `get_mcp_config()`, `update_mcp_config(servers)` | `{"mcp_servers": {...}}` |
-| Skills | `list_skills()`, `get_skill(name)`, `update_skill(name, enabled)`, `install_skill(path)` | `{"skills": [...]}` |
-| Memory | `get_memory()`, `reload_memory()`, `get_memory_config()`, `get_memory_status()` | dict |
-| Uploads | `upload_files(thread_id, files)`, `list_uploads(thread_id)`, `delete_upload(thread_id, filename)` | `{"success": true, "files": [...]}`, `{"files": [...], "count": N}` |
-| Artifacts | `get_artifact(thread_id, path)` ? `(bytes, mime_type)` | tuple |
-
-**Key difference from Gateway**: Upload accepts local `Path` objects instead of HTTP `UploadFile`, rejects directory paths before copying, and reuses a single worker when document conversion must run inside an active event loop. Artifact returns `(bytes, mime_type)` instead of HTTP Response. The new Gateway-only thread cleanup route deletes `.deer-flow/threads/{thread_id}` after LangGraph thread deletion; there is no matching `DeerFlowClient` method yet. `update_mcp_config()` and `update_skill()` automatically invalidate the cached agent.
-
-**Tests**: `tests/test_client.py` (77 unit tests including `TestGatewayConformance`), `tests/test_client_live.py` (live integration tests, requires config.yaml)
-
-**Gateway Conformance Tests** (`TestGatewayConformance`): Validate that every dict-returning client method conforms to the corresponding Gateway Pydantic response model. Each test parses the client output through the Gateway model ? if Gateway adds a required field that the client doesn't provide, Pydantic raises `ValidationError` and CI catches the drift. Covers: `ModelsListResponse`, `ModelResponse`, `SkillsListResponse`, `SkillResponse`, `SkillInstallResponse`, `McpConfigResponse`, `UploadResponse`, `MemoryConfigResponse`, `MemoryStatusResponse`.
-
-## Development Workflow
-
-### Test-Driven Development (TDD) ? MANDATORY
-
-**Every new feature or bug fix MUST be accompanied by unit tests. No exceptions.**
-
-- Write tests in `backend/tests/` following the existing naming convention `test_<feature>.py`
-- Run the full suite before and after your change: `make test`
-- Tests must pass before a feature is considered complete
-- For lightweight config/utility modules, prefer pure unit tests with no external dependencies
-- If a module causes circular import issues in tests, add a `sys.modules` mock in `tests/conftest.py` (see existing example for `deerflow.subagents.executor`)
-
-```bash
-# Run all tests
-make test
-
-# Run a specific test file
-PYTHONPATH=. uv run pytest tests/test_<feature>.py -v
-```
-
-### Running the Full Application
-
-From the **project root** directory:
-```bash
-make dev
-```
-
-This starts all services and makes the application available at `http://localhost:2026`.
-
-**All startup modes:**
-
-| | **Local Foreground** | **Local Daemon** | **Docker Dev** | **Docker Prod** |
-|---|---|---|---|---|
-| **Dev** | `./scripts/serve.sh --dev`<br/>`make dev` | `./scripts/serve.sh --dev --daemon`<br/>`make dev-daemon` | `./scripts/docker.sh start`<br/>`make docker-start` | ? |
-| **Prod** | `./scripts/serve.sh --prod`<br/>`make start` | `./scripts/serve.sh --prod --daemon`<br/>`make start-daemon` | ? | `./scripts/deploy.sh`<br/>`make up` |
-
-| Action | Local | Docker Dev | Docker Prod |
-|---|---|---|---|
-| **Stop** | `./scripts/serve.sh --stop`<br/>`make stop` | `./scripts/docker.sh stop`<br/>`make docker-stop` | `./scripts/deploy.sh down`<br/>`make down` |
-| **Restart** | `./scripts/serve.sh --restart [flags]` | `./scripts/docker.sh restart` | ? |
-
-**Nginx routing**:
-- `/api/langgraph/*` ? Gateway embedded runtime (8001), rewritten to `/api/*`
-- `/api/*` (other) ? Gateway API (8001)
-- `/` (non-API) ? Frontend (3000)
-
-### Running Backend Services Separately
-
-From the **backend** directory:
-
-```bash
-# Gateway API
-make gateway
-```
-
-Direct access (without nginx):
-- Gateway: `http://localhost:8001`
-
-### Frontend Configuration
-
-The frontend uses environment variables to connect to backend services:
-- `NEXT_PUBLIC_LANGGRAPH_BASE_URL` - Defaults to `/api/langgraph` (through nginx)
-- `NEXT_PUBLIC_BACKEND_BASE_URL` - Defaults to empty string (through nginx)
-
-When using `make dev` from root, the frontend automatically connects through nginx.
-
-## Key Features
-
-### File Upload
-
-Multi-file upload with automatic document conversion:
-- Endpoint: `POST /api/threads/{thread_id}/uploads`
-- Supports: PDF, PPT, Excel, Word documents (converted via `markitdown`)
-- Rejects directory inputs before copying so uploads stay all-or-nothing
-- Reuses one conversion worker per request when called from an active event loop
-- Files stored in thread-isolated directories
-- Agent receives uploaded file list via `UploadsMiddleware`
-
-See [docs/FILE_UPLOAD.md](docs/FILE_UPLOAD.md) for details.
-
-### Plan Mode
-
-TodoList middleware for complex multi-step tasks:
-- Controlled via runtime config: `config.configurable.is_plan_mode = True`
-- Provides `write_todos` tool for task tracking
-- One task in_progress at a time, real-time updates
-
-See [docs/plan_mode_usage.md](docs/plan_mode_usage.md) for details.
-
-### Context Summarization
-
-Automatic conversation summarization when approaching token limits:
-- Configured in `config.yaml` under `summarization` key
-- Trigger types: tokens, messages, or fraction of max input
-- Keeps recent messages while summarizing older ones
-
-See [docs/summarization.md](docs/summarization.md) for details.
-
-### Vision Support / Image Recognition (??)
-
-Two paths, selected by `config.yaml` ? `vision`:
-
-**Dedicated vision model (delegated recognition)** ? set `vision.model_name` to a model from `models[]`. The `view_image` tool sends the image to that model and returns its **text description** as the tool result, so a text-only main chat model can still "see" images. `view_image` accepts an optional `question` argument for targeted queries; the default prompt is `vision.prompt` (overridable), and `vision.max_tokens` caps the recognition output. Resolution/gating lives in `deerflow/models/vision.py` (`resolve_vision_model_name`, `vision_enabled`, `is_delegated_vision`, `recognize_image`).
-
-**Native vision (fallback)** ? when `vision.model_name` is unset, behavior falls back to the main chat model, which only works if it has `supports_vision: true`:
-- `ViewImageMiddleware` injects base64 image data into the conversation before the LLM call
-- `view_image` stores the image in `state.viewed_images` for that middleware
-
-Both the `view_image` tool and `ViewImageMiddleware` are registered whenever `vision_enabled(model_name)` is true (dedicated model configured **or** main model supports vision). Config schema: `deerflow/config/vision_config.py` (`VisionConfig`); tests: `tests/test_vision.py`.
-
-### Upload symlink-safety (cross-platform)
-
-`deerflow/uploads/manager.py:open_upload_file_no_symlink` writes uploads without following a pre-existing destination symlink. `O_NOFOLLOW` is applied where available (POSIX) but is **optional** ? Windows lacks `os.O_NOFOLLOW`/`O_NONBLOCK`, so requiring it previously made every Windows upload fail with `UnsafeUploadPathError`. Pre-existing symlinks are still rejected via `lstat`, and the opened fd is re-checked with `fstat`. `O_BINARY` is added on Windows. Regression: `tests/test_uploads_windows_safe.py`.
-
-## Code Style
-
-- Uses `ruff` for linting and formatting
-- Line length: 240 characters
-- Python 3.12+ with type hints
-- Double quotes, space indentation
-
-## Documentation
-
-See `docs/` directory for detailed documentation:
-- [CONFIGURATION.md](docs/CONFIGURATION.md) - Configuration options
-- [ARCHITECTURE.md](docs/ARCHITECTURE.md) - Architecture details
-- [API.md](docs/API.md) - API reference
-- [SETUP.md](docs/SETUP.md) - Setup guide
-- [FILE_UPLOAD.md](docs/FILE_UPLOAD.md) - File upload feature
-- [PATH_EXAMPLES.md](docs/PATH_EXAMPLES.md) - Path types and usage
-- [summarization.md](docs/summarization.md) - Context summarization
-- [plan_mode_usage.md](docs/plan_mode_usage.md) - Plan mode with TodoList
-- [BACKEND_RUN_STOP_ZH.md](docs/BACKEND_RUN_STOP_ZH.md) - Windows-first backend start/stop and troubleshooting guide
+# CLAUDE.md
+??
+This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
+
+## Project Overview
+
+DeerFlow is a LangGraph-based AI super agent system with a full-stack architecture. The backend provides a "super agent" with sandbox execution, persistent memory, subagent delegation, and extensible tool integration - all operating in per-thread isolated environments.
+
+**Architecture**:
+- **Gateway API** (port 8001): REST API plus embedded LangGraph-compatible agent runtime
+- **Frontend** (port 3000): Next.js web interface
+- **Nginx** (port 2026): Unified reverse proxy entry point
+- **Provisioner** (port 8002, optional in Docker dev): Started only when sandbox is configured for provisioner/Kubernetes mode
+
+**Runtime**:
+- `make dev`, Docker dev, and production all run the agent runtime in Gateway via `RunManager` + `run_agent()` + `StreamBridge` (`packages/harness/deerflow/runtime/`). Nginx exposes that runtime at `/api/langgraph/*` and rewrites it to Gateway's native `/api/*` routers.
+
+**Project Structure**:
+```
+deer-flow/
+??? Makefile                    # Root commands (check, install, dev, stop)
+??? config.yaml                 # Main application configuration
+??? extensions_config.json      # MCP servers and skills configuration
+??? backend/                    # Backend application (this directory)
+?   ??? Makefile               # Backend-only commands (dev, gateway, lint)
+?   ??? langgraph.json         # LangGraph Studio graph configuration
+?   ??? packages/
+?   ?   ??? harness/           # deerflow-harness package (import: deerflow.*)
+?   ?       ??? pyproject.toml
+?   ?       ??? deerflow/
+?   ?           ??? agents/            # LangGraph agent system
+?   ?           ?   ??? lead_agent/    # Main agent (factory + system prompt)
+?   ?           ?   ??? middlewares/   # 10 middleware components
+?   ?           ?   ??? memory/        # Memory V2: provider/manager/builtin/hindsight
+?   ?           ?   ??? thread_state.py # ThreadState schema
+?   ?           ??? sandbox/           # Sandbox execution system
+?   ?           ?   ??? local/         # Local filesystem provider
+?   ?           ?   ??? sandbox.py     # Abstract Sandbox interface
+?   ?           ?   ??? tools.py       # bash, ls, read/write/str_replace
+?   ?           ?   ??? middleware.py  # Sandbox lifecycle management
+?   ?           ??? subagents/         # Subagent delegation system
+?   ?           ?   ??? builtins/      # general-purpose, bash agents
+?   ?           ?   ??? executor.py    # Background execution engine
+?   ?           ?   ??? registry.py    # Agent registry
+?   ?           ??? tools/builtins/    # Built-in tools (present_files, ask_clarification, view_image)
+?   ?           ??? mcp/               # MCP integration (tools, cache, client)
+?   ?           ??? models/            # Model factory with thinking/vision support
+?   ?           ??? skills/            # Skills discovery, loading, parsing
+?   ?           ??? config/            # Configuration system (app, model, sandbox, tool, etc.)
+?   ?           ??? community/         # Community tools (tavily, jina_ai, firecrawl, image_search, aio_sandbox)
+?   ?           ??? reflection/        # Dynamic module loading (resolve_variable, resolve_class)
+?   ?           ??? utils/             # Utilities (network, readability)
+?   ?           ??? client.py          # Embedded Python client (DeerFlowClient)
+?   ??? app/                   # Application layer (import: app.*)
+?   ?   ??? gateway/           # FastAPI Gateway API
+?   ?   ?   ??? app.py         # FastAPI application
+?   ?   ?   ??? routers/       # FastAPI route modules (models, mcp, memory, skills, uploads, threads, artifacts, agents, suggestions, channels)
+?   ?   ??? channels/          # IM platform integrations
+?   ??? tests/                 # Test suite
+?   ??? docs/                  # Documentation
+??? frontend/                   # Next.js frontend application
+??? skills/                     # Agent skills directory
+    ??? public/                # Public skills (committed)
+    ??? custom/                # Custom skills (gitignored)
+```
+
+## Important Development Guidelines
+
+### Documentation Update Policy
+**CRITICAL: Always update README.md and CLAUDE.md after every code change**
+
+When making code changes, you MUST update the relevant documentation:
+- Update `README.md` for user-facing changes (features, setup, usage instructions)
+- Update `CLAUDE.md` for development changes (architecture, commands, workflows, internal systems)
+- Keep documentation synchronized with the codebase at all times
+- Ensure accuracy and timeliness of all documentation
+
+## Commands
+
+**Root directory** (for full application):
+```bash
+make check      # Check system requirements
+make install    # Install all dependencies (frontend + backend)
+make dev        # Start all services (Gateway + Frontend + Nginx), with config.yaml preflight
+make start      # Start production services locally
+make stop       # Stop all services
+```
+
+**Backend directory** (for backend development only):
+```bash
+make install    # Install backend dependencies
+make dev        # Run Gateway API with reload (port 8001)
+make gateway    # Run Gateway API only (port 8001)
+make test       # Run all backend tests
+make lint       # Lint with ruff
+make format     # Format code with ruff
+```
+
+Regression tests related to Docker/provisioner behavior:
+- `tests/test_docker_sandbox_mode_detection.py` (mode detection from `config.yaml`)
+- `tests/test_provisioner_kubeconfig.py` (kubeconfig file/directory handling)
+
+Boundary check (harness ? app import firewall):
+- `tests/test_harness_boundary.py` ? ensures `packages/harness/deerflow/` never imports from `app.*`
+
+CI runs these regression tests for every pull request via [.github/workflows/backend-unit-tests.yml](../.github/workflows/backend-unit-tests.yml).
+
+## Architecture
+
+### Harness / App Split
+
+The backend is split into two layers with a strict dependency direction:
+
+- **Harness** (`packages/harness/deerflow/`): Publishable agent framework package (`deerflow-harness`). Import prefix: `deerflow.*`. Contains agent orchestration, tools, sandbox, models, MCP, skills, config ? everything needed to build and run agents.
+- **App** (`app/`): Unpublished application code. Import prefix: `app.*`. Contains the FastAPI Gateway API and IM channel integrations (Feishu, Slack, Telegram, DingTalk).
+
+**Dependency rule**: App imports deerflow, but deerflow never imports app. This boundary is enforced by `tests/test_harness_boundary.py` which runs in CI.
+
+**Import conventions**:
+```python
+# Harness internal
+from deerflow.agents import make_lead_agent
+from deerflow.models import create_chat_model
+
+# App internal
+from app.gateway.app import app
+from app.channels.service import start_channel_service
+
+# App ? Harness (allowed)
+from deerflow.config import get_app_config
+
+# Harness ? App (FORBIDDEN ? enforced by test_harness_boundary.py)
+# from app.gateway.routers.uploads import ...  # ? will fail CI
+```
+
+### Agent System
+
+**Lead Agent** (`packages/harness/deerflow/agents/lead_agent/agent.py`):
+- Entry point: `make_lead_agent(config: RunnableConfig)` registered in `langgraph.json`
+- Dynamic model selection via `create_chat_model()` with thinking/vision support
+- Tools loaded via `get_available_tools()` - combines sandbox, built-in, MCP, community, and subagent tools
+- System prompt generated by `apply_prompt_template()` with skills, memory, and subagent instructions
+
+**ThreadState** (`packages/harness/deerflow/agents/thread_state.py`):
+- Extends `AgentState` with: `sandbox`, `thread_data`, `title`, `artifacts`, `todos`, `uploaded_files`, `viewed_images`
+- Uses custom reducers: `merge_artifacts` (deduplicate), `merge_viewed_images` (merge/clear)
+
+**Runtime Configuration** (via `config.configurable`):
+- `thinking_enabled` - Enable model's extended thinking
+- `model_name` - Select specific LLM model
+- `is_plan_mode` - Enable TodoList middleware
+- `subagent_enabled` - Enable task delegation tool
+
+### Middleware Chain
+
+Lead-agent middlewares are assembled in strict append order across `packages/harness/deerflow/agents/middlewares/tool_error_handling_middleware.py` (`build_lead_runtime_middlewares`) and `packages/harness/deerflow/agents/lead_agent/agent.py` (`_build_middlewares`):
+
+1. **ThreadDataMiddleware** - Creates per-thread directories under the user's isolation scope (`backend/.deer-flow/users/{user_id}/threads/{thread_id}/user-data/{workspace,uploads,outputs}`); resolves `user_id` via `get_effective_user_id()` (falls back to `"default"` in no-auth mode); Web UI thread deletion now follows LangGraph thread removal with Gateway cleanup of the local thread directory
+2. **UploadsMiddleware** - Tracks and injects newly uploaded files into conversation
+3. **SandboxMiddleware** - Acquires sandbox, stores `sandbox_id` in state
+4. **DanglingToolCallMiddleware** - Injects placeholder ToolMessages for AIMessage tool_calls that lack responses (e.g., due to user interruption), including raw provider tool-call payloads preserved only in `additional_kwargs["tool_calls"]`
+5. **LLMErrorHandlingMiddleware** - Normalizes provider/model invocation failures into recoverable assistant-facing errors before later middleware/tool stages run
+6. **GuardrailMiddleware** - Pre-tool-call authorization via pluggable `GuardrailProvider` protocol (optional, if `guardrails.enabled` in config). Evaluates each tool call and returns error ToolMessage on deny. Three provider options: built-in `AllowlistProvider` (zero deps), OAP policy providers (e.g. `aport-agent-guardrails`), or custom providers. See [docs/GUARDRAILS.md](docs/GUARDRAILS.md) for setup, usage, and how to implement a provider.
+7. **SandboxAuditMiddleware** - Audits sandboxed shell/file operations for security logging before tool execution continues
+8. **ToolErrorHandlingMiddleware** - Converts tool exceptions into error `ToolMessage`s so the run can continue instead of aborting
+9. **SummarizationMiddleware** (`DeerFlowSummarizationMiddleware`) - Context reduction when approaching token limits (optional, if enabled). **Non-destructive**: overrides the stock `before_model` (which persists a `RemoveMessage(REMOVE_ALL_MESSAGES)`) to a no-op and instead compresses **transiently** in `wrap_model_call` ? the model request is replaced with `[summary, *recent]` while the full original conversation stays in state (so the frontend keeps rendering the user's real Q&A). The `summary` message is `hide_from_ui` and never persisted; summaries are cached **per-thread in a process-wide, thread-safe LRU** (`_GLOBAL_SUMMARY_CACHE`, bounded to 512 threads, sticky boundary). The cache is module-level — **not** on the middleware instance — because `make_lead_agent` rebuilds a fresh middleware on every run; an instance cache would be cold each turn and force a blocking summary LLM call on every turn of a long conversation. Keyed by `thread_id` (resolved like `ThreadDataMiddleware`), so a summary computed on one turn is reused on the next, paying for an LLM call only on the first threshold crossing or a genuine re-compaction. **Threshold-first gating**: `_plan_compression` returns passthrough (no compression at all) whenever the *full* current conversation is still under the trigger — the cache/summary substitution only engages once the real context crosses the configured trigger. **Stickiness headroom** (`_apply_stickiness_headroom`, `_KEEP_HEADROOM_FRACTION=0.75`): after the `keep` policy picks a cutoff, the kept suffix is forced below ~75% of *every* trigger before summarizing. Without it a `keep` in *messages* (e.g. 20) can itself exceed a *token* trigger (e.g. 60000) when recent messages are large (tool outputs / pasted docs), leaving the "compressed" context still over the trigger so it re-summarizes on every follow-up question; the headroom makes the sticky boundary actually stick (re-compaction only after genuinely new content). When the genuine (blocking) summarize path runs, it emits a `{"type":"context_compacting","message":"正在压缩上下文…"}` event on LangGraph's `custom` stream channel (the cheap reuse path stays silent); the chat frontend's `onCustomEvent` renders it as a transient toast. Tests: `tests/test_summarization_middleware.py`
+10. **TodoListMiddleware** - Task tracking with `write_todos` tool (optional, if plan_mode)
+11. **TokenUsageMiddleware** - Records token usage metrics when token tracking is enabled (optional)
+12. **TitleMiddleware** - Two-phase auto-title (never shows "untitled"): `before_model` sets an instant question-derived *provisional* title the moment the first user message arrives (no LLM, zero latency), then `after_model` asynchronously generates the final LLM title and replaces it (tracked via the `title_provisional` state flag). Normalizes structured message content before prompting the title model
+13. **MemoryMiddleware** - Memory V2: `wrap_model_call` injects Hindsight recall transiently; `after_agent` auto-retains the turn to Hindsight
+14. **ViewImageMiddleware** - Injects base64 image data before LLM call (conditional on vision support)
+15. **DeferredToolFilterMiddleware** - Hides deferred tool schemas from the bound model until tool search is enabled (optional)
+16. **SubagentLimitMiddleware** - Truncates excess `task` tool calls from model response to enforce `MAX_CONCURRENT_SUBAGENTS` limit (optional, if `subagent_enabled`)
+17. **LoopDetectionMiddleware** - Detects repeated tool-call loops; hard-stop responses clear both structured `tool_calls` and raw provider tool-call metadata before forcing a final text answer
+18. **ClarificationMiddleware** - Intercepts `ask_clarification` tool calls, interrupts via `Command(goto=END)` (must be last)
+8a. **ToolMetricsMiddleware** - Appended immediately after `ToolErrorHandlingMiddleware` in `_build_runtime_middlewares`, so it sits *deepest* in the `wrap_tool_call` chain. Records one `tool_call_metrics` row per tool/skill/MCP invocation ? timing duration and bucketing failures into a coarse `error_category` (rate_limited / ip_blocked / auth_error / timeout / network_error / not_found / empty_result / tool_error). It catches both raised exceptions (before #8 converts them) and *self-handled* errors returned as error `ToolMessage`s (search tools that swallow HTTP 429/403). It also captures the targeted Agent Skill into `tool_call_metrics.skill_name`: the lead agent invokes a skill by `read_file`-ing its `/mnt/skills/<category>/<name>/SKILL.md` path (per the `<skill_system>` prompt block ? there is no dedicated "use skill" tool), so `skill_name` is recovered by scanning every tool call's string arguments for the `skills/{public,custom}/{name}/` path shape (`_SKILL_PATH_RE`), plus the explicit `name` argument of `skill_view`/`skill_manage`/`skill_archive`. `skill_view`-specific success detection handles its swallowed failures (`Skill 'x' not found.` returned as a plain string rather than raised). Regression coverage: `tests/test_tool_metrics_middleware.py`. Writes are best-effort fire-and-forget. **Per-run skill merging**: `SqlToolMetricsStore.record` upserts skill rows by `(run_id, skill_name)` instead of inserting one row per call ? a skill touched by several tool calls / retries inside one run collapses to a single row. Any success in the run wins (a skill that failed then finally succeeded is recorded as one `success`); a skill that only ever failed yields one `error` row whose `error_message` accumulates every attempt's message; durations are summed. Non-skill tool calls, and skill calls with no `run_id`, keep one row each. The read?merge?write is guarded by a per-event-loop `asyncio.Lock`. Surfaced by the admin `GET /api/admin/tool-metrics` router (which has a `kind=skill` filter and a per-skill `/summary` breakdown).
+9. **PromptPrefixMiddleware** - Injects the admin-configured prompt prefix into the model request only. The prefix is tagged onto the first human message's `additional_kwargs['prompt_prefix']` by the Gateway (`app.gateway.services`) and prepended to content here at call time, so the persisted/displayed message stays prefix-free
+10. **SummarizationMiddleware** (`DeerFlowSummarizationMiddleware`) - Context reduction when approaching token limits (optional, if enabled). **Non-destructive**: overrides the stock `before_model` (which persists a `RemoveMessage(REMOVE_ALL_MESSAGES)`) to a no-op and instead compresses **transiently** in `wrap_model_call` ? the model request is replaced with `[summary, *recent]` while the full original conversation stays in state (so the frontend keeps rendering the user's real Q&A). The `summary` message is `hide_from_ui` and never persisted; summaries are cached **per-thread in a process-wide, thread-safe LRU** (`_GLOBAL_SUMMARY_CACHE`, bounded to 512 threads, sticky boundary). The cache is module-level — **not** on the middleware instance — because `make_lead_agent` rebuilds a fresh middleware on every run; an instance cache would be cold each turn and force a blocking summary LLM call on every turn of a long conversation. Keyed by `thread_id` (resolved like `ThreadDataMiddleware`), so a summary computed on one turn is reused on the next, paying for an LLM call only on the first threshold crossing or a genuine re-compaction. **Threshold-first gating**: `_plan_compression` returns passthrough (no compression at all) whenever the *full* current conversation is still under the trigger — the cache/summary substitution only engages once the real context crosses the configured trigger. **Stickiness headroom** (`_apply_stickiness_headroom`, `_KEEP_HEADROOM_FRACTION=0.75`): after the `keep` policy picks a cutoff, the kept suffix is forced below ~75% of *every* trigger before summarizing. Without it a `keep` in *messages* (e.g. 20) can itself exceed a *token* trigger (e.g. 60000) when recent messages are large (tool outputs / pasted docs), leaving the "compressed" context still over the trigger so it re-summarizes on every follow-up question; the headroom makes the sticky boundary actually stick (re-compaction only after genuinely new content). When the genuine (blocking) summarize path runs, it emits a `{"type":"context_compacting","message":"正在压缩上下文…"}` event on LangGraph's `custom` stream channel (the cheap reuse path stays silent); the chat frontend's `onCustomEvent` renders it as a transient toast. Tests: `tests/test_summarization_middleware.py`
+11. **TodoListMiddleware** - Task tracking with `write_todos` tool (optional, if plan_mode)
+12. **TokenUsageMiddleware** - Records token usage metrics when token tracking is enabled (optional)
+13. **TitleMiddleware** - Two-phase auto-title (never shows "untitled"): `before_model` sets an instant question-derived *provisional* title the moment the first user message arrives (no LLM, zero latency), then `after_model` asynchronously generates the final LLM title and replaces it (tracked via the `title_provisional` state flag). Normalizes structured message content before prompting the title model
+14. **MemoryMiddleware** - Queues conversations for async memory update (filters to user + final AI responses)
+15. **ViewImageMiddleware** - Injects base64 image data before LLM call (conditional on vision support)
+16. **DeferredToolFilterMiddleware** - Hides deferred tool schemas from the bound model until tool search is enabled (optional)
+17. **SubagentLimitMiddleware** - Truncates excess `task` tool calls from model response to enforce `MAX_CONCURRENT_SUBAGENTS` limit (optional, if `subagent_enabled`)
+18. **LoopDetectionMiddleware** - Detects repeated tool-call loops; hard-stop responses clear both structured `tool_calls` and raw provider tool-call metadata before forcing a final text answer
+19. **ClarificationMiddleware** - Intercepts `ask_clarification` tool calls, interrupts via `Command(goto=END)` (must be last)
+
+### Configuration System
+
+**Main Configuration** (`config.yaml`):
+
+Setup: Copy `config.example.yaml` to `config.yaml` in the **project root** directory.
+
+**Config Versioning**: `config.example.yaml` has a `config_version` field. On startup, `AppConfig.from_file()` compares user version vs example version and emits a warning if outdated. Missing `config_version` = version 0. Run `make config-upgrade` to auto-merge missing fields. When changing the config schema, bump `config_version` in `config.example.yaml`.
+
+**Config Caching**: `get_app_config()` caches the parsed config, but automatically reloads it when the resolved config path changes or the file's mtime increases. This keeps Gateway and LangGraph reads aligned with `config.yaml` edits without requiring a manual process restart.
+
+Configuration priority:
+1. Explicit `config_path` argument
+2. `DEER_FLOW_CONFIG_PATH` environment variable
+3. `config.yaml` in current directory (backend/)
+4. `config.yaml` in parent directory (project root - **recommended location**)
+
+Config values starting with `$` are resolved as environment variables (e.g., `$OPENAI_API_KEY`).
+`ModelConfig` also declares `use_responses_api` and `output_version` so OpenAI `/v1/responses` can be enabled explicitly while still using `langchain_openai:ChatOpenAI`.
+
+**Extensions Configuration** (`extensions_config.json`):
+
+MCP servers and skills are configured together in `extensions_config.json` in project root:
+
+Configuration priority:
+1. Explicit `config_path` argument
+2. `DEER_FLOW_EXTENSIONS_CONFIG_PATH` environment variable
+3. `extensions_config.json` in current directory (backend/)
+4. `extensions_config.json` in parent directory (project root - **recommended location**)
+
+### Gateway API (`app/gateway/`)
+
+FastAPI application on port 8001 with health check at `GET /health`. Set `GATEWAY_ENABLE_DOCS=false` to disable `/docs`, `/redoc`, and `/openapi.json` in production (default: enabled).
+
+**Routers**:
+
+| Router | Endpoints |
+|--------|-----------|
+| **Models** (`/api/models`) | `GET /` - list models; `GET /{name}` - model details |
+| **MCP** (`/api/mcp`) | `GET /config` - get config; `PUT /config` - update config (saves to extensions_config.json) |
+| **Skills** (`/api/skills`) | `GET /` - list skills; `GET /{name}` - details; `PUT /{name}` - update enabled; `POST /install` - install from .skill archive (accepts standard optional frontmatter like `version`, `author`, `compatibility`) |
+| **Memory V2** (`/api/memory/v2`) | `GET /` - builtin data (USER.md/MEMORY.md); `POST/PUT/DELETE /entries` - add/replace/remove entry; `GET/PATCH/DELETE /config` - effective config + per-user overrides; `POST /search` - Hindsight recall/reflect proxy |
+| **Agents** (`/api/agents`) | `GET /` - list built-in, owned, and published agents; `POST /` - create user-owned agent; `GET/PUT/DELETE /{id}` - read/update/delete by stable agent id; `GET /check` - validate non-empty display name (duplicates allowed) |
+| **Skills** (`/api/skills`) | `GET /` - list skills (optional `search` fuzzy-name + `tag_id` filters; each item carries `tags`); `GET /{name}` - details; `PUT /{name}` - update enabled; `POST /install` - install from .skill archive (accepts standard optional frontmatter like `version`, `author`, `compatibility`) |
+| **Skills** (`/api/skills`) | `GET /` - list skills (optional `search` fuzzy-name + `tag_id` filters; each item carries `tags` + owner-authored `detail`); `GET /{name}` - details; `PUT /{name}` - update enabled; `PUT /{name}/detail` - set the long-form `detail` text (custom skill: owner or admin; built-in/public: admin only ? materializes a `skills` row on demand); `POST /install` - install from .skill archive (accepts standard optional frontmatter like `version`, `author`, `compatibility`) |
+| **Skills** (`/api/skills`) | `GET /` - list skills (optional `search` fuzzy-name + `tag_id` filters; each item carries `tags` + owner-authored `detail`); `GET /{name}` - details; `PUT /{name}` - update enabled; `PUT /{name}/detail` - set the long-form `detail` text (custom skill: owner or admin; built-in/public: admin only ? materializes a `skills` row on demand); `GET /{name}/content` - raw SKILL.md source, restricted to an admin or the publisher (custom-skill owner); other users get 403 and see only description + detail; `POST /install` - install from .skill archive (accepts standard optional frontmatter like `version`, `author`, `compatibility`) |
+| **Skills** (`/api/skills`) | `GET /` - list skills (optional `search` fuzzy-name + repeatable `tag_ids` multi-select filter, OR-matched; each item carries `tags` + owner-authored `detail`); `GET /{name}` - details; `PUT /{name}` - update enabled; `PUT /{name}/detail` - set the long-form `detail` text (custom skill: owner or admin; built-in/public: admin only ? materializes a `skills` row on demand); `GET /{name}/content` - raw SKILL.md source, restricted to an admin or the publisher (custom-skill owner); other users get 403 and see only description + detail; `POST /install` - install from .skill archive (accepts standard optional frontmatter like `version`, `author`, `compatibility`) |
+| **Memory** (`/api/memory`) | `GET /` - memory data; `POST /reload` - force reload; `GET /config` - config; `GET /status` - config + data |
+| **Agents** (`/api/agents`) | `GET /` - list built-in, owned, and published agents (optional `search` + repeatable `tag_ids` multi-select filter, OR-matched; each item carries `tags` and `featured_order`); `POST /` - create user-owned agent; `GET/PUT/DELETE /{id}` - read/update/delete by stable agent id; `GET /check` - validate non-empty display name (duplicates allowed); `GET /selector` - homepage agent picker (single admin-curated list, visibility-filtered, sorted by `featured_order` asc); `PUT /{id}/featured` - **admin** sets/clears the homepage `featured_order` (must be `published=true` to feature; passing `order: null` clears unconditionally); `PUT /featured/order` - **admin** bulk-reorders the homepage list (body `{agent_ids: [...]}` ? indices become 0..N-1) |
+| **Tags** (`/api/tags`) | `GET /` - list tags (optional `tag_type` filter; open to all); `POST /` - create tag (admin); `PUT /{id}` / `DELETE /{id}` - update/delete tag (admin); `GET /assignments/{target_type}/{target_id}` - tags on a resource; `PUT /assignments/{target_type}/{target_id}` - replace a resource's tags (admin). Tags are typed `agent`/`skill`/`scheduled_task` and may only tag resources of their own type. |
+| **Uploads** (`/api/threads/{id}/uploads`) | `POST /` - upload files (auto-converts PDF/PPT/Excel/Word); `GET /list` - list; `DELETE /{filename}` - delete |
+| **Threads** (`/api/threads/{id}`) | `DELETE /` - remove DeerFlow-managed local thread data after LangGraph thread deletion; unexpected failures are logged server-side and return a generic 500 detail |
+| **Threads search** (`POST /threads/search`, LangGraph-compat) | Conversation listing. **System threads are excluded by default**: rows whose `metadata.system` is truthy (scheduler runs, roundtable workers) never crowd the user's list; a caller whose `metadata` filter explicitly references `thread_type`/`system` (e.g. studio notebooks) opts out of the exclusion. New `query` field = case-insensitive title (`display_name`) substring search, used by the chats page's server-side pagination. `POST /threads/count` takes the same request body (limit/offset ignored) and returns `{total}` under identical filtering/exclusion semantics ? powers the chats page's ??/??? display without fetching rows (`ThreadMetaStore.count`). `ThreadMetaStore.search` gained `exclude_system`/`query` params; JSON-blob filters batch-scan until the requested `limit`/`offset` window is full (so pagination over a heavily-filtered set never under-fills), with a `thread_id` tie-breaker on the `updated_at DESC` ordering for stable OFFSET paging. Tests: `tests/test_thread_meta_search.py` |
+| **Artifacts** (`/api/threads/{id}/artifacts`) | `GET /{path}` - serve artifacts; active content types (`text/html`, `application/xhtml+xml`, `image/svg+xml`) are always forced as download attachments to reduce XSS risk; `?download=true` still forces download for other file types |
+| **Suggestions** (`/api/threads/{id}/suggestions`) | `POST /` - generate follow-up questions; rich list/block model content is normalized before JSON parsing |
+| **Thread Runs** (`/api/threads/{id}/runs`) | `POST /` - create background run; `POST /stream` - create + SSE stream; `POST /wait` - create + block; `GET /` - list runs; `GET /{rid}` - run details; `POST /{rid}/cancel` - cancel; `GET /{rid}/join` - join SSE; `GET /{rid}/messages` - paginated messages `{data, has_more}`; `GET /{rid}/events` - full event stream; `GET /../messages` - thread messages with feedback; `GET /../token-usage` - aggregate tokens |
+| **Feedback** (`/api/threads/{id}/runs/{rid}/feedback`) | `PUT /` - upsert feedback; `DELETE /` - delete user feedback; `POST /` - create feedback; `GET /` - list feedback; `GET /stats` - aggregate stats; `DELETE /{fid}` - delete specific |
+| **Runs** (`/api/runs`) | `POST /stream` - stateless run + SSE; `POST /wait` - stateless run + block; `GET /{rid}/messages` - paginated messages by run_id `{data, has_more}` (cursor: `after_seq`/`before_seq`); `GET /{rid}/feedback` - list feedback by run_id |
+| **Thread Shares** (`/api/thread-shares`) | Per-conversation sharing with two modes (`thread_shares.mode`). **`import`**: `POST /` `{thread_id, mode}` - create (or reuse) a revocable share code; `GET /{code}/preview` + `POST /{code}/import` - copy that conversation into the caller's account (fresh `thread_id` via `app/gateway/thread_copy.py` + new `threads_meta` row + `user-data` copy; dedup on `metadata.shared_origin`). **`view`**: a public, read-only static page ? see Public Shares below. Both modes: `GET /` - list my codes; `DELETE /{code}` - revoke. |
+| **Public Shares** (`/api/public/shares`) | **Unauthenticated** (the `/api/public/` prefix is whitelisted in `auth_middleware._PUBLIC_PATH_PREFIXES`). `GET /{code}` - read-only conversation snapshot (title + serialized messages + artifacts) for a `view`-mode share; messages are read straight from the latest checkpoint. `GET /{code}/files/{path}` - serves a file referenced by that conversation (images, Markdown, text, PDF, ?) so the shared page's links are openable; path resolution is confined to the owner's thread directory. Active content (`text/html`, `application/xhtml+xml`, `image/svg+xml`) is forced to download (`_public_share_disposition`); everything else is served inline. Missing/revoked/non-`view` codes all return 404. |
+| **Roundtable Drafts** (`/api/roundtable-drafts`) | Per-user persistence for the roundtable-planning page (replaces the old browser `localStorage`). `GET /` - list drafts (lightweight meta); `POST /` - create; `GET/PUT/DELETE /{id}` - full read / partial update / delete. A draft = one full session (Step 1 intent + Step 2 roundtable), step snapshots stored as JSON in `PortableLongText`. Recommendation history bound to a draft: `GET /{id}/recommendations` - list newest-first; `POST /{id}/recommendations` - append one run (objective/status/rationale/picks/candidates). Backed by `deerflow.persistence.roundtable_drafts` (tables `roundtable_drafts`, `roundtable_recommend_history`); store wired in `deps.py` as `app.state.roundtable_draft_store`. |
+| **Roundtable Draft Shares** (`/api/roundtable-draft-shares`) | Per-draft sharing for the roundtable-planning page, mirroring Thread Shares but the shared resource is a `roundtable_drafts` row. Two modes (`roundtable_draft_shares.mode`). **`import`**: `POST /` `{draft_id, mode}` - create (or reuse) a revocable code; `GET /{code}/preview` + `POST /{code}/import` - copy that draft (step1/2/3) into the caller's account under a deterministic per-(code,recipient) id (`imp_<sha1>` ? re-import is a no-op). **`view`**: a public, read-only result page ? see Public Roundtable Shares. Both: `GET /` - list my codes; `DELETE /{code}` - revoke. Backed by `deerflow.persistence.roundtable_draft_shares` (table `roundtable_draft_shares`, migration `20260606_01`); store wired in `deps.py` as `app.state.roundtable_draft_share_store`. Tests: `tests/test_roundtable_draft_shares.py`. |
+| **Public Roundtable Shares** (`/api/public/roundtable-shares`) | **Unauthenticated** (`/api/public/` whitelisted). `GET /{code}` - read-only snapshot for a `view`-mode draft share: returns the Step 3??????report HTML (`step3.html`, a self-contained document) + title/summary/`has_report`. The frontend renders it in a sandboxed `iframe` (no `allow-scripts`) at `#/roundtable-share/<code>`. Missing/revoked/non-`view` codes all return 404. |
+| **Scheduled Tasks** (`/api/scheduled-tasks`) | Per-user cron-like agent tasks. `GET/POST /` - list/create owned tasks; `PATCH/DELETE /{id}` - update/delete (owner); `POST /{id}/{pause,resume,trigger}`; `GET /{id}/runs`, `GET /runs/{rid}`, run-file preview/edit/download; `GET /runs/{rid}/messages` - agent messages for the run's underlying agent run (used to inspect a stuck/running run; authorized via `_load_run`); `GET /runs/{rid}/html` - generated HTML document for an `html_page` run; `DELETE /runs/{rid}` - owner-scoped, with an admin fallback so admins can delete any user's run; publish/subscribe under `/{id}/{publish,unpublish,subscribe}` and `GET /published`. **Admin (cross-user):** `GET /admin/all` - every task; `PATCH/DELETE /admin/{id}` - edit/delete any task; `POST /admin/{id}/trigger`; `GET /admin/{id}/runs`. Admin routes require `system_role == "admin"`. |
+| **HTML Page Favorites** (`/api/html-page-favorites`) | Bookmarked generated HTML pages, owner-scoped. `GET /` - list; `POST /` `{page_id, title?, description?, tags?}` - bookmark a `scheduled_html_pages` row (extracts a reusable style/layout summary via GLM5 at favorite time); a page can only be favorited once per user (`DuplicateFavoriteError` ? HTTP 409). `GET/DELETE /{id}` - read/delete (the create-task reference picker deletes favorites here). A favorite can be selected as a style reference when creating a new `html_page` task. |
+| **Public Scheduled Tasks** (`/api/public/scheduled-tasks`) | **Unauthenticated** mirror for embedding. Adds `GET /runs/{rid}/html` and `GET /{task_id}/latest-html` for the `html_page` preview; the run result keeps `result.html` (only file paths are redacted). |
+| **Light Apps** (`/api/light-apps`) | Light-app registry (????? / ????), **global/shared** (no user scoping; `created_by`/`updated_by` audit-only). Two kinds: `window_open` (launched in a new tab from the ???? card grid) and `iframe` (embedded under a parent sidebar menu via `mount_location` and dynamically injected into the sidebar). Both carry `route_params` ? a list of `{key, paramType: login\|theme, source, customValue, themeValue}` descriptors the **frontend** resolves at launch time (login params read from `localStorage.userInfo`; theme params use a literal value) and appends to `url` as query string. `GET /` - list (filters `app_type`/`mount_location`/`enabled_only`/`search`); `GET /{app_id}` - read one (both open to any authenticated user); `POST /`, `PUT /{app_id}`, `DELETE /{app_id}` - **admin only**; iframe apps must carry a `mount_location` (400 otherwise); the unique `name` collides as HTTP 409. Backed by `deerflow.persistence.light_apps` (table `light_apps`, auto-created by `create_all`); store wired in `deps.py` as `app.state.light_app_store`. Tests: `tests/test_light_apps.py`. |
+| **Knowledge** (`/api/knowledge`) | Product knowledge base, **global/shared** (no user isolation; writes record `created_by`/`updated_by` for audit). `POST /ingest/thread/{thread_id}` - sediment one thread into a note (reads messages from the checkpointer, `threads:read` owner-checked); `POST /ingest/threads/batch` - batch-sediment the caller's recent threads (`skip_existing` via `.manifest.json` dedup). **Defaults to `background=true`**: the slow per-thread extraction loop is deferred to a FastAPI background task and the endpoint returns `{queued:true, processed:N}` immediately (each thread's transcript is truncated to `max_input_chars` before the LLM call). Running the loop inline for a large batch outlives the nginx gateway timeout (504), which the frontend's `parseResponse` surfaces as a generic "Request failed"; `background=false` keeps the blocking path (full per-thread `created/skipped/failed` tally) for small batches/scripts. Shared loop: `_run_batch_ingest`. **Optional model selection**: both ingest endpoints accept `model_name` (overrides `knowledge.extract_model_name` for that capture; threaded through `KnowledgeService.capture_thread` ? `extract_thread_llm_multi`). **Model-error surfacing**: when the chosen model produces nothing usable (timeout / API error / non-JSON) capture still **falls back to rule-based extraction and lands a draft** (`llm_fallback:true`); the single-thread (non-background) response and the batch response (`llm_fallback` count) carry this so the UI tells the user "???????". The "??????" chat button stays background/fire-and-forget (passes `model_name`, no per-save error toast); the batch-import dialog is synchronous and shows the fallback count. `POST /ingest/file` - **import an uploaded file** (multipart: `file` + optional `title`/`tags`/`status`/`model_name`) into wiki note(s): the file is converted to Markdown (md/txt read directly, pdf/word/ppt/excel via `convert_file_to_markdown`), distilled by the same wiki-capture LLM pipeline as thread capture (`source_type="file"`), and falls back to a verbatim **draft** note when the model returns nothing usable (`llm_fallback:true`); `POST /ingest/search` - sediment search/tool results as a reference note; `POST /notes` - manual note; `GET /notes` (keyword/tag/source_type/status + pagination); `GET/PUT/DELETE /notes/{id}` (DELETE is soft = archive, `?hard=true` removes the vault file); `POST /search` - `mode=keyword\|vector\|hybrid` (vector?keyword fallback when no embedding endpoint); `GET /notes/{id}/versions` - edit history; `GET /duplicates` + `POST /merge` - dedup; `POST /reindex` - rebuild vector index; `GET /graph` - product graph data (notes + entities, graph-blacklist filtered); `GET/POST /graph/blacklist` + `DELETE /graph/blacklist/{id}` - graph node blacklist (labels = entity names / note titles, case-insensitive match; blacklisted nodes and their edges are dropped from `/graph` and the wiki-export artifacts; table `knowledge_graph_blacklist`, auto-created); `POST /export/graph` + `GET /export/files/{graph.json\|graph.html\|graph.graphml\|cypher.txt}` - obsidian-wiki `wiki-export` artifacts (self-contained `graph.html` for iframe); `POST /resolve` `{targets:[?]}` + `GET /notes/{id}/wikilinks` - resolve `[[wikilink]]` targets to a note id / vault page / missing (powers clickable wikilinks); `GET /vault-page?path=` - read a raw vault meta/hub Markdown page (no DB note). `GET/POST /folders` + `PUT/DELETE /folders/{id}` - user-managed directories (notes carry a `folder` path; `GET /notes?folder=` filters a dir + sub-dirs); `GET/POST /extract-templates` + `GET/PUT/DELETE /extract-templates/{id}` - editable extraction-prompt templates (thread/document ingest accept `template_id`; capture/PUT-note accept `folder`). Backed by the `deerflow.knowledge` module (see below). |
+
+Proxied through nginx: `/api/langgraph/*` ? LangGraph, all other `/api/*` ? Gateway.
+
+### AI Writing Built-in Functional Agents (????????)
+
+The AI-writing graph (`packages/harness/deerflow/agents/ai_writing/`) has four stage "brains" ? ?????? / ????? / ?? / ?? ? that can each run in one of two modes, switched per-stage by `config.yaml ? ai_writing.use_builtin_agents.{researcher,outliner,writer,editor}` (default **false**, mtime hot-reloaded, new sessions pick up changes immediately):
+
+- **Legacy (flag off)**: hard-coded prompts in `prompts/*.py`, direct LLM call (unchanged behavior).
+- **Builtin-agent (flag on)**: the node loads the corresponding functional agent from `.deer-flow/agents/ai-writing-<stage>/` ? `SOUL.md` becomes the system prompt and `config.yaml`'s `model` overrides the stage model (priority: `agent.model > node explicit fallback (keyword_model/rank_model) > state.model_name`). Admins edit SOUL/model from the agent management page. Runner: `nodes/_agent_runner.py` (`run_writing_agent` / `stream_writing_agent` / `parse_marker_json`). Output contracts (marker + ```json``` blocks, intent.py-style): researcher `[KEYWORDS_READY]` / `[MATERIALS_RANKED]`, outliner `[OUTLINE_READY]`, editor (per-dimension) `[REVIEW_READY]`; writer emits raw section Markdown (keeps per-section streaming) with the legacy `[[SECTION_BLOCKED]]` sentinel in strict mode. **Orchestration stays in the nodes** (search fan-out, per-section loop, 3-dimension review loop, all pause points) ? the agent only does single structured generations, and a missing/empty SOUL silently falls back to the legacy path (`WritingAgentUnavailable`).
+
+Seeding mirrors the roundtable pattern: verified `config.yaml` + `SOUL.md` ship in `app/gateway/routers/_ai_writing_seed_assets/<id>/`; `_ai_writing_seed.py::ensure_ai_writing_functional_agents()` copies them to `.deer-flow/agents/` when missing (never overwrites local edits), called from the startup lifespan (before `_sync_legacy_agents`, so the agents appear in `/api/agents` on first boot) and self-heals from `GET /api/ai-writing/sessions` + `GET /api/ai-writing/article-types`. Tests: `tests/test_ai_writing_seed.py`, `tests/test_ai_writing_agent_runner.py`.
+
+**Skill-driven retrieval (检索类型).** `material_source` has exactly two values: `general` (通用检索, default) and `notebook` (我的空间); legacy `knowledge_base`/`skill` are normalized to `general` by the researcher node. General retrieval is orchestrated from the **skills configured on the `ai-writing-researcher` agent** (`config.yaml → skills`, editable from the agent management page): the built-in `knowledge-base-search` skill (知识库检索) runs the native keyword fan-out web/intranet search (the old default path), while any other configured skill searches with a `description`-targeted query; skills run in configured order with early-stop at `max_materials`, and an empty/unresolvable skills config falls back to the built-in knowledge search so retrieval always works. The skill itself is seeded from `_ai_writing_seed_assets/skills/knowledge-base-search/SKILL.md` into `skills/public/` (never overwrites), and a pre-existing researcher `config.yaml` lacking the `skills` key is patched in place with the default (explicit `skills: []` is respected). **PAUSE-1 input append flow**: when the user types natural language into the material-confirm input (`re_search`/legacy `enable_skill_search` + `user_query`), the node matches the most relevant configured skill via `SkillPicker(candidates=…)` (skipped when only one skill is configured), searches with the user's words, and **appends** the new materials to the existing package (URL-dedup, original order preserved, no LLM re-rank/truncation, 100-item hard cap). **Skill-failure fault tolerance**: a configured skill that times out / raises / has all its keyword searches fail (vs. legitimately returning 0 results) no longer just silently moves on — the researcher node emits a `skill_search_step` of type `skill_error` (carrying the error `reason`), and if the configured skills collected **no** new materials at all it runs a `fallback` pass over other available skills (`_fallback_candidate_skills`: the built-in `knowledge-base-search` native path first, then other enabled catalog skills not yet tried) instead of erroring out. The frontend `SkillStepsCard` renders `skill_error` (rose) and `fallback` (amber) so the timeline shows "技能报错 → 改用其他技能搜集". Tests: `tests/test_ai_writing_material_source.py`.
+
+### Admin Leaderboard Snapshots (`/api/admin/users/leaderboard*`)
+
+The admin ????? is served from **precomputed per-day snapshots** in `admin_leaderboard_daily_stats` (keyed by `(stat_date, settings_hash)`), not by aggregating raw rows on every request. A background scheduler inside the Gateway (`_leaderboard_snapshot_scheduler_loop` in `app/gateway/app.py`, singleton file-locked) ticks every `ADMIN_LEADERBOARD_BACKGROUND_INTERVAL_SECONDS` (default 300s): it prewarms recent days, then builds any `queued`/`error`/stale-`running` day. `GET /leaderboard` merges the per-day snapshots for the requested range. **Today** is rebuilt whenever its snapshot is older than `ADMIN_LEADERBOARD_TODAY_TTL_SECONDS` (default 43200s = half a day, so the leaderboard refreshes ~twice daily); a past day, once `ready`, is treated as fresh forever.
+
+Two on-demand/automatic refresh paths layer on top:
+- **Manual "????"** ? `POST /leaderboard/refresh-today` force-rebuilds *today's* snapshot **synchronously** (bounded by the same semaphore/timeout the scheduler uses), then returns the merged range. Lets an admin pull the latest real-time records into the stats without waiting for the TTL. Frontend: a "????" button on `AdminUserLeaderboardPage` ? `useRefreshTodayLeaderboard` (`core/admin/hooks.ts`), which primes + invalidates the `["admin-activity","leaderboard"]` query.
+- **Daily finalize of the previous day** ? once per day, on/after `ADMIN_LEADERBOARD_FINALIZE_HOUR` (local Beijing, default `1` = ?? 1 ?), the scheduler tick force-recomputes *yesterday* a single time so runs/tool-calls that only landed near midnight are fully captured. The decision is the pure `deerflow.persistence.admin_stats.finalize.should_finalize_previous_day` (status/generated_at/now/hour) ? the rebuild stamps `generated_at` to today, which self-debounces it for the rest of the day (no extra state file). Tests: `tests/test_leaderboard_finalize.py`.
+
+### Scheduled Task Runtime
+
+`ScheduledTaskService` (`deerflow/runtime/scheduler/service.py`) polls due tasks every `poll_interval_seconds` and executes each in its own throwaway thread, serialized by a shared `_agent_run_lock`.
+
+**Stuck-run recovery**: each execution attempt is bounded by `_RUN_ATTEMPT_TIMEOUT_SECONDS` (30 min) via `asyncio.wait_for`. On timeout the run is retried up to `_MAX_RUN_RETRIES` (3) times, then marked `failed`. A watchdog (`_reconcile_stuck_runs`, run every poll) recovers runs left in `running` with no in-process owner ? e.g. orphaned by a crash or restart ? re-driving them through the same retry budget so a task can never stay perpetually "executing". The `scheduled_task_runs.retry_count` column tracks consumed retries.
+
+**Unique task names**: task names are unique per user (`uq_scheduled_tasks_user_name`). `create_task`/`update_task`/`admin_update_task` proactively check via `_assert_name_available()` and raise `DuplicateTaskNameError` on a collision; the router translates that into HTTP `409` instead of letting the raw DB `IntegrityError` surface as a `500`.
+
+**HTML page tasks (`task_kind == "html_page"`)**: a second task kind that renders a full HTML document each run instead of a Markdown report. The kind + its config (`html_config`: `page_prompt`, `negative_prompt`, `reference_favorite_id`, `model_name`) live inside `execution_context` ? no schema change to `scheduled_tasks`. Data gathering is delegated to the agent's own search skills (driven by `page_prompt`), so there is no explicit search-query/data-source field; the render model is user-selectable (`model_name`, default `zai-org/GLM-5-FP8`). `_run_task_attempt()` dispatches by kind; `_run_html_page_attempt()` runs the pipeline: gather real data via the lead agent (its search tools) ? optionally load a reference favorite's reusable style/layout ? render with the chosen model (thinking off) ? fault-tolerant JSON parse ? sanitize (BeautifulSoup, strips `<script>`/`<iframe>`/`on*`/`javascript:`) ? structural validate ? one GLM5 repair pass if invalid ? **completeness gate**: if the HTML is still not a full valid document (missing doctype/html/head/body, empty body, or residual scripts) after the repair pass, raise `HtmlGenerationError` so the run is marked **failed** (a page only counts as successful when it is a complete, previewable HTML document) ? best-effort Playwright screenshot (no-op when Playwright absent) ? archive `index.html` ? persist a `scheduled_html_pages` row ? return `result` with `format: "html"` and the full HTML inline. Pure helpers live in `deerflow/runtime/scheduler/html_page.py`. The frontend previews the HTML in a sandboxed `iframe` (no `allow-scripts`) on the run-detail and public pages; a user can bookmark a page (style summary extracted via GLM5) and later pick it as a style reference. New tables `scheduled_html_pages` + `html_page_favorites` (`deerflow/persistence/html_pages/`, wired as `app.state.html_page_store`) are created by `Base.metadata.create_all` with no migration. Tests: `tests/test_html_page_tasks.py`.
+
+**Unified output switches (`require_html` + `require_markdown`).** The task form no longer splits "???? / HTML ??" into separate kinds. `execution_context` now carries two **independent** booleans that may both be on: `require_html` (???? HTML) and `require_markdown` (???? Markdown). `_task_requires_html()` returns true for `require_html` **or** legacy `task_kind == "html_page"` (so old tasks need no data migration), and `_run_task_attempt()` dispatches to `_run_html_page_attempt` when it is true, else the regular agent attempt. When **both** switches are on, `_gather_html_data()` returns `(text, files)` ? the gather agent run is additionally told to save a `.md` file ? and `_run_html_page_attempt` validates it (`MarkdownRequirementError` on failure) and surfaces it alongside the generated page, so the run delivers both artifacts. The single main `prompt` doubles as the HTML page requirement (`html_config.page_prompt`); there is no separate page-prompt field. Tests: `tests/test_scheduled_task_require_html.py` + the markdown-coexistence cases in `tests/test_html_page_tasks.py`.
+
+### Per-user Prompts & UI Preferences
+
+**Custom prompts can be scoped to one agent.** `user_custom_prompts` gains a nullable `agent_id` column (migration `20260603_02`): `NULL` = global (the chip shows in every chat), otherwise the prompt only surfaces when that agent is selected. The visibility filter is applied in the frontend `CustomPromptBar` (it receives the current agent id from `ChatPage`'s selector or `AgentChatPage`'s `assistantId`); the backend (`/api/user-prompts` CRUD) just stores/returns/clears `agent_id` (update accepts an explicit `null` to clear). Tests: `tests/test_user_prompts_agent_id.py`.
+
+**Per-user UI preferences** (`deerflow/persistence/user_preferences/`, wired as `app.state.user_preferences_store`; migration `20260603_03`, also auto-created by `create_all`): one JSON-blob row per user so new toggles need no schema change. Router `/api/user-preferences` (`GET`/`PUT`, account-scoped). Current key: `hide_admin_prompt_prefix` ? when true the chat input hides the admin-configured prompt-prefix switch and never applies the prefix (the user uses only their own prompts; enforced in `ChatPage` for both `shouldApplyPrefix` and `promptPrefixProp`). Toggled from the "???????" dialog. Tests: `tests/test_user_preferences.py`.
+
+**HTML skills (tools the render step uses).** Two offline skills under `skills/public/` shape the generated HTML and are toggled in the skills settings page (enabled by default): `html-page-builder` (a self-contained inline-CSS design system ? colors/cards/tables/KPI/responsive) and `html-charts` (inline-SVG charts that render in the no-`allow-scripts` preview iframe; optional local ECharts documented in its `assets/README.md`). A skill opts into HTML injection via `metadata.kind: html_page` in its SKILL.md frontmatter; `_collect_html_skill_instructions()` reads the bodies of **enabled** such skills and `build_generation_prompt(skill_instructions=?)` injects them, so the (raw-model) render step honors the user's skill selection. The data-gathering step and the render/repair/style passes all use the task's selected `model_name` (no silent fallback to the default model ? and there is **no hardcoded GLM-5**; an unset selection falls back to config `models[0]`). To stop auxiliary LLM calls from leaking onto the default model, **`TitleMiddleware` is skipped when `is_scheduled_run` is set** (`_build_middlewares` in `lead_agent/agent.py`) ? scheduled/HTML runs use throwaway threads whose titles are never shown, and `config.title.model_name` is usually unset (? would call `models[0]`, not the chosen model). The completeness gate (`validate_html`) also requires a non-empty `<title>` and forbids external stylesheets, on top of the doctype/html/head/body + non-empty-body + no-script checks.
+
+### Knowledge Base (`packages/harness/deerflow/knowledge/`)
+
+Self-contained product knowledge base. **All knowledge capability lives in this one module** (harness layer, `deerflow.*` only ? no `app.*` imports). Wired as `app.state.knowledge_service` in `deps.py` via `make_knowledge_service(sf, config.knowledge)` (returns `None` on the memory backend or when `knowledge.enabled=false` ? router 503). Surfaced by `app/gateway/routers/knowledge.py` (`/api/knowledge`). Config: `config.yaml` ? `knowledge` (`deerflow/config/knowledge_config.py`).
+
+**Two stores, kept in sync on every write:**
+- **DB mirror** ? tables `knowledge_notes` / `knowledge_sources` / `knowledge_tags` / `knowledge_note_tags` (`knowledge/models.py`, registered in `persistence/models/__init__.py`, auto-created by `Base.metadata.create_all`; portable types `PortableLongText`/`BeijingDateTime`). Powers listing/filter/search/audit. Global/shared ? no `user_id` column; `created_by`/`updated_by` are audit-only (resolved via `get_effective_user_id()`).
+- **Markdown vault** ? obsidian-wiki compatible, default `{$DEER_FLOW_HOME}/data/knowledge/obsidian-vault`. `ObsidianWikiAdapter` (`obsidian_adapter.py`) writes wiki-capture frontmatter+body, and after **every** write syncs `index.md` / `log.md` (CAPTURE/INGEST/UPDATE/ARCHIVE/EXPORT) / `hot.md` / `.manifest.json` (source-hash dedup for `skip_existing`). `markdown_vault.py` handles frontmatter build/parse, slugging, wikilink extraction and **path-traversal-proof** resolution.
+
+**Module layout:** `models.py` (ORM: notes/sources/tags/note_tags + embeddings/versions/entities/relations) ? `repository.py` (DB only, no LLM) ? `schemas.py` (NoteDraft/SourceDraft DTOs) ? `skill_templates.py` ? `markdown_vault.py` ? `manifest.py` ? `extractor.py` (rule-based fallback) ? `llm_extractor.py` (**default**: distills declarative knowledge per wiki-capture, structured JSON incl. entities/relations; falls back to `extractor.py`) ? `sources.py` ? `exporter.py` (`graph.json`/`graph.graphml`/`cypher.txt` + self-contained vanilla-JS `graph.html`; entity nodes + relation edges. **`graph.html` is theme-aware** ? it reads `?theme=light|dark` from its own URL since the embedding iframe can't inherit the app's CSS; `KnowledgeGraphPage.tsx` appends the active day/night mode and re-skins on toggle. The layout cools down and **freezes** so dense graphs stay clickable, node radius scales with degree, and hovering/clicking a node focuses it + its direct neighbours while dimming the rest + a detail panel ? taming heavy node/edge clutter) ? `chunking.py` + `embeddings.py` (OpenAI-compatible embedding client, keyword fallback) + `search.py` (keyword/vector/hybrid) ? `sensitive.py` (secret/PII redaction) ? `dedup.py` (Jaccard + hash dedup) ? `auto_ingest.py` (background queue/worker) ? `obsidian_adapter.py` ? `service.py` (`KnowledgeService` orchestrator ? the only thing the router touches; capture entry points `capture_thread` / `capture_search` / `capture_file` / `create_manual_note`, all converging on `_persist_draft`). Tests: `tests/test_knowledge.py` (20 cases). Frontend: `frontend-web/src/core/knowledge/*` + `pages/KnowledgeBasePage.tsx` / `KnowledgeGraphPage.tsx` + chat `knowledge-save-trigger.tsx` / `knowledge-rag-toggle.tsx`; spec at `frontend-web/docs/product-knowledge-base-dev.md`.
+
+**Wikilinks & multi-note capture** (`knowledge.{related_notes_enabled,related_notes_limit,multi_note_enabled}`, all default-on; `project_name` defaults to `zncm`):
+- **Navigable `[[wikilinks]]`** ? `service.resolve_wikilink(s)` maps a `[[target]]` to a knowledge note (by id ? `vault_path` ? title) or a vault meta page (`read_vault_page`, path-traversal-proof), else `missing`. `markdown_vault.normalize_wikilink_target` canonicalizes targets (drops `|alias`, slashes, `.md`). The frontend `MarkdownRenderer` renders non-numeric `[[...]]` as clickable links from the resolver map (numeric `[[123]]` stay citation badges); note links navigate, vault links open a `VaultPageDialog`, missing links are greyed.
+- **Real related notes only (no structural meta-links)** ? `extractor.related_links()` now returns `[]`: notes no longer auto-link `index` / `_meta/taxonomy` / a project hub (they were noise in every `## Related`). On capture, `service._find_related_notes` cross-links **only by shared *distinctive* tags** (project/`knowledge-base` defaults excluded; free-text title/summary matching was removed because CJK tokenization linked unrelated notes via generic words like ??/????) and `_inject_related_notes` writes real `[[Title]]` links into `## Related` (plus note?note `related` graph edges); no distinctive tags ? no related notes. The renderers omit the `## Related` section entirely when there are no related notes and no sources. The vault's own `index.md`/`_meta/taxonomy.md` still exist (obsidian scaffold) and remain resolvable for manually-written links. The router enriches `GET /notes/{id}` with `created_by_name`/`updated_by_name` (auth-provider email lookup) so the UI shows a person, not a UUID. The frontend `MarkdownRenderer` renders wiki-capture provenance markers `^[inferred]`/`^[ambiguous]` as small "??"/"??" badges.
+- **LLM-extraction outcomes** ? capture distinguishes three LLM outcomes: (1) **success** ? declarative note(s); (2) **deliberate rejection** ? the model returns `should_ingest:false` with no notes (greeting / meta chatter) ? `extract_thread_llm_multi` raises `LLMRejectedIngest` and `capture_thread` raises `EmptyThreadError` (skip ? nothing persisted, so worthless chats never enter the KB); (3) **genuine failure** ? API error / non-JSON ? returns `[]`, capture falls back to the rule-based verbatim extractor and lands the note as a **`draft`** (not `approved`), logs a WARNING, and returns `llm_fallback:true` (surfaced by the router). `llm_extractor` logs an actionable WARNING/INFO for each non-success path so extraction issues are diagnosable.
+- **Multi-note split** ? `llm_extractor.extract_thread_llm_multi` lets one conversation distill into 1?4 cross-linked notes (`notes:[?]` JSON array; legacy single-object output still accepted). `capture_thread` persists each, sibling-cross-links them by title, and returns `{note, created, notes:[?]}`. Gated by `multi_note_enabled` (false ? first note only). Tests: `tests/test_knowledge_wikilinks.py`.
+
+**Phases 2?4 (all config-gated, defaults safe):**
+- **Phase 2 ? vector RAG**: `knowledge.embedding.{enabled,base_url,api_key,model,...}` enables an OpenAI-compatible embedding endpoint; notes are chunked + embedded on write (`knowledge_embeddings`), `POST /search` supports `mode=keyword|vector|hybrid` (vector degrades to keyword when unconfigured). `KnowledgeRagMiddleware` (lead-agent chain) injects a transient `<knowledge-context>` recall before the model call, gated **per-conversation** by the `knowledge_rag_enabled` runtime flag (frontend chat toggle) falling back to global `knowledge.rag_enabled`. `POST /reindex` rebuilds the index.
+- **Phase 3 ? governance**: `sensitive.py` redacts secrets/PII before persistence (`sensitive_redaction_enabled`); every update snapshots `knowledge_note_versions` (`GET /notes/{id}/versions`); `GET /duplicates` + `POST /merge` for dedup; background `AutoIngestQueue` (wired in `deps.py`, started/stopped in the gateway lifespan) auto-sediments finished conversations **off the chat path** when `auto_ingest_enabled` ? enqueued from `KnowledgeRagMiddleware.after_agent`, value-judged by min-chars + `skip_existing`, landing as `draft`.
+- **Phase 4 ? graph**: LLM extraction also yields entities/relations ? `knowledge_entities`/`knowledge_relations`; `exporter.build_graph` adds entity nodes + note?entity/entity?entity edges (reflected in `graph.html`/`graph.json`); `GET /graph` returns product graph data.
+
+**User directories + editable extraction templates + display polish (knowledge UX overhaul):**
+- **Custom directories (`knowledge_notes.folder` + `knowledge_folders`)** ? notes carry a free-form, slash-joined `folder` path (e.g. `投研/行业`) independent of the wiki-capture `category` and of `vault_path`. `knowledge_folders` is the manageable directory list (supports empty folders + rename/move that re-points descendant folders **and** filed notes). Router CRUD under `/api/knowledge/folders` (`GET`/`POST`/`PUT/{id}`/`DELETE/{id}?reassign_to=`); `GET /notes?folder=` filters a directory + its sub-folders (`folder=""` ? unfiled). Capture (`capture_thread`/`capture_file`/`create_manual_note`) and `PUT /notes/{id}` accept `folder` (the note PUT uses `model_fields_set` so an absent key leaves the folder untouched, a present `null` clears it). Frontend `buildKnowledgeTree(notes, folderPaths)` builds the left tree from `folder` (no longer from `vault_path`); folder management dialog + `FolderSelect` in the create/import dialogs and the note detail header.
+- **Editable extraction prompts (`knowledge_extract_templates`)** ? the hard-coded distillation prompts are seeded as editable templates on startup (`KnowledgeService.seed_default_templates`, idempotent — `默认 · 对话提炼` / `默认 · 文档提炼`, each `is_default` for its scope). `scope ∈ {thread, document, both}`; capture resolves an explicit `template_id` → the scope default → the built-in constant, and passes it to `llm_extractor.extract_thread_llm_multi` / `extract_document_llm_multi` as `system_prompt` (used verbatim). Router CRUD under `/api/knowledge/extract-templates` (409 on duplicate name). Frontend: extraction-template management dialog + `TemplateSelect` in the batch/file import dialogs. Editing a template is the supported way to fix "提炼太精简" (loosen the prompt to keep more detail) and to add alternate extraction logics.
+- **Single source list** ? `extractor._related_section_md` no longer emits a `### Sources` sub-section into the body; the note detail shows sources once via the structured `来源` block (`note.sources`). `## Related` stays pure wikilinks so the frontend extracts it cleanly into the related-docs module.
+- **Chinese section headings (display-only)** ? the frontend `localizeHeadings` (wikiUtils) maps `Context/Finding / Decision/Reasoning/Implications/Related/Sources` to 中文 at render time (applied to the body fed to both `extractToc` and the viewer, so anchors stay consistent). No data migration — historical English-heading notes render in Chinese too.
+- **Historical-data cleanup** ? `scripts/cleanup_knowledge_related.py` (one-off, idempotent, `--dry-run`/`--note-id`) strips legacy `## Related` structural links (vault scaffold: `index`/`_meta/taxonomy`/`projects/…`, any slash-bearing target) and the old `### Sources` sub-section from existing notes (DB `content_md` + vault files, via `KnowledgeService.update_note`).
+- Migration `20260616_02` + idempotent `scripts/sql/column_additions_mysql.sql` cover `knowledge_notes.folder` (+ index); the two new tables are auto-created by `create_all` / `_ensure_orm_columns_sync` on startup. Tests: `tests/test_knowledge.py`.
+
+### Custom Agent Persistence
+
+Custom agent metadata lives in the `agents` ORM table (`deerflow.persistence.agents`). `agents.id` is the stable runtime id and the filesystem directory name under `.deer-flow/agents/{id}/`; `agents.name` is display-only, can contain Chinese characters, and is not unique. `user_id IS NULL` marks built-in agents, user-owned agents are visible only to their owner unless `published = true`, and update/delete operations require ownership. The runtime accepts `agent_id` in context/configurable, and non-default `assistant_id` values are treated as agent ids after visibility checks.
+
+**Listing cache + pagination** ? to keep `GET /api/agents` fast, the gallery's per-card extras (skills + tool_groups) are denormalized into two cache tables instead of reading every agent's `config.yaml` on each request: `agent_skills` (relational, one row per (agent, skill) with `position`) and `agent_extras` (one row per agent: `tool_groups` JSON + `model` + `skills_synced_at`). `config.yaml` remains authoritative for the agent runtime; these tables are a read cache, write-through on create/update and backfilled from disk by `_sync_legacy_agents` (mtime-vs-`skills_synced_at` drift check). The list endpoint enriches via `AgentStore.get_extras_for([ids])` (two batched `IN` queries). Both tables are auto-created by `Base.metadata.create_all` (no migration needed). `GET /api/agents` supports **opt-in server-side pagination**: `scope` (`mine` = owned + favorited / `square` = built-in/published public), `page`, `page_size`, returning `{agents, total, page, page_size}`; omitting `scope`/`page` preserves the legacy "return everything" behavior. Name `search` matches the raw OR desensitized name (the frontend's static term?abbrev map is mirrored server-side in `deerflow/config/agent_name_desensitize.py`); `tag_ids` is an SQL OR-filter; `owner` keeps only agents published by that user id (the card's ??? ? built-in agents never match), powering the gallery's click-???-to-filter. `POST /api/agents/private-square/list` takes the same pagination params (incl. `owner`). Repository entry point: `AgentStore.list_paginated(...)`. Tests: `tests/test_agent_listing_cache.py::test_list_paginated_owner_filter_keeps_one_publisher`.
+
+**Homepage selector (agent picker on the new-chat page)** ? single admin-curated list backed by `agents.featured_order` (nullable BIGINT). NULL = not featured. Only `published = true` agents can be featured (router enforces on writes); any published agent is eligible regardless of ownership. Cleared with `order: null`. The list rendered to end users is `WHERE featured_order IS NOT NULL` + visibility-filter, sorted by `featured_order ASC`. Admins reorder by sending the desired ids to `PUT /api/agents/featured/order` (assigns indices 0..N-1). BIGINT is used rather than INT so the management UI can seed initial values from `Date.now()` (overflows INT on MySQL).
+
+### Publish Squares (????)
+
+Admin-configurable, **access-control-free** categories that published agents/skills are filed under (distinct from the password-gated ????). The square list lives in `system_settings.json` ? `publish_squares` (`PublishSquaresSettings`: ordered `squares: [{id,name,description}]` + `default_square_id`); helpers `list_publish_squares` / `resolve_default_square_id` / `normalize_square_id` always yield a valid square (built-in "????" fallback when none configured). Admin CRUD: `GET/PUT /api/system-settings/publish-squares` (GET readable by any user ? the gallery sub-tabs + publish picker need it; PUT admin-only, auto-slugs missing ids and re-anchors a dangling default).
+
+Each agent/skill carries a single `square_id` column (`AgentRow` / `SkillRow`, `String(64)` NOT NULL default `''`, migration `20260611_01`, idempotent SQL in `column_additions_mysql.sql`). One column = at most one square, so an item can never be "published twice". `''` = unassigned ? resolved to the default square at read/filter time. Writes normalize via `normalize_square_id`: agents on create/update (`square_id` field), skills on publish (`PUT /api/skills/custom/{name}/published` accepts optional `square_id`). Listing filters: `GET /api/agents?square=` and `GET /api/skills?square=` keep only that square's items; the **default** square additionally sweeps in legacy `''` rows (`square_is_default`). Tests: `tests/test_publish_squares.py`. Frontend: `core/system-settings` (`usePublishSquares`/`useUpdatePublishSquares`), admin `PublishSquaresSettingsPage` (embedded in the appearance/settings hub), square picker in the agent edit dialog + per-skill card, and a ??-tab square sub-filter in both galleries.
+
+### Agent Module Tag Config (zzJC / zzXD 按类型聚合)
+
+Backs the zzJC / zzXD agent-chat left sidebar (「按类型聚合智能体 + 聊天记录」). An admin picks which agent **tags** belong to a module; the module's left sidebar then shows only agents under those tags (empty/unconfigured = show all visible agents). Stored in `system_settings.json` → `agent_modules` (`AgentModuleSettings`: `modules: {module_key → [tag_id, ...]}`); pure helpers `get_module_tag_ids` / `set_module_tag_ids` (de-dupe + strip). Endpoints (in `app/gateway/routers/system_settings.py`): `GET /api/system-settings/agent-modules/{key}` (readable by any user — the sidebar needs it to filter), `PUT /api/system-settings/agent-modules/{key}` (admin only, full replacement of the tag-id list). Module keys are frontend-defined (`AGENT_MODULES` in `components/workspace/agents/agent-module-sidebar.tsx`, keyed by the module's landing agent id: `6bf`→`zzjc`, `10d69…`→`zzxd`). Tests: `tests/test_agent_modules.py`. Frontend: `core/system-settings` (`useAgentModuleConfig`/`useUpdateAgentModuleConfig`), admin `AgentModuleSettingsDialog` (gear button in the agent-chat header).
+
+### 岗位 (Position) Subscription System
+
+An admin-managed aggregation layer that decides, per user, which agents / skills
+/ scheduled tasks / recommended questions they are subscribed to or can see.
+**A user has at most one position** (`users.position_id`, 1:1); unassigned users
+inherit the single `is_default` position. Note: the admin-configured prompts
+shown to users ARE the **recommended questions** (`recommended_questions` table)
+— there is no separate prompt-template store; the per-user `user_custom_prompts`
+(personal 常用提示词 bar) is untouched by this system.
+
+**Persistence** (`deerflow/persistence/positions/`, wired as
+`app.state.position_store`; tables auto-created by `create_all`, migration
+`20260616_01`):
+- `positions` (`PositionRow`): `id`, `name` (unique), `description`,
+  `is_default` (only one at a time — the store clears the flag on others).
+- `position_tag_bindings` (`PositionTagBindingRow`): one row per
+  `(position, resource_type, tag)`. **No binding for a resource_type = "use the
+  default set".** `POSITION_RESOURCE_TYPES = (agent, skill, scheduled_task,
+  recommended_question)` — the same set drives the tag types (`TAG_TYPES` was
+  extended with `recommended_question`).
+
+**Sync engine** (`deerflow/services/position_sync.py`, wired as
+`app.state.position_sync`). `sync_user(user)` **materializes** three resource
+types into the existing "我的" tables, idempotently:
+- agents → `agent_favorites` rows with `origin='position'`
+- **skills → `skill_favorites` rows with `origin='position'`** (table
+  `SkillFavoriteRow`, migration `20260616_02`, auto-created by `create_all`)
+- scheduled tasks → `scheduled_task_subscriptions` rows with `origin='position'`
+
+Skills behave **exactly like agents**: a position only *adds* its tagged skills
+to the member's 我的技能 — the browse-all 广场 is never restricted. (This replaced
+the earlier query-time visibility filter, which wrongly shrank the square to the
+bound skill.) `PositionSyncService.resolve_visible(user, type)` returns
+`set | None` (None = unrestricted) and is now used **only** for the one remaining
+query-time type, `recommended_question`.
+
+Sync only ever adds/removes `origin='position'` rows, so a user's own
+favorites/subscriptions are never clobbered. Default sets (no binding for a
+type): agents = built-ins (`user_id IS NULL`); skills = empty (square stays
+full, nothing added to 我的); scheduled tasks = empty; questions = unrestricted.
+`origin`/`position_id` columns were added to `agent_favorites` and
+`scheduled_task_subscriptions` — see migration `20260616_01` +
+`scripts/sql/column_additions_mysql.sql`. The running app also auto-adds these
+via `_ensure_orm_columns_sync` on startup.
+
+**Triggers**: assigning/removing members, changing tag bindings (re-syncs all
+members), becoming the default position (`sync_default_members`), and login
+(`auth.py::_sync_user_position`, best-effort so a brand-new user inheriting the
+default gets their grants materialized).
+
+**API** (`app/gateway/routers/positions.py`, `/api/positions`, **admin only**):
+position CRUD; `GET/PUT /{id}/tag-bindings`; `GET/POST/DELETE /{id}/members`
+(batch assign/unassign → triggers sync); `POST /{id}/resync`;
+`GET /users` (all users + their position); `PUT /users/{user_id}` (set one
+user's position). **`GET /me` is the one non-admin route** — it returns the
+caller's own effective position (explicit assignment, else the default position
+unassigned users inherit, else `null`); powers the top-bar 岗位 label
+(`strategy-components/Header.tsx` user cluster, via `useMyPosition`). Related
+changes: `GET /api/recommended-questions` is now
+filtered by the caller's position (`?all=1` bypasses for admin management) and
+admin can tag questions via `/api/tags/assignments/recommended_question/{id}`.
+Each listed question now carries its assigned `tags` (bulk-attached via
+`tag_store.list_assignments`), and the list accepts a repeatable `tag_ids`
+OR-filter — so the 推荐问题配置 page can display per-question 标签 and search by
+them (frontend `RecommendedQuestionsPage` + `TagSearchBar`). Tests:
+`tests/test_recommended_questions_tags.py`.
+`GET /api/skills` no longer restricts by position; each skill carries a
+`favorited` flag (true = in the caller's 我的技能, e.g. position-granted) so the
+skills page's 我的 tab includes owned ∪ position-granted skills while the 广场
+stays complete. Frontend: `core/positions`
++ admin `PositionManagementPage` (侧边栏「综合管理 → 岗位管理」 + 右下角设置菜单).
+Recommended questions stay rendered **above** the chat input (their original
+position), just filtered by the caller's position. Tests:
+`tests/test_positions.py`.
+
+### Portable Column Types (`deerflow/persistence/types.py`)
+
+Two `TypeDecorator`s keep ORM columns correct across the `sqlite` / `postgres` / `mysql` backends. **All ORM timestamp/JSON columns must use these, not the raw SQLAlchemy types.**
+
+- **`PortableJSON`** ? native `JSON` on SQLite/Postgres; `TEXT` (with manual `json.dumps`/`loads`) on MySQL, where older servers lack JSON support.
+- **`PortableLongText`** ? a large text column: maps to MySQL `LONGTEXT` (4 GB) and plain `TEXT` on SQLite/Postgres (already unbounded). Use for big JSON/markdown blobs that would overflow MySQL's 64 KB `TEXT` cap ? e.g. `ai_writing_sessions.transcript`.
+- **`BeijingDateTime`** ? stores timestamps as Beijing (UTC+8) wall-clock time. The app records time with `datetime.now(UTC)`, but MySQL `DATETIME` / SQLite have no timezone and persist whatever wall clock the driver serializes ? so a raw `DateTime` column would store UTC and read 8 hours behind Beijing. `BeijingDateTime` converts UTC?Beijing on bind and re-attaches `+08:00` on result, so the database holds Beijing time while Python still gets timezone-aware datetimes (instant comparisons stay correct, including in WHERE clauses). PostgreSQL `timestamptz` stores the instant unambiguously, so values pass through unchanged there. Uses a fixed UTC+8 offset (China has no DST), so no `tzdata` dependency.
+
+### Schema Migrations / Manual Column-Add SQL (`scripts/sql/`)
+
+Alembic auto-migration is **disabled** ? `_run_alembic_upgrade(url)` is commented out in `persistence/engine.py`. On startup only `Base.metadata.create_all` runs, which creates **new tables** but never patches **existing** ones (no new columns, no new indexes). So a column added to an ORM model appears automatically on a brand-new deployment, but is **missing on any existing MySQL database** ? and code reading/writing that column will error there until the column is added by hand.
+
+**Convention (required):** whenever code adds a field to an existing table (a new `mapped_column` on an ORM model), you must, in the same change:
+1. Keep the matching alembic version file under `persistence/migrations/versions/` (guarded with `_has_table`/`_has_column`), and
+2. Append the idempotent `ADD COLUMN` statement to **`scripts/sql/column_additions_mysql.sql`**, with a definition matching the ORM column exactly (type, length, nullability).
+
+Operators upgrading an existing DB run the whole file (`USE <db>;` then execute) ? it is idempotent (existing columns are skipped via `information_schema.columns`), so it is safe to re-run. Index-only changes go in the sibling `scripts/sql/leaderboard_indexes_mysql.sql` (same idempotent pattern).
+
+### Sandbox System (`packages/harness/deerflow/sandbox/`)
+
+**Interface**: Abstract `Sandbox` with `execute_command`, `read_file`, `write_file`, `list_dir`
+**Provider Pattern**: `SandboxProvider` with `acquire`, `get`, `release` lifecycle
+**Implementations**:
+- `LocalSandboxProvider` - Singleton local filesystem execution with path mappings
+- `AioSandboxProvider` (`packages/harness/deerflow/community/`) - Docker-based isolation
+
+**Virtual Path System**:
+- Agent sees: `/mnt/user-data/{workspace,uploads,outputs}`, `/mnt/skills`
+- Physical: `backend/.deer-flow/users/{user_id}/threads/{thread_id}/user-data/...`, `deer-flow/skills/`
+- Translation: `replace_virtual_path()` / `replace_virtual_paths_in_command()`
+- Detection: `is_local_sandbox()` checks `sandbox_id == "local"`
+
+**Sandbox Tools** (in `packages/harness/deerflow/sandbox/tools.py`):
+- `bash` - Execute commands with path translation and error handling
+- `ls` - Directory listing (tree format, max 2 levels)
+- `read_file` - Read file contents with optional line range
+- `write_file` - Write/append to files, creates directories
+- `str_replace` - Substring replacement (single or all occurrences); same-path serialization is scoped to `(sandbox.id, path)` so isolated sandboxes do not contend on identical virtual paths inside one process
+
+### Subagent System (`packages/harness/deerflow/subagents/`)
+
+**Built-in Agents**: `general-purpose` (all tools except `task`) and `bash` (command specialist)
+**Execution**: Dual thread pool - `_scheduler_pool` (3 workers) + `_execution_pool` (3 workers)
+**Concurrency**: `MAX_CONCURRENT_SUBAGENTS = 3` enforced by `SubagentLimitMiddleware` (truncates excess tool calls in `after_model`), 15-minute timeout
+**Flow**: `task()` tool ? `SubagentExecutor` ? background thread ? poll 5s ? SSE events ? result
+**Events**: `task_started`, `task_running`, `task_completed`/`task_failed`/`task_timed_out`
+
+### Tool System (`packages/harness/deerflow/tools/`)
+
+`get_available_tools(groups, include_mcp, model_name, subagent_enabled)` assembles:
+1. **Config-defined tools** - Resolved from `config.yaml` via `resolve_variable()`
+2. **MCP tools** - From enabled MCP servers (lazy initialized, cached with mtime invalidation)
+3. **Built-in tools**:
+   - `present_files` - Make output files visible to user (only `/mnt/user-data/outputs`)
+   - `ask_clarification` - Request clarification (intercepted by ClarificationMiddleware ? interrupts)
+   - `view_image` - Read image as base64 (added only if model supports vision)
+4. **Subagent tool** (if enabled):
+   - `task` - Delegate to subagent (description, prompt, subagent_type, max_turns)
+
+**Community tools** (`packages/harness/deerflow/community/`):
+- `tavily/` - Web search (5 results default) and web fetch (4KB limit)
+- `jina_ai/` - Web fetch via Jina reader API with readability extraction
+- `firecrawl/` - Web scraping via Firecrawl API
+
+**ACP agent tools**:
+- `invoke_acp_agent` - Invokes external ACP-compatible agents from `config.yaml`
+- ACP launchers must be real ACP adapters. The standard `codex` CLI is not ACP-compatible by itself; configure a wrapper such as `npx -y @zed-industries/codex-acp` or an installed `codex-acp` binary
+- Missing ACP executables now return an actionable error message instead of a raw `[Errno 2]`
+- Each ACP agent uses a per-thread workspace at `{base_dir}/users/{user_id}/threads/{thread_id}/acp-workspace/`. The workspace is accessible to the lead agent via the virtual path `/mnt/acp-workspace/` (read-only). In docker sandbox mode, the directory is volume-mounted into the container at `/mnt/acp-workspace` (read-only); in local sandbox mode, path translation is handled by `tools.py`
+- `image_search/` - Image search via DuckDuckGo
+
+### MCP System (`packages/harness/deerflow/mcp/`)
+
+- Uses `langchain-mcp-adapters` `MultiServerMCPClient` for multi-server management
+- **Lazy initialization**: Tools loaded on first use via `get_cached_mcp_tools()`
+- **Cache invalidation**: Detects config file changes via mtime comparison
+- **Transports**: stdio (command-based), SSE, HTTP
+- **OAuth (HTTP/SSE)**: Supports token endpoint flows (`client_credentials`, `refresh_token`) with automatic token refresh + Authorization header injection
+- **Runtime updates**: Gateway API saves to extensions_config.json; LangGraph detects via mtime
+
+### Skills System (`packages/harness/deerflow/skills/`)
+
+- **Location**: `deer-flow/skills/{public,custom}/`
+- **Format**: Directory with `SKILL.md` (YAML frontmatter: name, description, license, allowed-tools)
+- **Loading**: `load_skills()` recursively scans `skills/{public,custom}` for `SKILL.md`, parses metadata, and reads enabled state from extensions_config.json. **A directory containing `SKILL.md` is a self-contained leaf** — discovery does not descend into its subdirectories, so a skill *package* that bundles its own `examples/`/`references/` sub-skills (each with a SKILL.md) surfaces as **one** skill, not a swarm of phantom entries.
+- **Discovery↔delete consistency**: a skill is keyed for discovery by its frontmatter `name`, but its on-disk directory name may differ (hand-dropped / mis-extracted package). `LocalSkillStorage._resolve_skill_dir(name, category)` resolves the *discovered* directory (canonical `<category>/<name>` fast-path, else scan-by-frontmatter-name); `custom_skill_exists`/`read_custom_skill`/`delete_custom_skill` all go through it, so **anything that gets listed can also be read and deleted** regardless of directory name (the old code assumed `custom/<name>/` and 404'd "Custom skill 'x' not found." on any non-canonical layout). Delete is path-guarded to never escape the `custom/` root. Tests: `tests/test_skill_storage_delete.py`.
+- **Injection**: Enabled skills listed in agent system prompt with container paths
+- **Installation**: `POST /api/skills/install` extracts .skill ZIP archive to custom/ directory
+- **Chinese name (`name_zh`)**: a Chinese display name stored in the **`skills` DB table** (`SkillRow.name_zh`, `String(128)`, migration `20260609_01`) ? never written to the on-disk SKILL.md, so it works for built-in skills too. Threaded through `SkillStore` create/ensure_legacy/update/update_admin exactly like `detail`; surfaced on `SkillResponse.name_zh` from the ownership record in `_skill_to_response`. Authorization mirrors `detail`: custom skill ? owner or admin; built-in/public ? admin only (row materialized on demand via `ensure_legacy`). Endpoints (in `app/gateway/routers/skills.py`): `PUT /api/skills/{name}/name-zh` (set/clear), `POST /api/skills/{name}/name-zh/generate` (LLM generates one), `POST /api/skills/name-zh/generate-missing` (????: **admin-only** bulk-generate for every skill that lacks one, bounded-concurrency `asyncio.gather`). **Optional model selection**: both generate endpoints accept an optional JSON body `{model_name}` ? an explicit name wins, else the curator model setting, else the main model (`_resolve_skill_name_zh_model_name`). **Model-error surfacing**: `_generate_skill_name_zh` raises `SkillNameZhModelError` when the model itself fails (build/timeout/API error) ? distinct from the model returning an unusable name (? `None`). The single endpoint maps that to `502 ?????,?????????:?`; the bulk endpoint fails fast on a model-build error (502) and otherwise carries a representative `model_error` string in `SkillNameZhBulkResponse` so the UI can say "N ?????????". The frontend (`skill-settings-page.tsx` list + `skill-detail-dialog.tsx` editor, plus the agent skill pickers/badges in `NewAgentPage.tsx` + `agent-card.tsx`) shows `name_zh || name`, offers per-skill "AI ??" plus an **admin-only** header "???????" button, and a compact model dropdown (`components/workspace/model-name-select.tsx`, default = configured model) beside each action.
+
+#### Skill Compression (技能压缩)
+
+When the skill catalogue grows large, injecting every enabled skill full-text
+into the lead agent's system prompt bloats it and wastes tokens. **Skill
+compression** is an admin-gated mode that keeps only *常驻(免压缩)* skills in the
+prompt body and surfaces the rest on demand via keyword retrieval.
+
+- **Master switch** — `system_settings.json → skill_compression`
+  (`SkillCompressionSettings`: `enabled` default **false**, `search_top_n`
+  default 5, `keep_index_in_prompt` default true). Read/written by
+  `GET`/`PUT /api/system-settings/skill-compression` (GET open to any user — the
+  skill detail page reads it; PUT admin-only, refreshes the prompt cache).
+- **Per-skill 常驻 flag** — stored in the **DB**: `SkillRow.always_on`
+  (`Boolean`, migration `20260616_03`, idempotent SQL in
+  `column_additions_mysql.sql`), **not** in `extensions_config.json` (the
+  source-tree config file is read-only / unsafe to rewrite in some deployments,
+  and built-in skills carry no file entry). Threaded through `SkillStore`
+  create/ensure_legacy/update/update_admin + `list_always_on_names()` exactly
+  like `name_zh`/`detail`; surfaced on `SkillResponse.always_on` from the
+  ownership record in `_skill_to_response`. Toggled via
+  `PUT /api/skills/{name}/always-on` (custom skill → owner or admin; built-in /
+  public → admin only, materializes a row on demand via `ensure_legacy`). The
+  prompt builder stamps `Skill.always_on` from `list_always_on_names()` via a
+  sync DB bridge (`_apply_always_on` in `lead_agent/prompt.py`).
+- **Prompt split** — `get_skills_prompt_section` (`lead_agent/prompt.py`) reads
+  the compression settings (skipped for restricted custom agents — they already
+  have a curated allowlist). When **off**, behaviour is unchanged (all enabled
+  skills full-text). When **on**, `always_on` skills render as full `<skill>`
+  blocks; the rest are dropped from the body, replaced by a `search_skills`
+  usage note plus (when `keep_index_in_prompt`) a name-only `<compressed_skills>`
+  index. The `_get_cached_skills_prompt_section` cache key includes the
+  `always_on` flag + compression flags so changes invalidate correctly.
+- **Retrieval tool** — `search_skills` (`tools/builtins/skill_tools.py`,
+  `search_skills_impl`) keyword-scores `name + description` over enabled,
+  visible, non-archived skills (same visibility filter as `skill_list`) and
+  returns top-N `{name, description, category, location}`. It is registered in
+  `get_available_tools` **only when compression is enabled** (offline,
+  deterministic, zero external deps). The agent then `read_file`s the returned
+  `location`.
+
+Tests: `tests/test_skill_compression.py`.
+
+#### Skill Curator (`packages/harness/deerflow/skills/curator.py`)
+
+Background skill lifecycle manager. The Gateway `lifespan` starts a resident
+scheduler task (`_curator_scheduler_loop` in `app/gateway/app.py`) that ? after a
+5-minute startup delay ? calls `maybe_run_curator()` every hour. A cross-process
+file lock (`.curator_scheduler.lock`) elects a single leader worker so the
+scheduler runs once per deployment. Per run:
+
+- **Master switch** ? `skill_evolution.enabled` (default **`false`**) gates the
+  whole feature; `curator.enabled` (default **`false`**) gates the curator
+  specifically. Both must be `true` for the curator to act. `should_run_now()`
+  and `_curator_thread_target()` both enforce this ? so manual `POST
+  /api/curator/run` also respects the master switch. Config is
+  `config.yaml.skill_evolution` deep-merged with per-user
+  `users/{user_id}/skill_evolution.json`.
+- **Per-user run lock** ? `.curator_state` carries `running` / `started_at`;
+  `_try_acquire_run_lock()` prevents the same user's curator from running
+  concurrently (scheduler vs. manual trigger, or double manual clicks). A lock
+  whose `started_at` is older than `_CURATOR_STALE_LOCK_HOURS` (6h) is treated as
+  process-crash residue and can be taken over.
+- **Concurrency cap** ? `maybe_run_curator()` collects all due users and hands
+  them to `_dispatch_curator_runs()`, which runs them through a
+  `ThreadPoolExecutor` capped at `_CURATOR_MAX_CONCURRENCY` (5) ? avoids a
+  thundering herd of LLM calls when many users come due at once (e.g. first
+  deploy).
+- **Failure back-off** ? `_release_run_lock()` always stamps `last_attempt_at`.
+  When a first run keeps crashing (`last_run_at` never set), `should_run_now()`
+  retries on `_CURATOR_RETRY_HOURS` (6h) instead of every hourly tick.
+- **`should_run_now()`** is the real gate: the hourly tick only checks "is the
+  per-user `interval_hours` (default 168h) elapsed". Timestamp parsing tolerates
+  naive (tz-less) strings ? a corrupt `last_run_at` triggers a re-run rather than
+  a permanent deadlock.
+- **Published-skill filter** ? `SkillStore.list_published_names()` is a batch
+  query; the curator fetches the published-name set once per run instead of one
+  DB round-trip per candidate skill.
+- **LLM review timeout** ? `_run_llm_review()` runs `agent.invoke()` in a worker
+  thread bounded by `curator.review_timeout_seconds` (default 600s). The model's
+  per-request `request_timeout` can't bound a multi-turn React agent loop, so on
+  timeout the run is abandoned (a warning is written to `curator_log`, visible in
+  the Web UI) and the curator finishes cleanly + releases the run lock. The
+  worker context is copied via `contextvars.copy_context()` so tool calls still
+  resolve the correct `user_id`.
+- **Review model** ? `_run_llm_review()` uses `curator.model_name` (a
+  `config.yaml` `models[]` entry name) if set, else the primary model; it is
+  editable from the Web UI evolution panel. The skill security scanner
+  (`scan_skill_content` in `security_scanner.py`) reuses the same
+  `skill_evolution.curator.model_name` setting ? review and moderation share one
+  model. There is no separate moderation-model config field.
+
+### Model Factory (`packages/harness/deerflow/models/factory.py`)
+
+- `create_chat_model(name, thinking_enabled)` instantiates LLM from config via reflection
+- Supports `thinking_enabled` flag with per-model `when_thinking_enabled` overrides
+- Supports vLLM-style thinking toggles via `when_thinking_enabled.extra_body.chat_template_kwargs.enable_thinking` for Qwen reasoning models, while normalizing legacy `thinking` configs for backward compatibility
+- Supports `supports_vision` flag for image understanding models
+- Config values starting with `$` resolved as environment variables
+- Missing provider modules surface actionable install hints from reflection resolvers (for example `uv add langchain-google-genai`)
+
+### vLLM Provider (`packages/harness/deerflow/models/vllm_provider.py`)
+
+- `VllmChatModel` subclasses `langchain_openai:ChatOpenAI` for vLLM 0.19.0 OpenAI-compatible endpoints
+- Preserves vLLM's non-standard assistant `reasoning` field on full responses, streaming deltas, and follow-up tool-call turns
+- Designed for configs that enable thinking through `extra_body.chat_template_kwargs.enable_thinking` on vLLM 0.19.0 Qwen reasoning models, while accepting the older `thinking` alias
+
+### IM Channels System (`app/channels/`)
+
+Bridges external messaging platforms (Feishu, Slack, Telegram, DingTalk) to the DeerFlow agent via the LangGraph Server.
+
+
+**Architecture**: Channels communicate with Gateway through the `langgraph-sdk` HTTP client (same as the frontend), ensuring threads are created and managed server-side. The internal SDK client injects process-local internal auth plus a matching CSRF cookie/header pair so Gateway accepts state-changing thread/run requests from channel workers without relying on browser session cookies.
+
+**Components**:
+- `message_bus.py` - Async pub/sub hub (`InboundMessage` ? queue ? dispatcher; `OutboundMessage` ? callbacks ? channels)
+- `store.py` - JSON-file persistence mapping `channel_name:chat_id[:topic_id]` ? `thread_id` (keys are `channel:chat` for root conversations and `channel:chat:topic` for threaded conversations)
+- `manager.py` - Core dispatcher: creates threads via `client.threads.create()`, routes commands, keeps Slack/Telegram on `client.runs.wait()`, and uses `client.runs.stream(["messages-tuple", "values"])` for Feishu incremental outbound updates
+- `base.py` - Abstract `Channel` base class (start/stop/send lifecycle)
+- `service.py` - Manages lifecycle of all configured channels from `config.yaml`
+- `slack.py` / `feishu.py` / `telegram.py` / `dingtalk.py` - Platform-specific implementations (`feishu.py` tracks the running card `message_id` in memory and patches the same card in place; `dingtalk.py` optionally uses AI Card streaming for in-place updates when `card_template_id` is configured)
+
+**Message Flow**:
+1. External platform -> Channel impl -> `MessageBus.publish_inbound()`
+2. `ChannelManager._dispatch_loop()` consumes from queue
+3. For chat: look up/create thread through Gateway's LangGraph-compatible API
+4. Feishu chat: `runs.stream()` ? accumulate AI text ? publish multiple outbound updates (`is_final=False`) ? publish final outbound (`is_final=True`)
+5. Slack/Telegram chat: `runs.wait()` ? extract final response ? publish outbound
+6. Feishu channel sends one running reply card up front, then patches the same card for each outbound update (card JSON sets `config.update_multi=true` for Feishu's patch API requirement)
+7. DingTalk AI Card mode (when `card_template_id` configured): `runs.stream()` ? create card with initial text ? stream updates via `PUT /v1.0/card/streaming` ? finalize on `is_final=True`. Falls back to `sampleMarkdown` if card creation or streaming fails
+8. For commands (`/new`, `/status`, `/models`, `/memory`, `/help`): handle locally or query Gateway API
+9. Outbound ? channel callbacks ? platform reply
+
+**Configuration** (`config.yaml` -> `channels`):
+- `langgraph_url` - LangGraph-compatible Gateway API base URL (default: `http://localhost:8001/api`)
+- `gateway_url` - Gateway API URL for auxiliary commands (default: `http://localhost:8001`)
+- In Docker Compose, IM channels run inside the `gateway` container, so `localhost` points back to that container. Use `http://gateway:8001/api` for `langgraph_url` and `http://gateway:8001` for `gateway_url`, or set `DEER_FLOW_CHANNELS_LANGGRAPH_URL` / `DEER_FLOW_CHANNELS_GATEWAY_URL`.
+- Per-channel configs: `feishu` (app_id, app_secret), `slack` (bot_token, app_token), `telegram` (bot_token), `dingtalk` (client_id, client_secret, optional `card_template_id` for AI Card streaming)
+
+
+### Memory System ? Memory V2 (`packages/harness/deerflow/agents/memory/`)
+
+Pluggable memory architecture (Hermes-style). See `docs/MEMORY_V2_DESIGN_ZH.md`
+and `docs/HINDSIGHT_SETUP_ZH.md` for the full design.
+
+**Components**:
+- `provider.py` - `MemoryProvider` abstract base (12 lifecycle/hook methods)
+- `manager.py` - `MemoryManager` orchestrates builtin + ?1 external provider; `build_memory_manager()` is the construction entry point
+- `providers/builtin.py` - `BuiltinFileProvider`: local USER.md / MEMORY.md, ? delimiter, char quotas, file locking, atomic writes, frozen snapshot
+- `providers/hindsight.py` - `HindsightProvider`: HTTP client to a self-hosted Hindsight Docker instance, dual-bank routing
+- `security.py` - injection/exfiltration scan + invisible-unicode block for memory content
+- `message_processing.py` - `filter_messages_for_memory` (last-turn extraction for Hindsight retain)
+
+**Two stores (binary classification)**:
+- `USER.md` at `{base_dir}/users/{user_id}/USER.md` ? user profile, shared across agents
+- `MEMORY.md` at `{base_dir}/users/{user_id}/agents/{agent_id}/MEMORY.md` ? agent-private working notes
+- `user_id` resolved via `get_effective_user_id()`; `"default"` in no-auth mode
+
+**Write paths**:
+- LLM actively calls the `memory` tool (`tools/builtins/memory_tool.py`) ? writes USER.md/MEMORY.md, mirrors to Hindsight
+- LLM calls `hindsight_recall`/`reflect`/`retain` tools (`tools/builtins/hindsight_tools.py`) ? only in `tools`/`hybrid` memory_mode
+- Hindsight `auto_retain` ? every turn auto-stored to the work bank (via `MemoryMiddleware.after_agent`)
+
+**Injection**:
+- builtin frozen snapshot ? injected into the system prompt via `lead_agent/prompt.py` `_get_memory_context`
+- Hindsight dynamic recall ? injected per model call via `MemoryMiddleware.wrap_model_call` (transient ? not persisted to state, so it never leaks to the frontend)
+
+**Hindsight dual-bank isolation** (B-scheme):
+- global user bank `deerflow-user-{user_id}` ? cross-agent
+- agent work bank `deerflow-work-{user_id}-{agent_id}` ? per-agent
+- `hindsight-client` is an optional dependency; when missing, `HindsightProvider.is_available()` returns False and the system falls back to builtin-only
+
+**Migration**: `PYTHONPATH=. python scripts/migrate_memory_to_v2.py [--all|--user-id X] [--dry-run]` converts legacy `memory.json` into USER.md/MEMORY.md.
+
+**Configuration** (`config.yaml` ? `memory`, nested): `enabled`, `provider` (builtin/hindsight),
+`builtin.{memory_char_limit,user_char_limit,...}`, `hindsight.{mode,api_url,api_key,bank_id_template,work_bank_id_template,memory_mode,...}`,
+`injection.{enabled,max_tokens,include_builtin,include_hindsight,context_tag}`, `security.{scan_content,block_invisible_unicode}`.
+Per-user runtime overrides (whitelisted fields) live at `{base_dir}/users/{user_id}/memory_config.json` ? see `get_effective_memory_config()`.
+
+**Gateway API** ? see the Memory router below (`/api/memory/v2*`).
+
+### Reflection System (`packages/harness/deerflow/reflection/`)
+
+- `resolve_variable(path)` - Import module and return variable (e.g., `module.path:variable_name`)
+- `resolve_class(path, base_class)` - Import and validate class against base class
+
+### Config Schema
+
+**`config.yaml`** key sections:
+- `models[]` - LLM configs with `use` class path, `supports_thinking`, `supports_vision`, provider-specific fields
+- vLLM reasoning models should use `deerflow.models.vllm_provider:VllmChatModel`; for Qwen-style parsers prefer `when_thinking_enabled.extra_body.chat_template_kwargs.enable_thinking`, and DeerFlow will also normalize the older `thinking` alias
+- `tools[]` - Tool configs with `use` variable path and `group`
+- `tool_groups[]` - Logical groupings for tools
+- `sandbox.use` - Sandbox provider class path
+- `skills.path` / `skills.container_path` - Host and container paths to skills directory
+- `title` - Auto-title generation (enabled, max_words, max_chars, prompt_template)
+- `summarization` - Context summarization (enabled, trigger conditions, keep policy)
+- `subagents.enabled` - Master switch for subagent delegation
+- `memory` - Memory system (enabled, storage_path, debounce_seconds, model_name, max_facts, fact_confidence_threshold, injection_enabled, max_injection_tokens)
+- `auth_login` - Custom login modes layered on the base auth (`deerflow/config/login_config.py`):
+  - `username_login.{require_password, password}` - Password gate for `POST /api/v1/auth/login/username`. When `require_password=true`, the (otherwise passwordless) username login additionally requires `?password=<shared-secret>`; empty `password` falls back to `123ewq`. Gate only ? it does not change which account logs in.
+  - `token_login.{enabled, token_info_url, service_authorization, timeout_seconds, verify_ssl, ca_cert_path}` - Token-exchange login (`POST /api/v1/auth/login/token`). The gateway POSTs to `token_info_url` with the caller-supplied token placed directly in the `Authorization: Bearer <token>` header (no cookie; the static `service_authorization` field is deprecated/unused, kept only for backward compat), reads `data.username` from a `{"state":"200",...}` body, then signs the matching local user in (auto-registering if absent). `token_info_url` is HTTPS ? `verify_ssl: false` skips cert verification for self-signed/internal UAT hosts, or set `ca_cert_path` to a PEM bundle (takes precedence, keeps verification on). See [docs/AUTH_LOGIN.md](docs/AUTH_LOGIN.md) for the full request/response contract.
+
+Both endpoints are whitelisted in `auth_middleware._PUBLIC_EXACT_PATHS`. The shared `_get_or_create_user_by_username()` helper (in `routers/auth.py`) backs both `/login/username` and `/login/token`.
+
+- `log_system` - Admin-only external log-system entry (`deerflow/config/log_system_config.py`): `{enabled, url, label}`. Surfaced read-only by `GET /api/system-settings/log-system` (admin only); the frontend renders it as a "????" item in the bottom-right settings menu (`workspace-nav-menu.tsx`) that opens `url` in a new tab. Disabled by default.
+
+**`extensions_config.json`**:
+- `mcpServers` - Map of server name ? config (enabled, type, command, args, env, url, headers, oauth, description)
+- `skills` - Map of skill name ? state (enabled). 常驻(免压缩)/always_on lives in the `skills` DB table, not here.
+
+Both can be modified at runtime via Gateway API endpoints or `DeerFlowClient` methods.
+
+### Embedded Client (`packages/harness/deerflow/client.py`)
+
+`DeerFlowClient` provides direct in-process access to all DeerFlow capabilities without HTTP services. All return types align with the Gateway API response schemas, so consumer code works identically in HTTP and embedded modes.
+
+**Architecture**: Imports the same `deerflow` modules that Gateway API uses. Shares the same config files and data directories. No FastAPI dependency.
+
+**Agent Conversation**:
+- `chat(message, thread_id)` ? synchronous, accumulates streaming deltas per message-id and returns the final AI text
+- `stream(message, thread_id)` ? subscribes to LangGraph `stream_mode=["values", "messages", "custom"]` and yields `StreamEvent`:
+  - `"values"` ? full state snapshot (title, messages, artifacts); AI text already delivered via `messages` mode is **not** re-synthesized here to avoid duplicate deliveries
+  - `"messages-tuple"` ? per-chunk update: for AI text this is a **delta** (concat per `id` to rebuild the full message); tool calls and tool results are emitted once each
+  - `"custom"` ? forwarded from `StreamWriter`
+  - `"end"` ? stream finished (carries cumulative `usage` counted once per message id)
+- Agent created lazily via `create_agent()` + `_build_middlewares()`, same as `make_lead_agent`
+- Supports `checkpointer` parameter for state persistence across turns
+- `reset_agent()` forces agent recreation (e.g. after memory or skill changes)
+- See [docs/STREAMING.md](docs/STREAMING.md) for the full design: why Gateway and DeerFlowClient are parallel paths, LangGraph's `stream_mode` semantics, the per-id dedup invariants, and regression testing strategy
+
+**Gateway Equivalent Methods** (replaces Gateway API):
+
+| Category | Methods | Return format |
+|----------|---------|---------------|
+| Models | `list_models()`, `get_model(name)` | `{"models": [...]}`, `{name, display_name, ...}` |
+| MCP | `get_mcp_config()`, `update_mcp_config(servers)` | `{"mcp_servers": {...}}` |
+| Skills | `list_skills()`, `get_skill(name)`, `update_skill(name, enabled)`, `install_skill(path)` | `{"skills": [...]}` |
+| Memory | `get_memory()`, `reload_memory()`, `get_memory_config()`, `get_memory_status()` | dict |
+| Uploads | `upload_files(thread_id, files)`, `list_uploads(thread_id)`, `delete_upload(thread_id, filename)` | `{"success": true, "files": [...]}`, `{"files": [...], "count": N}` |
+| Artifacts | `get_artifact(thread_id, path)` ? `(bytes, mime_type)` | tuple |
+
+**Key difference from Gateway**: Upload accepts local `Path` objects instead of HTTP `UploadFile`, rejects directory paths before copying, and reuses a single worker when document conversion must run inside an active event loop. Artifact returns `(bytes, mime_type)` instead of HTTP Response. The new Gateway-only thread cleanup route deletes `.deer-flow/threads/{thread_id}` after LangGraph thread deletion; there is no matching `DeerFlowClient` method yet. `update_mcp_config()` and `update_skill()` automatically invalidate the cached agent.
+
+**Tests**: `tests/test_client.py` (77 unit tests including `TestGatewayConformance`), `tests/test_client_live.py` (live integration tests, requires config.yaml)
+
+**Gateway Conformance Tests** (`TestGatewayConformance`): Validate that every dict-returning client method conforms to the corresponding Gateway Pydantic response model. Each test parses the client output through the Gateway model ? if Gateway adds a required field that the client doesn't provide, Pydantic raises `ValidationError` and CI catches the drift. Covers: `ModelsListResponse`, `ModelResponse`, `SkillsListResponse`, `SkillResponse`, `SkillInstallResponse`, `McpConfigResponse`, `UploadResponse`, `MemoryConfigResponse`, `MemoryStatusResponse`.
+
+## Development Workflow
+
+### Test-Driven Development (TDD) ? MANDATORY
+
+**Every new feature or bug fix MUST be accompanied by unit tests. No exceptions.**
+
+- Write tests in `backend/tests/` following the existing naming convention `test_<feature>.py`
+- Run the full suite before and after your change: `make test`
+- Tests must pass before a feature is considered complete
+- For lightweight config/utility modules, prefer pure unit tests with no external dependencies
+- If a module causes circular import issues in tests, add a `sys.modules` mock in `tests/conftest.py` (see existing example for `deerflow.subagents.executor`)
+
+```bash
+# Run all tests
+make test
+
+# Run a specific test file
+PYTHONPATH=. uv run pytest tests/test_<feature>.py -v
+```
+
+### Running the Full Application
+
+From the **project root** directory:
+```bash
+make dev
+```
+
+This starts all services and makes the application available at `http://localhost:2026`.
+
+**All startup modes:**
+
+| | **Local Foreground** | **Local Daemon** | **Docker Dev** | **Docker Prod** |
+|---|---|---|---|---|
+| **Dev** | `./scripts/serve.sh --dev`<br/>`make dev` | `./scripts/serve.sh --dev --daemon`<br/>`make dev-daemon` | `./scripts/docker.sh start`<br/>`make docker-start` | ? |
+| **Prod** | `./scripts/serve.sh --prod`<br/>`make start` | `./scripts/serve.sh --prod --daemon`<br/>`make start-daemon` | ? | `./scripts/deploy.sh`<br/>`make up` |
+
+| Action | Local | Docker Dev | Docker Prod |
+|---|---|---|---|
+| **Stop** | `./scripts/serve.sh --stop`<br/>`make stop` | `./scripts/docker.sh stop`<br/>`make docker-stop` | `./scripts/deploy.sh down`<br/>`make down` |
+| **Restart** | `./scripts/serve.sh --restart [flags]` | `./scripts/docker.sh restart` | ? |
+
+**Nginx routing**:
+- `/api/langgraph/*` ? Gateway embedded runtime (8001), rewritten to `/api/*`
+- `/api/*` (other) ? Gateway API (8001)
+- `/` (non-API) ? Frontend (3000)
+
+### Running Backend Services Separately
+
+From the **backend** directory:
+
+```bash
+# Gateway API
+make gateway
+```
+
+Direct access (without nginx):
+- Gateway: `http://localhost:8001`
+
+### Frontend Configuration
+
+The frontend uses environment variables to connect to backend services:
+- `NEXT_PUBLIC_LANGGRAPH_BASE_URL` - Defaults to `/api/langgraph` (through nginx)
+- `NEXT_PUBLIC_BACKEND_BASE_URL` - Defaults to empty string (through nginx)
+
+When using `make dev` from root, the frontend automatically connects through nginx.
+
+## Key Features
+
+### File Upload
+
+Multi-file upload with automatic document conversion:
+- Endpoint: `POST /api/threads/{thread_id}/uploads`
+- Supports: PDF, PPT, Excel, Word documents (converted via `markitdown`)
+- Rejects directory inputs before copying so uploads stay all-or-nothing
+- Reuses one conversion worker per request when called from an active event loop
+- Files stored in thread-isolated directories
+- Agent receives uploaded file list via `UploadsMiddleware`
+
+See [docs/FILE_UPLOAD.md](docs/FILE_UPLOAD.md) for details.
+
+### Plan Mode
+
+TodoList middleware for complex multi-step tasks:
+- Controlled via runtime config: `config.configurable.is_plan_mode = True`
+- Provides `write_todos` tool for task tracking
+- One task in_progress at a time, real-time updates
+
+See [docs/plan_mode_usage.md](docs/plan_mode_usage.md) for details.
+
+### Context Summarization
+
+Automatic conversation summarization when approaching token limits:
+- **On/off switch is per-request, not the config file.** The chat input box has a
+  「上下文压缩」 toggle (below 参考文献, **default off**) → frontend sends
+  `context.summarization_enabled` → whitelisted in `_CONTEXT_CONFIGURABLE_KEYS`
+  (`app/gateway/services.py`) → `_build_middlewares` reads it from the run config
+  and calls `_create_summarization_middleware(force_enabled=summarization_enabled)`.
+  `force_enabled=True` mounts the middleware even when `config.summarization.enabled`
+  is false; when the flag is absent (channels / scheduled runs) it falls back to
+  the config flag. `config.summarization.enabled` now defaults to **false**.
+- `config.yaml` `summarization` still supplies the **tuning**: trigger types
+  (tokens, messages, or fraction of max input) and the keep policy.
+- Keeps recent messages while summarizing older ones.
+
+Tests: `tests/test_summarization_toggle.py` (gating + whitelist),
+`tests/test_summarization_middleware.py` (cache behavior).
+See [docs/summarization.md](docs/summarization.md) for details.
+
+### Vision Support / Image Recognition (??)
+
+Two paths, selected by `config.yaml` ? `vision`:
+
+**Dedicated vision model (delegated recognition)** ? set `vision.model_name` to a model from `models[]`. The `view_image` tool sends the image to that model and returns its **text description** as the tool result, so a text-only main chat model can still "see" images. `view_image` accepts an optional `question` argument for targeted queries; the default prompt is `vision.prompt` (overridable), and `vision.max_tokens` caps the recognition output. Resolution/gating lives in `deerflow/models/vision.py` (`resolve_vision_model_name`, `vision_enabled`, `is_delegated_vision`, `recognize_image`).
+
+**Native vision (fallback)** ? when `vision.model_name` is unset, behavior falls back to the main chat model, which only works if it has `supports_vision: true`:
+- `ViewImageMiddleware` injects base64 image data into the conversation before the LLM call
+- `view_image` stores the image in `state.viewed_images` for that middleware
+
+Both the `view_image` tool and `ViewImageMiddleware` are registered whenever `vision_enabled(model_name)` is true (dedicated model configured **or** main model supports vision). Config schema: `deerflow/config/vision_config.py` (`VisionConfig`); tests: `tests/test_vision.py`.
+
+### Upload symlink-safety (cross-platform)
+
+`deerflow/uploads/manager.py:open_upload_file_no_symlink` writes uploads without following a pre-existing destination symlink. `O_NOFOLLOW` is applied where available (POSIX) but is **optional** ? Windows lacks `os.O_NOFOLLOW`/`O_NONBLOCK`, so requiring it previously made every Windows upload fail with `UnsafeUploadPathError`. Pre-existing symlinks are still rejected via `lstat`, and the opened fd is re-checked with `fstat`. `O_BINARY` is added on Windows. Regression: `tests/test_uploads_windows_safe.py`.
+
+## Code Style
+
+- Uses `ruff` for linting and formatting
+- Line length: 240 characters
+- Python 3.12+ with type hints
+- Double quotes, space indentation
+
+## Documentation
+
+See `docs/` directory for detailed documentation:
+- [CONFIGURATION.md](docs/CONFIGURATION.md) - Configuration options
+- [ARCHITECTURE.md](docs/ARCHITECTURE.md) - Architecture details
+- [API.md](docs/API.md) - API reference
+- [SETUP.md](docs/SETUP.md) - Setup guide
+- [FILE_UPLOAD.md](docs/FILE_UPLOAD.md) - File upload feature
+- [PATH_EXAMPLES.md](docs/PATH_EXAMPLES.md) - Path types and usage
+- [summarization.md](docs/summarization.md) - Context summarization
+- [plan_mode_usage.md](docs/plan_mode_usage.md) - Plan mode with TodoList
+- [BACKEND_RUN_STOP_ZH.md](docs/BACKEND_RUN_STOP_ZH.md) - Windows-first backend start/stop and troubleshooting guide
diff --git a/offline-backend-20260512/backend/app/gateway/app.py b/offline-backend-20260512/backend/app/gateway/app.py
index 622280af..bae7f335 100644
--- a/offline-backend-20260512/backend/app/gateway/app.py
+++ b/offline-backend-20260512/backend/app/gateway/app.py
@@ -50,6 +50,7 @@ from app.gateway.routers import (
     notifications,
     parallel_agent_tasks,
     parallel_agents,
+    positions,
     public_embed,
     public_leaderboard,
     public_scheduled_tasks,
@@ -1409,6 +1410,9 @@ This gateway provides custom endpoints for models, MCP configuration, skills, an
     # Tags API is mounted at /api/tags
     app.include_router(tags.router)
 
+    # 岗位 (position) management API is mounted at /api/positions
+    app.include_router(positions.router)
+
     # Thread Runs API (LangGraph Platform-compatible runs lifecycle)
     app.include_router(thread_runs.router)
 
diff --git a/offline-backend-20260512/backend/app/gateway/deps.py b/offline-backend-20260512/backend/app/gateway/deps.py
index 7813ffc2..e403a7cc 100644
--- a/offline-backend-20260512/backend/app/gateway/deps.py
+++ b/offline-backend-20260512/backend/app/gateway/deps.py
@@ -144,6 +144,14 @@ async def langgraph_runtime(app: FastAPI) -> AsyncGenerator[None, None]:
 
         app.state.tag_store = make_tag_store(sf)
 
+        from deerflow.persistence.positions import make_position_store
+
+        app.state.position_store = make_position_store(sf)
+
+        from deerflow.services.position_sync import PositionSyncService
+
+        app.state.position_sync = PositionSyncService(sf, app.state.position_store, app.state.tag_store)
+
         from deerflow.persistence.thread_shares import make_thread_share_store
 
         app.state.thread_share_store = make_thread_share_store(sf)
@@ -168,6 +176,10 @@ async def langgraph_runtime(app: FastAPI) -> AsyncGenerator[None, None]:
 
         app.state.article_type_store = make_article_type_store(sf)
 
+        from deerflow.persistence.research_templates import make_research_template_store
+
+        app.state.research_template_store = make_research_template_store(sf)
+
         from deerflow.persistence.memory_config import make_user_memory_config_store
 
         app.state.memory_config_store = make_user_memory_config_store(sf)
@@ -176,6 +188,14 @@ async def langgraph_runtime(app: FastAPI) -> AsyncGenerator[None, None]:
 
         app.state.knowledge_service = make_knowledge_service(sf, config.knowledge)
 
+        # Seed the built-in extraction prompts as editable templates so operators
+        # can immediately view/edit them (idempotent; never overwrites edits).
+        if app.state.knowledge_service is not None:
+            try:
+                await app.state.knowledge_service.seed_default_templates()
+            except Exception:
+                logger.exception("Failed to seed knowledge extraction templates")
+
         # Background auto-sedimentation queue (phase 3) — drains finished
         # conversations off the chat path. Worker only does work when
         # ``knowledge.auto_ingest_enabled`` is set; the queue is always wired so
@@ -389,6 +409,22 @@ def get_tag_store(request: Request) -> TagStore:
     return val
 
 
+def get_position_store(request: Request):
+    """Return the 岗位 (position) store."""
+    val = getattr(request.app.state, "position_store", None)
+    if val is None:
+        raise HTTPException(status_code=503, detail="Position store not available")
+    return val
+
+
+def get_position_sync(request: Request):
+    """Return the position → subscription sync service."""
+    val = getattr(request.app.state, "position_sync", None)
+    if val is None:
+        raise HTTPException(status_code=503, detail="Position sync service not available")
+    return val
+
+
 def get_thread_share_store(request: Request):
     """Return the conversation share-link store."""
     val = getattr(request.app.state, "thread_share_store", None)
@@ -423,6 +459,11 @@ def get_article_type_store(request: Request):
     return getattr(request.app.state, "article_type_store", None)
 
 
+def get_research_template_store(request: Request):
+    """Return the research template store, or None when not configured."""
+    return getattr(request.app.state, "research_template_store", None)
+
+
 def get_memory_config_store(request: Request):
     """Return the per-user memory config store."""
     val = getattr(request.app.state, "memory_config_store", None)
diff --git a/offline-backend-20260512/backend/app/gateway/routers/ai_writing.py b/offline-backend-20260512/backend/app/gateway/routers/ai_writing.py
index 83475e64..ab11de2c 100644
--- a/offline-backend-20260512/backend/app/gateway/routers/ai_writing.py
+++ b/offline-backend-20260512/backend/app/gateway/routers/ai_writing.py
@@ -90,6 +90,48 @@ class ArticleTypeResponse(BaseModel):
     updated_at: str
 
 
+class ResearchTemplateRequest(BaseModel):
+    name: str
+    description: Optional[str] = None
+    article_type_key: Optional[str] = None
+    outline: Optional[str] = None
+    material_source: str = "general"
+    word_count: int = 800
+    audience: Optional[str] = None
+    strict_mode: bool = False
+    topic: Optional[str] = None
+    sort_order: int = 0
+
+
+class UpdateResearchTemplateRequest(BaseModel):
+    name: Optional[str] = None
+    description: Optional[str] = None
+    article_type_key: Optional[str] = None
+    outline: Optional[str] = None
+    material_source: Optional[str] = None
+    word_count: Optional[int] = None
+    audience: Optional[str] = None
+    strict_mode: Optional[bool] = None
+    topic: Optional[str] = None
+    sort_order: Optional[int] = None
+
+
+class ResearchTemplateResponse(BaseModel):
+    id: str
+    name: str
+    description: Optional[str] = None
+    article_type_key: Optional[str] = None
+    outline: Optional[str] = None
+    material_source: str
+    word_count: int
+    audience: Optional[str] = None
+    strict_mode: bool
+    topic: Optional[str] = None
+    sort_order: int
+    created_at: str
+    updated_at: str
+
+
 class SessionResponse(BaseModel):
     id: str
     user_id: Optional[str] = None
@@ -268,6 +310,94 @@ async def delete_article_type(article_type_id: str, request: Request):
     return {"success": True}
 
 
+# ── 研究模板（课题研究左栏一键预填表单，需求 0616 #5）────────────────────────────
+
+
+@router.get("/research-templates", response_model=list[ResearchTemplateResponse])
+async def list_research_templates(request: Request):
+    """列出所有研究模板，按 sort_order 升序。
+
+    新机器/空表时自动写入内置范例模板（网摘/周报/要讯），幂等仅在表为空时执行。
+    """
+    from app.gateway.deps import get_research_template_store
+
+    store = get_research_template_store(request)
+    if store is None:
+        return []
+    try:
+        await store.seed_defaults()
+    except Exception:
+        logger.exception("ai_writing: failed to seed default research templates")
+    return await store.list()
+
+
+@router.post("/research-templates", response_model=ResearchTemplateResponse, status_code=201)
+async def create_research_template(body: ResearchTemplateRequest, request: Request):
+    """新增研究模板。"""
+    from app.gateway.deps import get_research_template_store
+
+    store = get_research_template_store(request)
+    if store is None:
+        raise HTTPException(status_code=503, detail="Research template store not available")
+    return await store.create(
+        name=body.name,
+        description=body.description,
+        article_type_key=body.article_type_key,
+        outline=body.outline,
+        material_source=body.material_source,
+        word_count=body.word_count,
+        audience=body.audience,
+        strict_mode=body.strict_mode,
+        topic=body.topic,
+        sort_order=body.sort_order,
+    )
+
+
+@router.patch("/research-templates/{template_id}", response_model=ResearchTemplateResponse)
+async def update_research_template(template_id: str, body: UpdateResearchTemplateRequest, request: Request):
+    """更新研究模板字段（只传需要修改的字段）。"""
+    from app.gateway.deps import get_research_template_store
+
+    store = get_research_template_store(request)
+    if store is None:
+        raise HTTPException(status_code=503, detail="Research template store not available")
+
+    kwargs = {}
+    for field in (
+        "name",
+        "description",
+        "article_type_key",
+        "outline",
+        "material_source",
+        "word_count",
+        "audience",
+        "strict_mode",
+        "topic",
+        "sort_order",
+    ):
+        val = getattr(body, field)
+        if val is not None:
+            kwargs[field] = val
+
+    updated = await store.update(template_id, **kwargs)
+    if updated is None:
+        raise HTTPException(status_code=404, detail="Research template not found")
+    return updated
+
+
+@router.delete("/research-templates/{template_id}")
+async def delete_research_template(template_id: str, request: Request):
+    """删除研究模板。"""
+    from app.gateway.deps import get_research_template_store
+
+    store = get_research_template_store(request)
+    if store is None:
+        raise HTTPException(status_code=503, detail="Research template store not available")
+    deleted = await store.delete(template_id)
+    if not deleted:
+        raise HTTPException(status_code=404, detail="Research template not found")
+    return {"success": True}
+
 
 # ── 管理接口 ──────────────────────────────────────────────────────────────────
 
diff --git a/offline-backend-20260512/backend/app/gateway/routers/auth.py b/offline-backend-20260512/backend/app/gateway/routers/auth.py
index ab671066..3eaefc13 100644
--- a/offline-backend-20260512/backend/app/gateway/routers/auth.py
+++ b/offline-backend-20260512/backend/app/gateway/routers/auth.py
@@ -202,6 +202,21 @@ def _username_to_email(username: str) -> str:
     return f"{local}@{domain}"
 
 
+async def _sync_user_position(request: Request, user_id: str) -> None:
+    """Best-effort: materialize the user's 岗位 (position) grants on login.
+
+    Ensures a user who has never been synced — e.g. a brand-new account that
+    inherits the ``is_default`` position — gets the default agents / tasks /
+    prompts loaded into their "我的" tables. Failures never block login.
+    """
+    try:
+        sync = getattr(request.app.state, "position_sync", None)
+        if sync is not None:
+            await sync.sync_user(str(user_id))
+    except Exception:
+        logger.debug("position sync on login failed for user %s", user_id, exc_info=True)
+
+
 async def _get_or_create_user_by_username(username: str):
     """Resolve a username to a local user, auto-registering when absent.
 
@@ -476,6 +491,7 @@ async def login_local(
     _record_login_success(client_ip)
     token = create_access_token(str(user.id), token_version=user.token_version)
     _set_session_cookie(response, token, request)
+    await _sync_user_position(request, user.id)
 
     return LoginResponse(
         expires_in=get_auth_config().token_expiry_days * 24 * 3600,
@@ -520,6 +536,7 @@ async def login_by_username(
 
     token = create_access_token(str(user.id), token_version=user.token_version)
     _set_session_cookie(response, token, request)
+    await _sync_user_position(request, user.id)
 
     return UsernameLoginResponse(
         access_token=token,
@@ -570,6 +587,7 @@ async def login_by_token(
 
     access_token = create_access_token(str(user.id), token_version=user.token_version)
     _set_session_cookie(response, access_token, request)
+    await _sync_user_position(request, user.id)
 
     return UsernameLoginResponse(
         access_token=access_token,
@@ -600,6 +618,7 @@ async def register(request: Request, response: Response, body: RegisterRequest):
 
     token = create_access_token(str(user.id), token_version=user.token_version)
     _set_session_cookie(response, token, request)
+    await _sync_user_position(request, user.id)
 
     return UserResponse(id=str(user.id), email=user.email, system_role=user.system_role)
 
diff --git a/offline-backend-20260512/backend/app/gateway/routers/knowledge.py b/offline-backend-20260512/backend/app/gateway/routers/knowledge.py
index 8562dde9..b3af7de8 100644
--- a/offline-backend-20260512/backend/app/gateway/routers/knowledge.py
+++ b/offline-backend-20260512/backend/app/gateway/routers/knowledge.py
@@ -54,6 +54,8 @@ class IngestThreadRequest(BaseModel):
         description="When true, sediment the thread asynchronously in a background task and return immediately.",
     )
     model_name: str | None = Field(default=None, description="可选:指定用于知识提炼的大模型名称;留空使用默认配置")
+    template_id: str | None = Field(default=None, description="可选:抽取提示词模板 id;留空使用对话默认模板")
+    folder: str | None = Field(default=None, description="可选:沉淀结果归入的目录(斜杠分隔路径)")
 
 
 class BatchIngestRequest(BaseModel):
@@ -63,7 +65,13 @@ class BatchIngestRequest(BaseModel):
     include_sources: bool = True
     skip_existing: bool = True
     mode: str = "summary"
+    background: bool = Field(
+        default=True,
+        description="When true (default), the per-thread extraction loop runs in a background task and the endpoint returns immediately ({queued:true}). Batch capture of many threads is far too slow to finish inside one HTTP request — running it inline makes nginx time out (504), which the frontend surfaces as a generic 'request failed'.",
+    )
     model_name: str | None = Field(default=None, description="可选:指定用于知识提炼的大模型名称;留空使用默认配置")
+    template_id: str | None = Field(default=None, description="可选:抽取提示词模板 id;留空使用对话默认模板")
+    folder: str | None = Field(default=None, description="可选:沉淀结果归入的目录(斜杠分隔路径)")
 
 
 class SearchResultItem(BaseModel):
@@ -90,6 +98,7 @@ class CreateNoteRequest(BaseModel):
     summary: str | None = None
     tags: list[str] = Field(default_factory=list)
     status: str = "approved"
+    folder: str | None = None
 
 
 class UpdateNoteRequest(BaseModel):
@@ -98,6 +107,34 @@ class UpdateNoteRequest(BaseModel):
     summary: str | None = None
     tags: list[str] | None = None
     status: str | None = None
+    # Present (even as null) ⇒ (re)assign the note's directory; absent ⇒ leave as-is.
+    folder: str | None = None
+
+
+class CreateFolderRequest(BaseModel):
+    path: str = Field(description="目录全路径,斜杠分隔,如 投研/行业")
+
+
+class RenameFolderRequest(BaseModel):
+    path: str = Field(description="新的目录全路径")
+
+
+class ExtractTemplateRequest(BaseModel):
+    name: str
+    system_prompt: str
+    scope: str = Field(default="both", description="thread | document | both")
+    description: str | None = None
+    is_default: bool = False
+    enabled: bool = True
+
+
+class UpdateExtractTemplateRequest(BaseModel):
+    name: str | None = None
+    system_prompt: str | None = None
+    scope: str | None = None
+    description: str | None = None
+    is_default: bool | None = None
+    enabled: bool | None = None
 
 
 class SearchRequest(BaseModel):
@@ -187,6 +224,8 @@ async def _ingest_thread_background(service, checkpointer, thread_id: str, body:
             reference_batches=batches,
             created_by=created_by,
             model_name=body.model_name,
+            template_id=body.template_id,
+            folder=body.folder,
         )
     except EmptyThreadError:
         logger.warning("Background knowledge ingest skipped: thread %s has no sediment-able content", sanitize_log_param(thread_id))
@@ -237,6 +276,8 @@ async def ingest_thread(thread_id: str, body: IngestThreadRequest, request: Requ
             reference_batches=batches,
             created_by=get_effective_user_id(),
             model_name=body.model_name,
+            template_id=body.template_id,
+            folder=body.folder,
         )
     except EmptyThreadError:
         raise HTTPException(status_code=400, detail="Thread has no sediment-able content")
@@ -248,15 +289,14 @@ async def ingest_thread(thread_id: str, body: IngestThreadRequest, request: Requ
     }
 
 
-@router.post("/ingest/threads/batch")
-@require_auth
-async def ingest_threads_batch(body: BatchIngestRequest, request: Request) -> dict[str, Any]:
-    """Batch-sediment the caller's recent threads into knowledge notes."""
-    service = get_knowledge_service(request)
-    checkpointer = get_checkpointer(request)
-    thread_store = get_thread_store(request)
+async def _run_batch_ingest(service, checkpointer, threads: list[dict[str, Any]], body: BatchIngestRequest, created_by: str | None) -> dict[str, Any]:
+    """Run the per-thread sediment loop and tally the outcome.
 
-    threads = await thread_store.search(limit=body.limit, offset=body.offset)
+    Shared by the inline and background paths; ``created_by`` is passed in
+    explicitly because the user contextvar is gone once a background task runs.
+    Each thread's transcript is truncated by the extractor (``max_input_chars``)
+    before the LLM call, so a single huge conversation can't blow the context.
+    """
     created = skipped = failed = 0
     # Notes that landed via the rule-based fallback because the chosen model
     # produced nothing usable (timeout / API error / non-JSON). Surfaced so the
@@ -282,9 +322,11 @@ async def ingest_threads_batch(body: BatchIngestRequest, request: Request) -> di
                 status=body.status,
                 include_sources=body.include_sources,
                 reference_batches=batches,
-                created_by=get_effective_user_id(),
+                created_by=created_by,
                 skip_existing=body.skip_existing,
                 model_name=body.model_name,
+                template_id=body.template_id,
+                folder=body.folder,
             )
             if result.get("llm_fallback"):
                 llm_fallback += 1
@@ -308,6 +350,49 @@ async def ingest_threads_batch(body: BatchIngestRequest, request: Request) -> di
     return {"ok": True, "processed": len(items), "created": created, "skipped": skipped, "failed": failed, "llm_fallback": llm_fallback, "items": items}
 
 
+async def _ingest_threads_batch_background(service, checkpointer, threads: list[dict[str, Any]], body: BatchIngestRequest, created_by: str | None) -> None:
+    """Background variant of the batch loop (failures swallowed/logged)."""
+    try:
+        result = await _run_batch_ingest(service, checkpointer, threads, body, created_by)
+        logger.info(
+            "Background batch ingest finished: processed=%d created=%d skipped=%d failed=%d llm_fallback=%d",
+            result["processed"], result["created"], result["skipped"], result["failed"], result["llm_fallback"],
+        )
+    except Exception:
+        logger.exception("Background batch ingest failed")
+
+
+@router.post("/ingest/threads/batch")
+@require_auth
+async def ingest_threads_batch(body: BatchIngestRequest, request: Request, background_tasks: BackgroundTasks) -> dict[str, Any]:
+    """Batch-sediment the caller's recent threads into knowledge notes.
+
+    Defaults to ``background=true``: the (slow) per-thread extraction runs in a
+    background task and the endpoint returns ``{"queued": true, "processed": N}``
+    immediately, so the request never outlives the nginx gateway timeout. Pass
+    ``background=false`` to block until every thread is processed (small batches /
+    scripts that want the full per-thread tally).
+    """
+    service = get_knowledge_service(request)
+    checkpointer = get_checkpointer(request)
+    thread_store = get_thread_store(request)
+
+    threads = await thread_store.search(limit=body.limit, offset=body.offset)
+    if body.background:
+        # Capture the user id now — the contextvar is gone once the task runs.
+        background_tasks.add_task(
+            _ingest_threads_batch_background,
+            service,
+            checkpointer,
+            threads,
+            body,
+            get_effective_user_id(),
+        )
+        return {"ok": True, "queued": True, "processed": len(threads), "created": 0, "skipped": 0, "failed": 0, "llm_fallback": 0, "items": []}
+
+    return await _run_batch_ingest(service, checkpointer, threads, body, get_effective_user_id())
+
+
 async def _extract_upload_text(file: UploadFile) -> tuple[str, str]:
     """Read an uploaded file and return ``(filename, markdown_text)``.
 
@@ -352,6 +437,8 @@ async def ingest_file(
     tags: str | None = Form(default=None),
     status: str = Form(default="approved"),
     model_name: str | None = Form(default=None),
+    template_id: str | None = Form(default=None),
+    folder: str | None = Form(default=None),
 ) -> dict[str, Any]:
     """Import an uploaded file into the knowledge base as wiki note(s).
 
@@ -371,6 +458,8 @@ async def ingest_file(
             status=status,
             created_by=get_effective_user_id(),
             model_name=(model_name or None),
+            template_id=(template_id or None),
+            folder=(folder or None),
         )
     except EmptyThreadError:
         raise HTTPException(status_code=422, detail="文件内容无可沉淀的知识")
@@ -413,6 +502,7 @@ async def create_note(body: CreateNoteRequest, request: Request) -> dict[str, An
         summary=body.summary,
         tags=body.tags,
         status=body.status,
+        folder=body.folder,
         created_by=get_effective_user_id(),
     )
     return {"note": note}
@@ -427,13 +517,14 @@ async def list_notes(
     entity: str | None = None,
     source_type: str | None = None,
     status: str | None = None,
+    folder: str | None = None,
     limit: int = 20,
     offset: int = 0,
 ) -> dict[str, Any]:
     service = get_knowledge_service(request)
-    limit = max(1, min(limit, 100))
+    limit = max(1, min(limit, 1000))
     offset = max(0, offset)
-    return await service.list_notes(keyword=keyword, tag=tag, entity=entity, source_type=source_type, status=status, limit=limit, offset=offset)
+    return await service.list_notes(keyword=keyword, tag=tag, entity=entity, source_type=source_type, status=status, folder=folder, limit=limit, offset=offset)
 
 
 @router.get("/notes/{note_id}")
@@ -451,6 +542,9 @@ async def get_note(note_id: str, request: Request) -> dict[str, Any]:
 @require_auth
 async def update_note(note_id: str, body: UpdateNoteRequest, request: Request) -> dict[str, Any]:
     service = get_knowledge_service(request)
+    # ``folder`` present in the request body (even as null) ⇒ (re)assign the
+    # directory; absent ⇒ leave it untouched.
+    folder_provided = "folder" in body.model_fields_set
     note = await service.update_note(
         note_id,
         title=body.title,
@@ -458,6 +552,8 @@ async def update_note(note_id: str, body: UpdateNoteRequest, request: Request) -
         summary=body.summary,
         tags=body.tags,
         status=body.status,
+        folder=body.folder,
+        set_folder=folder_provided,
         updated_by=get_effective_user_id(),
     )
     if note is None:
@@ -634,3 +730,119 @@ async def export_file(filename: str, request: Request) -> FileResponse:
         raise HTTPException(status_code=404, detail="Export file not found")
     media_type = mimetypes.guess_type(filename)[0] or "application/octet-stream"
     return FileResponse(path, media_type=media_type)
+
+
+# --------------------------------------------------------------------------- #
+# Folders (user-managed directories)
+# --------------------------------------------------------------------------- #
+@router.get("/folders")
+@require_auth
+async def list_folders(request: Request) -> dict[str, Any]:
+    """List every managed knowledge directory (for the directory tree + picker)."""
+    service = get_knowledge_service(request)
+    return {"items": await service.list_folders()}
+
+
+@router.post("/folders")
+@require_auth
+async def create_folder(body: CreateFolderRequest, request: Request) -> dict[str, Any]:
+    """Create a directory (parents auto-created). Notes are filed into it later."""
+    service = get_knowledge_service(request)
+    try:
+        folder = await service.create_folder(body.path, created_by=get_effective_user_id())
+    except ValueError as exc:
+        raise HTTPException(status_code=400, detail=str(exc))
+    return {"folder": folder}
+
+
+@router.put("/folders/{folder_id}")
+@require_auth
+async def rename_folder(folder_id: str, body: RenameFolderRequest, request: Request) -> dict[str, Any]:
+    """Rename/move a directory; descendants + filed notes follow the new prefix."""
+    service = get_knowledge_service(request)
+    try:
+        folder = await service.rename_folder(folder_id, body.path)
+    except ValueError as exc:
+        raise HTTPException(status_code=400, detail=str(exc))
+    if folder is None:
+        raise HTTPException(status_code=404, detail="Folder not found")
+    return {"folder": folder}
+
+
+@router.delete("/folders/{folder_id}")
+@require_auth
+async def delete_folder(folder_id: str, request: Request, reassign_to: str | None = None) -> dict[str, Any]:
+    """Delete a directory (and sub-dirs); notes inside move to ``reassign_to`` or unfiled."""
+    service = get_knowledge_service(request)
+    if not await service.delete_folder(folder_id, reassign_to=reassign_to):
+        raise HTTPException(status_code=404, detail="Folder not found")
+    return {"ok": True}
+
+
+# --------------------------------------------------------------------------- #
+# Extraction prompt templates
+# --------------------------------------------------------------------------- #
+@router.get("/extract-templates")
+@require_auth
+async def list_extract_templates(request: Request, scope: str | None = None, enabled_only: bool = False) -> dict[str, Any]:
+    """List editable extraction-prompt templates (optionally filtered by scope)."""
+    service = get_knowledge_service(request)
+    return {"items": await service.list_extract_templates(scope=scope, enabled_only=enabled_only)}
+
+
+@router.get("/extract-templates/{template_id}")
+@require_auth
+async def get_extract_template(template_id: str, request: Request) -> dict[str, Any]:
+    service = get_knowledge_service(request)
+    tpl = await service.get_extract_template(template_id)
+    if tpl is None:
+        raise HTTPException(status_code=404, detail="Template not found")
+    return {"template": tpl}
+
+
+@router.post("/extract-templates")
+@require_auth
+async def create_extract_template(body: ExtractTemplateRequest, request: Request) -> dict[str, Any]:
+    service = get_knowledge_service(request)
+    if await service.template_name_taken(body.name):
+        raise HTTPException(status_code=409, detail="同名模板已存在")
+    tpl = await service.create_extract_template(
+        name=body.name,
+        system_prompt=body.system_prompt,
+        scope=body.scope,
+        description=body.description,
+        is_default=body.is_default,
+        enabled=body.enabled,
+        created_by=get_effective_user_id(),
+    )
+    return {"template": tpl}
+
+
+@router.put("/extract-templates/{template_id}")
+@require_auth
+async def update_extract_template(template_id: str, body: UpdateExtractTemplateRequest, request: Request) -> dict[str, Any]:
+    service = get_knowledge_service(request)
+    if body.name is not None and await service.template_name_taken(body.name, exclude_id=template_id):
+        raise HTTPException(status_code=409, detail="同名模板已存在")
+    tpl = await service.update_extract_template(
+        template_id,
+        name=body.name,
+        system_prompt=body.system_prompt,
+        scope=body.scope,
+        description=body.description,
+        is_default=body.is_default,
+        enabled=body.enabled,
+        updated_by=get_effective_user_id(),
+    )
+    if tpl is None:
+        raise HTTPException(status_code=404, detail="Template not found")
+    return {"template": tpl}
+
+
+@router.delete("/extract-templates/{template_id}")
+@require_auth
+async def delete_extract_template(template_id: str, request: Request) -> dict[str, Any]:
+    service = get_knowledge_service(request)
+    if not await service.delete_extract_template(template_id):
+        raise HTTPException(status_code=404, detail="Template not found")
+    return {"ok": True}
diff --git a/offline-backend-20260512/backend/app/gateway/routers/positions.py b/offline-backend-20260512/backend/app/gateway/routers/positions.py
new file mode 100644
index 00000000..80572aa8
--- /dev/null
+++ b/offline-backend-20260512/backend/app/gateway/routers/positions.py
@@ -0,0 +1,337 @@
+"""岗位 (position) management API — admin only.
+
+A position aggregates, per resource type, the tags whose resources are granted
+to its members. Assigning a user to a position (1:1) materializes those grants
+into the user's "我的" tables via the sync engine; recommended-question and
+skill visibility are filtered live elsewhere.
+
+All endpoints require ``system_role == "admin"`` (auth-disabled mode lets calls
+through, matching the other admin routers).
+"""
+
+from __future__ import annotations
+
+import logging
+import uuid
+
+from fastapi import APIRouter, Depends, HTTPException, Request
+from pydantic import BaseModel, Field
+
+from app.gateway.deps import get_optional_user_from_request, get_position_store, get_position_sync, get_tag_store
+from deerflow.persistence.positions import POSITION_RESOURCE_TYPES, PositionStore
+from deerflow.persistence.tags import TagStore
+
+logger = logging.getLogger(__name__)
+
+router = APIRouter(prefix="/api/positions", tags=["positions"])
+
+
+async def _require_admin(request: Request) -> None:
+    user = await get_optional_user_from_request(request)
+    if user is None:
+        return
+    if getattr(user, "system_role", None) != "admin":
+        raise HTTPException(status_code=403, detail="岗位管理仅限管理员")
+
+
+def _validate_bindings(bindings: dict[str, list[str]]) -> dict[str, list[str]]:
+    out: dict[str, list[str]] = {}
+    for resource_type, tag_ids in bindings.items():
+        if resource_type not in POSITION_RESOURCE_TYPES:
+            raise HTTPException(status_code=422, detail=f"Invalid resource type '{resource_type}'. Allowed: {', '.join(POSITION_RESOURCE_TYPES)}")
+        out[resource_type] = list(tag_ids or [])
+    return out
+
+
+# ------------------------------------------------------------------ schemas
+
+
+class PositionResponse(BaseModel):
+    id: str
+    name: str
+    description: str = ""
+    is_default: bool = False
+    member_count: int = 0
+    created_at: str | None = None
+    updated_at: str | None = None
+
+
+class PositionListResponse(BaseModel):
+    positions: list[PositionResponse]
+    resource_types: list[str] = Field(default_factory=lambda: list(POSITION_RESOURCE_TYPES))
+
+
+class PositionCreateRequest(BaseModel):
+    name: str = Field(..., min_length=1, max_length=191)
+    description: str = Field(default="", max_length=1024)
+    is_default: bool = False
+
+
+class PositionUpdateRequest(BaseModel):
+    name: str | None = Field(default=None, min_length=1, max_length=191)
+    description: str | None = Field(default=None, max_length=1024)
+    is_default: bool | None = None
+
+
+class TagBindingsResponse(BaseModel):
+    position_id: str
+    bindings: dict[str, list[str]]
+
+
+class TagBindingsRequest(BaseModel):
+    bindings: dict[str, list[str]] = Field(default_factory=dict, description="resource_type -> [tag_id]; omitted/empty type = default set")
+
+
+class MemberResponse(BaseModel):
+    id: str
+    email: str = ""
+
+
+class MembersResponse(BaseModel):
+    members: list[MemberResponse]
+
+
+class MembersMutationRequest(BaseModel):
+    user_ids: list[str] = Field(..., min_length=1)
+
+
+class UserPositionRequest(BaseModel):
+    position_id: str | None = Field(default=None, description="null clears the user's position")
+
+
+class PositionUserResponse(BaseModel):
+    id: str
+    email: str = ""
+    system_role: str = "user"
+    position_id: str | None = None
+
+
+class PositionUsersResponse(BaseModel):
+    users: list[PositionUserResponse]
+
+
+class MyPositionResponse(BaseModel):
+    position: PositionResponse | None = None
+
+
+# ----------------------------------------------------------------- position CRUD
+
+
+@router.get("", response_model=PositionListResponse, summary="List Positions")
+async def list_positions(request: Request, store: PositionStore = Depends(get_position_store)) -> PositionListResponse:
+    await _require_admin(request)
+    records = await store.list_positions()
+    return PositionListResponse(positions=[PositionResponse(**r) for r in records])
+
+
+@router.get("/users", response_model=PositionUsersResponse, summary="List All Users + Their Position")
+async def list_position_users(request: Request, store: PositionStore = Depends(get_position_store)) -> PositionUsersResponse:
+    await _require_admin(request)
+    users = await store.list_all_users()
+    return PositionUsersResponse(users=[PositionUserResponse(**u) for u in users])
+
+
+@router.get("/me", response_model=MyPositionResponse, summary="Get My Effective Position")
+async def get_my_position(request: Request, store: PositionStore = Depends(get_position_store)) -> MyPositionResponse:
+    """The caller's own position — open to any authenticated user (not admin).
+
+    Returns the explicitly-assigned position, falling back to the default
+    position (which unassigned users inherit). ``null`` when neither exists or
+    the caller is anonymous. Powers the top-bar 岗位 label.
+    """
+    user = await get_optional_user_from_request(request)
+    if user is None:
+        return MyPositionResponse(position=None)
+    record = None
+    pid = await store.get_user_position_id(str(user.id))
+    if pid:
+        record = await store.get_position(pid)
+    if record is None:
+        default_id = await store.get_default_position_id()
+        if default_id:
+            record = await store.get_position(default_id)
+    return MyPositionResponse(position=PositionResponse(**record) if record else None)
+
+
+@router.post("", response_model=PositionResponse, status_code=201, summary="Create Position")
+async def create_position(request: Request, body: PositionCreateRequest, store: PositionStore = Depends(get_position_store)) -> PositionResponse:
+    await _require_admin(request)
+    try:
+        record = await store.create_position({"id": uuid.uuid4().hex, "name": body.name.strip(), "description": body.description, "is_default": body.is_default})
+        return PositionResponse(**record)
+    except ValueError as e:
+        raise HTTPException(status_code=409, detail=str(e))
+
+
+@router.put("/{position_id}", response_model=PositionResponse, summary="Update Position")
+async def update_position(
+    request: Request,
+    position_id: str,
+    body: PositionUpdateRequest,
+    store: PositionStore = Depends(get_position_store),
+    sync=Depends(get_position_sync),
+) -> PositionResponse:
+    await _require_admin(request)
+    update_data = body.model_dump(exclude_unset=True)
+    if update_data.get("name"):
+        update_data["name"] = update_data["name"].strip()
+    try:
+        record = await store.update_position(position_id, update_data)
+    except ValueError as e:
+        raise HTTPException(status_code=409, detail=str(e))
+    if record is None:
+        raise HTTPException(status_code=404, detail="Position not found")
+    # Becoming the default changes what unassigned users inherit.
+    if update_data.get("is_default"):
+        await sync.sync_default_members()
+    return PositionResponse(**record)
+
+
+@router.delete("/{position_id}", status_code=204, summary="Delete Position")
+async def delete_position(
+    request: Request,
+    position_id: str,
+    store: PositionStore = Depends(get_position_store),
+    sync=Depends(get_position_sync),
+) -> None:
+    await _require_admin(request)
+    member_ids = await store.list_member_ids(position_id)
+    deleted = await store.delete_position(position_id)
+    if not deleted:
+        raise HTTPException(status_code=404, detail="Position not found")
+    # Former members now fall back to the default position; re-sync them.
+    for uid in member_ids:
+        await sync.sync_user(uid)
+
+
+# ----------------------------------------------------------------- tag bindings
+
+
+@router.get("/{position_id}/tag-bindings", response_model=TagBindingsResponse, summary="Get Tag Bindings")
+async def get_tag_bindings(request: Request, position_id: str, store: PositionStore = Depends(get_position_store)) -> TagBindingsResponse:
+    await _require_admin(request)
+    if await store.get_position(position_id) is None:
+        raise HTTPException(status_code=404, detail="Position not found")
+    bindings = await store.get_tag_bindings(position_id)
+    return TagBindingsResponse(position_id=position_id, bindings=bindings)
+
+
+@router.put("/{position_id}/tag-bindings", response_model=TagBindingsResponse, summary="Set Tag Bindings")
+async def set_tag_bindings(
+    request: Request,
+    position_id: str,
+    body: TagBindingsRequest,
+    store: PositionStore = Depends(get_position_store),
+    tag_store: TagStore = Depends(get_tag_store),
+    sync=Depends(get_position_sync),
+) -> TagBindingsResponse:
+    await _require_admin(request)
+    if await store.get_position(position_id) is None:
+        raise HTTPException(status_code=404, detail="Position not found")
+    cleaned = _validate_bindings(body.bindings)
+    # Validate every tag id exists and matches its resource type.
+    for resource_type, tag_ids in cleaned.items():
+        if not tag_ids:
+            continue
+        known = {t["id"] for t in await tag_store.list_tags(resource_type)}
+        for tid in tag_ids:
+            if tid not in known:
+                raise HTTPException(status_code=422, detail=f"Tag '{tid}' is not a '{resource_type}' tag")
+    bindings = await store.set_tag_bindings(position_id, cleaned)
+    await sync.sync_position(position_id)
+    return TagBindingsResponse(position_id=position_id, bindings=bindings)
+
+
+# --------------------------------------------------------------------- members
+
+
+@router.get("/{position_id}/members", response_model=MembersResponse, summary="List Position Members")
+async def list_members(request: Request, position_id: str, store: PositionStore = Depends(get_position_store)) -> MembersResponse:
+    await _require_admin(request)
+    if await store.get_position(position_id) is None:
+        raise HTTPException(status_code=404, detail="Position not found")
+    members = await store.list_members(position_id)
+    return MembersResponse(members=[MemberResponse(**m) for m in members])
+
+
+@router.post("/{position_id}/members", response_model=MembersResponse, summary="Add Position Members")
+async def add_members(
+    request: Request,
+    position_id: str,
+    body: MembersMutationRequest,
+    store: PositionStore = Depends(get_position_store),
+    sync=Depends(get_position_sync),
+) -> MembersResponse:
+    await _require_admin(request)
+    if await store.get_position(position_id) is None:
+        raise HTTPException(status_code=404, detail="Position not found")
+    # Re-sync users leaving their previous position too, so its grants are dropped.
+    prev_positions = {uid: await store.get_user_position_id(uid) for uid in body.user_ids}
+    await store.assign_users(position_id, body.user_ids)
+    for uid in body.user_ids:
+        await sync.sync_user(uid)
+        prev = prev_positions.get(uid)
+        if prev and prev != position_id:
+            logger.debug("user %s moved from position %s to %s", uid, prev, position_id)
+    members = await store.list_members(position_id)
+    return MembersResponse(members=[MemberResponse(**m) for m in members])
+
+
+@router.delete("/{position_id}/members", response_model=MembersResponse, summary="Remove Position Members")
+async def remove_members(
+    request: Request,
+    position_id: str,
+    body: MembersMutationRequest,
+    store: PositionStore = Depends(get_position_store),
+    sync=Depends(get_position_sync),
+) -> MembersResponse:
+    await _require_admin(request)
+    if await store.get_position(position_id) is None:
+        raise HTTPException(status_code=404, detail="Position not found")
+    # Only unassign users that are actually in this position.
+    member_ids = set(await store.list_member_ids(position_id))
+    victims = [uid for uid in body.user_ids if uid in member_ids]
+    await store.unassign_users(victims)
+    for uid in victims:
+        await sync.sync_user(uid)
+    members = await store.list_members(position_id)
+    return MembersResponse(members=[MemberResponse(**m) for m in members])
+
+
+@router.post("/{position_id}/resync", summary="Re-sync Position Members")
+async def resync_position(
+    request: Request,
+    position_id: str,
+    store: PositionStore = Depends(get_position_store),
+    sync=Depends(get_position_sync),
+) -> dict:
+    await _require_admin(request)
+    if await store.get_position(position_id) is None:
+        raise HTTPException(status_code=404, detail="Position not found")
+    count = await sync.sync_position(position_id)
+    return {"synced": count}
+
+
+# --------------------------------------------------- per-user position (1:1)
+
+
+@router.put("/users/{user_id}", response_model=PositionResponse | None, summary="Set User Position")
+async def set_user_position(
+    request: Request,
+    user_id: str,
+    body: UserPositionRequest,
+    store: PositionStore = Depends(get_position_store),
+    sync=Depends(get_position_sync),
+) -> PositionResponse | None:
+    await _require_admin(request)
+    if body.position_id:
+        if await store.get_position(body.position_id) is None:
+            raise HTTPException(status_code=404, detail="Position not found")
+        await store.assign_users(body.position_id, [user_id])
+    else:
+        await store.unassign_users([user_id])
+    await sync.sync_user(user_id)
+    if not body.position_id:
+        return None
+    record = await store.get_position(body.position_id)
+    return PositionResponse(**record) if record else None
diff --git a/offline-backend-20260512/backend/app/gateway/routers/recommended_questions.py b/offline-backend-20260512/backend/app/gateway/routers/recommended_questions.py
index 06ea6c60..a267c901 100644
--- a/offline-backend-20260512/backend/app/gateway/routers/recommended_questions.py
+++ b/offline-backend-20260512/backend/app/gateway/routers/recommended_questions.py
@@ -6,18 +6,28 @@ import logging
 from datetime import datetime
 from uuid import uuid4
 
-from fastapi import APIRouter, HTTPException, Request
+from fastapi import APIRouter, HTTPException, Query, Request
 from pydantic import BaseModel, Field
 
+from deerflow.runtime.user_context import get_effective_user_id
+
 logger = logging.getLogger(__name__)
 router = APIRouter(prefix="/api", tags=["recommended-questions"])
 
 
+def _current_user_id(request: Request) -> str:
+    user = getattr(request.state, "user", None)
+    if user is not None:
+        return str(user.id)
+    return get_effective_user_id()
+
+
 class RecommendedQuestionResponse(BaseModel):
     id: str
     title: str
     content: str
     sort_order: int = 0
+    tags: list[dict] = Field(default_factory=list, description="Tags assigned to this question (按岗位分发标签)")
     created_at: datetime | str | None = None
     updated_at: datetime | str | None = None
 
@@ -50,10 +60,46 @@ def _get_store(request: Request):
 
 
 @router.get("/recommended-questions", response_model=RecommendedQuestionsListResponse)
-async def list_recommended_questions(request: Request) -> RecommendedQuestionsListResponse:
+async def list_recommended_questions(
+    request: Request,
+    all: bool = Query(default=False, description="Admin management view: return every question, unfiltered by 岗位."),
+    tag_ids: list[str] | None = Query(default=None, description="Keep questions carrying any of these tags (OR)."),
+) -> RecommendedQuestionsListResponse:
     store = _get_store(request)
     rows = await store.list_all()
-    return RecommendedQuestionsListResponse(questions=[RecommendedQuestionResponse(**r) for r in rows])
+    if not all:
+        # Filter by the caller's 岗位 (position) visibility. ``None`` =
+        # unrestricted (no skill/question tag binding → show all default
+        # questions); a set = show only the questions carrying the bound tags.
+        sync = getattr(request.app.state, "position_sync", None)
+        if sync is not None:
+            try:
+                visible = await sync.resolve_visible(_current_user_id(request), "recommended_question")
+            except Exception:
+                logger.debug("recommended-question visibility resolve failed", exc_info=True)
+                visible = None
+            if visible is not None:
+                rows = [r for r in rows if r["id"] in visible]
+    responses = [RecommendedQuestionResponse(**r) for r in rows]
+
+    # Bulk-attach the tags assigned to each question so the management list can
+    # display them and filter by 标签. Best-effort: a missing/failing tag store
+    # never breaks the question list.
+    tag_store = getattr(request.app.state, "tag_store", None)
+    if tag_store is not None and responses:
+        try:
+            assignments = await tag_store.list_assignments("recommended_question", [r.id for r in responses])
+        except Exception:
+            logger.debug("recommended-question tag attach failed", exc_info=True)
+            assignments = {}
+        for resp in responses:
+            resp.tags = assignments.get(resp.id, [])
+
+    if tag_ids:
+        wanted = set(tag_ids)
+        responses = [r for r in responses if any(t.get("id") in wanted for t in r.tags)]
+
+    return RecommendedQuestionsListResponse(questions=responses)
 
 
 @router.post("/recommended-questions", response_model=RecommendedQuestionResponse, status_code=201)
diff --git a/offline-backend-20260512/backend/app/gateway/routers/skills.py b/offline-backend-20260512/backend/app/gateway/routers/skills.py
index ad284ce8..dad9758f 100644
--- a/offline-backend-20260512/backend/app/gateway/routers/skills.py
+++ b/offline-backend-20260512/backend/app/gateway/routers/skills.py
@@ -154,6 +154,7 @@ class SkillResponse(BaseModel):
     license: str | None = Field(None, description="License information")
     category: SkillCategory = Field(..., description="Category of the skill (public or custom)")
     enabled: bool = Field(default=True, description="Whether this skill is enabled")
+    always_on: bool = Field(default=False, description="常驻(免压缩):开启技能压缩后仍全文注入系统提示词")
     use_count: int = Field(default=0, description="Number of times the skill has been used")
     view_count: int = Field(default=0, description="Number of times the skill has been viewed")
     patch_count: int = Field(default=0, description="Number of times the skill has been patched")
@@ -168,6 +169,7 @@ class SkillResponse(BaseModel):
     published: bool = Field(default=False, description="Whether the owner has made this skill visible to other users")
     square_id: str = Field(default="", description="Publish square (发布广场) the skill is filed under once published. Empty string = the default square.")
     builtin: bool = Field(default=False, description="True when owner_user_id is null (no human owner)")
+    favorited: bool = Field(default=False, description="True when this skill is in the caller's 我的技能 collection (e.g. granted by their 岗位/position). Owned skills are detected separately via owner_user_id.")
     detail: str | None = Field(default=None, description="Human-authored long-form detail text (Markdown), editable by owner/admin")
     display: dict[str, Any] | None = Field(default=None, description="Optional frontend display adapter")
     display_sample: dict[str, Any] | None = Field(default=None, alias="displaySample", description="Persisted sample output used by the display adapter editor")
@@ -186,6 +188,12 @@ class SkillUpdateRequest(BaseModel):
     enabled: bool = Field(..., description="Whether to enable or disable the skill")
 
 
+class SkillAlwaysOnUpdateRequest(BaseModel):
+    """Request model for updating a skill's 常驻(免压缩) flag."""
+
+    always_on: bool = Field(..., description="Whether this skill bypasses skill compression (full-text injected)")
+
+
 class SkillDisplayUpdateRequest(BaseModel):
     """Request model for updating a skill's frontend display adapter."""
 
@@ -273,6 +281,7 @@ def _skill_to_response(
     *,
     display_override: dict[str, Any] | None | object = _UNSET,
     display_sample: dict[str, Any] | None = None,
+    favorited: bool = False,
 ) -> SkillResponse:
     """Convert a Skill object to a SkillResponse.
 
@@ -310,6 +319,9 @@ def _skill_to_response(
     detail: str | None = None
     name_zh: str | None = None
     square_id = ""
+    # 常驻(免压缩) is stored in the DB (SkillRow.always_on), surfaced via the
+    # ownership record — never read from the on-disk extensions_config.json.
+    always_on = False
     if ownership is not None:
         owner_user_id = ownership.get("owner_user_id")
         published = bool(ownership.get("published", False))
@@ -317,6 +329,7 @@ def _skill_to_response(
         detail = ownership.get("detail")
         name_zh = ownership.get("name_zh")
         square_id = str(ownership.get("square_id") or "")
+        always_on = bool(ownership.get("always_on", False))
 
     display = None if display_override is _UNSET else display_override
     if display_override is _UNSET:
@@ -334,6 +347,7 @@ def _skill_to_response(
         license=skill.license,
         category=skill.category,
         enabled=skill.enabled,
+        always_on=always_on,
         use_count=entry.get("use_count", 0),
         view_count=entry.get("view_count", 0),
         patch_count=entry.get("patch_count", 0),
@@ -347,6 +361,7 @@ def _skill_to_response(
         published=published,
         square_id=square_id,
         builtin=builtin,
+        favorited=favorited,
         detail=detail,
         display=display,
         displaySample=display_sample,
@@ -777,6 +792,7 @@ async def list_skills(
     search: str | None = Query(default=None, description="Case-insensitive fuzzy match on skill name or Chinese name (name_zh)"),
     tag_ids: list[str] | None = Query(default=None, description="Keep skills carrying any of these tags (OR)"),
     square: str | None = Query(default=None, description="Keep only skills filed under this publish square (发布广场). Omit for all squares."),
+    scope: str | None = Query(default=None, description="``mine`` filters to the skills the caller's 岗位 (position) makes visible."),
     sort: str = Query(
         default="time_desc",
         pattern="^(time_desc|time_asc|activity_desc|activity_asc|usage_desc|usage_asc|name_asc|name_desc)$",
@@ -801,6 +817,15 @@ async def list_skills(
             ownership = {record["name"]: record for record in ownership_records}
         else:
             ownership = await _build_ownership_index(skill_store, user_id)
+        # Skills the caller's 岗位 (position) granted into their "我的技能"
+        # collection. Admins browse everything, so the badge is moot for them.
+        favorite_names: set[str] = set()
+        if not admin:
+            try:
+                favorite_names = await skill_store.list_favorite_skill_names(user_id)
+            except Exception:
+                logger.debug("skill favorites lookup failed", exc_info=True)
+                favorite_names = set()
         display_adapters = await _list_skill_display_adapters([skill.name for skill in all_skills])
         responses: list[SkillResponse] = []
         for skill in all_skills:
@@ -816,22 +841,30 @@ async def list_skills(
                         ownership.get(skill.name),
                         display_override=_display_override_from_adapter(adapter),
                         display_sample=adapter.get("sample") if adapter else None,
+                        favorited=skill.name in favorite_names,
                     )
                 )
                 continue
             record = ownership.get(skill.name)
             if record is None:
-                if not admin:
+                if admin:
+                    # Admin sees every custom skill on disk even without an
+                    # ownership row (legacy / orphaned), so they can clean it up.
+                    record = None
+                elif skill.name in favorite_names:
+                    # A position granted this skill even though it isn't
+                    # otherwise visible to the user (e.g. an unpublished custom
+                    # skill). Surface it so it lands in their 我的技能.
+                    record = await skill_store.get_any(skill.name)
+                else:
                     continue
-                # Admin sees every custom skill on disk even without an
-                # ownership row (legacy / orphaned), so they can clean it up.
-                record = None
             responses.append(
                 _skill_to_response(
                     skill,
                     record,
                     display_override=_display_override_from_adapter(adapter),
                     display_sample=adapter.get("sample") if adapter else None,
+                    favorited=skill.name in favorite_names,
                 )
             )
         responses = await _attach_and_filter_tags(tag_store, responses, search, tag_ids)
@@ -846,6 +879,12 @@ async def list_skills(
                 for r in responses
                 if (r.square_id or "") == square_id or (square_is_default and not (r.square_id or ""))
             ]
+        # 岗位 (position) no longer *restricts* the skill list: a position grants
+        # skills by materializing them into the member's 我的技能 (the ``favorited``
+        # flag above + the ``skill_favorites`` table, synced by PositionSyncService),
+        # mirroring how agents are added to 我的智能体. The browse-all 广场 therefore
+        # always shows every visible skill — the ``scope`` param is kept for API
+        # compatibility but intentionally does not filter here.
         responses = _sort_skill_responses(responses, sort)
         return SkillsListResponse(skills=responses)
     except HTTPException:
@@ -2200,6 +2239,64 @@ async def update_skill(skill_name: str, request: SkillUpdateRequest, config: App
         raise HTTPException(status_code=500, detail=f"Failed to update skill: {str(e)}")
 
 
+@router.put(
+    "/skills/{skill_name}/always-on",
+    response_model=SkillResponse,
+    summary="Update Skill 常驻(免压缩) Flag",
+    description="Mark a skill as 常驻(免压缩) so it is injected full-text into the system prompt even when skill compression is enabled.",
+)
+async def update_skill_always_on(
+    skill_name: str,
+    request: SkillAlwaysOnUpdateRequest,
+    http_request: Request,
+    config: AppConfig = Depends(get_config),
+    skill_store: SkillStore = Depends(get_skill_store),
+) -> SkillResponse:
+    """Persist the 常驻(免压缩) flag in the **DB** (skills table), not in
+    extensions_config.json — the source-tree config file is read-only / unsafe to
+    rewrite in some deployments, and built-in skills carry no file entry anyway.
+    Mirrors the detail/name_zh endpoints: custom skill → owner or admin; built-in
+    / public skill → admin only (materializes a row on demand)."""
+    try:
+        skill_name = skill_name.replace("\r\n", "").replace("\n", "")
+        user_id = _current_user_id(http_request)
+        is_admin = _is_admin_user(http_request)
+
+        skills = get_or_new_skill_storage(app_config=config).load_skills(enabled_only=False)
+        skill = next((s for s in skills if s.name == skill_name), None)
+        if skill is None:
+            raise HTTPException(status_code=404, detail=f"Skill '{skill_name}' not found")
+
+        if skill.category == SkillCategory.CUSTOM:
+            _record, mode = await _get_editable_skill_or_raise(skill_store, skill_name, user_id, is_admin=is_admin)
+            if mode == "admin":
+                updated = await skill_store.update_admin(skill_name, {"always_on": request.always_on})
+            else:
+                updated = await skill_store.update(skill_name, user_id, {"always_on": request.always_on})
+        else:
+            # Built-in / public skill — admin only; no DB row by default.
+            if not is_admin:
+                raise HTTPException(status_code=403, detail="Only admin users can change a built-in skill's 常驻 flag")
+            if await skill_store.get_any(skill_name) is None:
+                await skill_store.ensure_legacy({"name": skill_name, "owner_user_id": None, "published": True})
+            updated = await skill_store.update_admin(skill_name, {"always_on": request.always_on})
+
+        if updated is None:
+            raise HTTPException(status_code=500, detail="Failed to persist skill 常驻 flag")
+
+        # always_on lives in the system prompt's skill injection — refresh the cache.
+        await refresh_skills_system_prompt_cache_async()
+
+        logger.info("Skill '%s' always_on updated to %s", skill_name, request.always_on)
+        return _skill_to_response(skill, updated)
+
+    except HTTPException:
+        raise
+    except Exception as e:
+        logger.error(f"Failed to update skill always_on {skill_name}: {e}", exc_info=True)
+        raise HTTPException(status_code=500, detail=f"Failed to update skill: {str(e)}")
+
+
 @router.put(
     "/skills/{skill_name}/display",
     response_model=SkillResponse,
diff --git a/offline-backend-20260512/backend/app/gateway/routers/system_settings.py b/offline-backend-20260512/backend/app/gateway/routers/system_settings.py
index 42e095ef..fc014c8d 100644
--- a/offline-backend-20260512/backend/app/gateway/routers/system_settings.py
+++ b/offline-backend-20260512/backend/app/gateway/routers/system_settings.py
@@ -16,11 +16,14 @@ from deerflow.config.system_settings import (
     PromptPrefixSettings,
     PublishSquare,
     PublishSquaresSettings,
+    SkillCompressionSettings,
     SystemSettings,
+    get_module_tag_ids,
     list_publish_squares,
     load_system_settings,
     resolve_default_square_id,
     save_system_settings,
+    set_module_tag_ids,
 )
 
 logger = logging.getLogger(__name__)
@@ -113,6 +116,64 @@ async def update_prompt_prefix(body: PromptPrefixUpdateRequest, request: Request
     return _to_response(saved.prompt_prefix)
 
 
+class SkillCompressionResponse(BaseModel):
+    enabled: bool = Field(..., description="Whether skill compression is active")
+    search_top_n: int = Field(..., description="Max skills returned by the search_skills tool")
+    keep_index_in_prompt: bool = Field(..., description="Whether compressed skills keep a compact name index in the prompt")
+
+
+class SkillCompressionUpdateRequest(BaseModel):
+    enabled: bool | None = None
+    search_top_n: int | None = Field(default=None, ge=1, le=50)
+    keep_index_in_prompt: bool | None = None
+
+
+def _skill_compression_to_response(settings: SkillCompressionSettings) -> SkillCompressionResponse:
+    return SkillCompressionResponse(
+        enabled=settings.enabled,
+        search_top_n=settings.search_top_n,
+        keep_index_in_prompt=settings.keep_index_in_prompt,
+    )
+
+
+@router.get("/skill-compression", response_model=SkillCompressionResponse)
+async def get_skill_compression() -> SkillCompressionResponse:
+    """Return the skill-compression configuration (readable by any user).
+
+    The frontend reads this to decide whether to surface the per-skill 常驻
+    (免压缩) toggle on the skill detail page.
+    """
+    return _skill_compression_to_response(load_system_settings().skill_compression)
+
+
+@router.put("/skill-compression", response_model=SkillCompressionResponse)
+async def update_skill_compression(body: SkillCompressionUpdateRequest, request: Request) -> SkillCompressionResponse:
+    """Update the skill-compression configuration. Admin only."""
+    await _require_admin(request)
+
+    current = load_system_settings()
+    fields = body.model_dump(exclude_unset=True)
+    update: dict[str, object] = {}
+    if "enabled" in fields and fields["enabled"] is not None:
+        update["enabled"] = bool(fields["enabled"])
+    if "keep_index_in_prompt" in fields and fields["keep_index_in_prompt"] is not None:
+        update["keep_index_in_prompt"] = bool(fields["keep_index_in_prompt"])
+    if "search_top_n" in fields and fields["search_top_n"] is not None:
+        update["search_top_n"] = max(1, min(int(fields["search_top_n"]), 50))
+    compression = current.skill_compression.model_copy(update=update)
+    saved = save_system_settings(current.model_copy(update={"skill_compression": compression}))
+
+    # 压缩开关/参数变化会改变系统提示词的技能注入,清掉提示词缓存让其立即生效。
+    try:
+        from deerflow.agents.lead_agent.prompt import refresh_skills_system_prompt_cache_async
+
+        await refresh_skills_system_prompt_cache_async()
+    except Exception:
+        logger.debug("Failed to refresh skills prompt cache after compression update", exc_info=True)
+
+    return _skill_compression_to_response(saved.skill_compression)
+
+
 def _leaderboard_to_response(settings: LeaderboardSettings) -> LeaderboardSettingsResponse:
     return LeaderboardSettingsResponse(
         exclude_user_ids=list(dict.fromkeys([uid for uid in settings.exclude_user_ids if uid])),
@@ -377,3 +438,35 @@ async def get_log_system(request: Request) -> LogSystemResponse:
     await _require_admin(request)
     cfg = get_app_config().log_system
     return LogSystemResponse(enabled=cfg.enabled, url=cfg.url, label=cfg.label)
+
+
+class AgentModuleConfigResponse(BaseModel):
+    """Selected agent-tag ids for a「按类型聚合」module (zzJC / zzXD)."""
+
+    key: str = Field(..., description="Module key, e.g. zzjc / zzxd")
+    tag_ids: list[str] = Field(default_factory=list, description="Selected agent tag ids")
+
+
+class AgentModuleConfigUpdateRequest(BaseModel):
+    tag_ids: list[str] = Field(default_factory=list, description="Full replacement list of selected agent tag ids")
+
+
+@router.get("/agent-modules/{key}", response_model=AgentModuleConfigResponse)
+async def get_agent_module_config(key: str) -> AgentModuleConfigResponse:
+    """Return the agent tag ids configured for ``key`` (readable by any user —
+    the module's left sidebar needs it to filter the agent list)."""
+    settings = load_system_settings().agent_modules
+    return AgentModuleConfigResponse(key=key, tag_ids=get_module_tag_ids(settings, key))
+
+
+@router.put("/agent-modules/{key}", response_model=AgentModuleConfigResponse)
+async def update_agent_module_config(
+    key: str, body: AgentModuleConfigUpdateRequest, request: Request
+) -> AgentModuleConfigResponse:
+    """Replace the agent tag ids configured for ``key``. Admin only."""
+    await _require_admin(request)
+
+    current = load_system_settings()
+    updated = set_module_tag_ids(current.agent_modules, key, body.tag_ids)
+    saved = save_system_settings(current.model_copy(update={"agent_modules": updated}))
+    return AgentModuleConfigResponse(key=key, tag_ids=get_module_tag_ids(saved.agent_modules, key))
diff --git a/offline-backend-20260512/backend/app/gateway/services.py b/offline-backend-20260512/backend/app/gateway/services.py
index 12c8781b..362c5b9c 100644
--- a/offline-backend-20260512/backend/app/gateway/services.py
+++ b/offline-backend-20260512/backend/app/gateway/services.py
@@ -243,6 +243,9 @@ _CONTEXT_CONFIGURABLE_KEYS: frozenset[str] = frozenset(
         # 关掉个人记忆召回 / builtin 记忆注入(见 multi_agent._run_payload)。
         "memory_recall_disabled",
         "memory_injection_enabled",
+        # 上下文压缩开关:由聊天框下方的按钮控制(默认关闭),透传到 run config 后
+        # 在 _build_middlewares 里决定是否挂载 SummarizationMiddleware。
+        "summarization_enabled",
     }
 )
 
diff --git a/offline-backend-20260512/backend/config.yaml b/offline-backend-20260512/backend/config.yaml
index 5c28a316..0ca799fe 100644
--- a/offline-backend-20260512/backend/config.yaml
+++ b/offline-backend-20260512/backend/config.yaml
@@ -190,6 +190,31 @@ title:
     User: {user_msg}
     Assistant: {assistant_msg}
 
+# 上下文压缩(对话摘要):长对话接近上下文上限时,把较早的消息总结成一段摘要,
+# 只保留最近若干条原始消息,避免撑爆模型上下文窗口。
+# - trigger:触发条件,可配多个,任意一个先满足就压缩(OR 关系)。真正决定何时
+#     压缩的是 trigger(不是 keep)。type=tokens 按 token 总数;type=messages 按条数。
+# - keep:压缩后保留多少最近的上下文(tokens / messages)。
+# - trim_tokens_to_summarize:喂给摘要模型的消息最多裁到多少 token。
+# 当前为「测试值(易触发)」:发到服务器后稍微聊几句就会触发,可在 Gateway 日志里
+# grep「上下文压缩」/「context compaction」确认生效;确认后再调回生产值
+# (trigger tokens 60000 / messages 80,keep tokens 16000)。
+summarization:
+  # 开关已移到聊天框下方(参考文献下面)的「上下文压缩」按钮,默认关闭。
+  # 这里 enabled 仅作为非 Web 路径(渠道/定时任务)的后备开关;下面的 trigger/keep
+  # 等仅是调参,Web 端是否压缩由前端按钮(summarization_enabled)决定。
+  enabled: false
+  # model_name: deepseek-chat   # 留空则用主对话模型
+  trigger:
+    - type: tokens
+      value: 8000
+    - type: messages
+      value: 12
+  keep:
+    type: messages
+    value: 6
+  trim_tokens_to_summarize: 4000
+
 tool_groups:
   - name: web
   - name: file:read
diff --git a/offline-backend-20260512/backend/docs/summarization.md b/offline-backend-20260512/backend/docs/summarization.md
index a3037a0f..ba1a6939 100644
--- a/offline-backend-20260512/backend/docs/summarization.md
+++ b/offline-backend-20260512/backend/docs/summarization.md
@@ -22,6 +22,9 @@ LangChain's stock `SummarizationMiddleware` runs in `before_model` and persists
 - Compression runs in `wrap_model_call` / `awrap_model_call`: for a single model call, the message list is replaced with `[summary, *recent]` via `request.override(messages=...)`. **Nothing is written back to state.**
 - The full original messages stay in the checkpoint, so the frontend keeps showing the user's real conversation. The summary message is named `summary` and flagged `hide_from_ui`, and because it only ever lives in the transient request it never reaches the frontend at all.
 - The generated summary is cached **per-thread** so we don't pay an extra LLM call on every model invocation. The cached summary is reused (no LLM call) until enough *new* messages accumulate to push the compressed context back over the trigger threshold (a sticky boundary mirroring the stock middleware's post-compaction behavior). The cache is bounded (oldest threads evicted first).
+- **Threshold-first**: while the full current conversation is still under the trigger, the request is passed through untouched — no cache lookup, no summary substitution. Compression only ever engages once the real context crosses the configured trigger.
+- **Stickiness headroom**: after `keep` picks a cutoff, the kept window is forced below ~75% of *every* trigger before summarizing (`_apply_stickiness_headroom`). This prevents the every-turn re-compaction that happens when `keep` is in *messages* (e.g. 20) but a trigger is in *tokens* (e.g. 60000) and the recent messages are large — without headroom the "compressed" context would still sit over the trigger and re-summarize on every follow-up question.
+- When the genuine (blocking) summarize path actually runs an LLM call, the middleware emits a `{"type":"context_compacting","message":"正在压缩上下文…"}` event on LangGraph's `custom` stream channel so the chat UI can show a transient "正在压缩" toast (the cheap cache-reuse path stays silent). Frontend handler: `onCustomEvent` in `frontend-web/src/core/threads/hooks.ts`.
 
 > **Legacy threads:** conversations compacted by an older (destructive) build already lost their original messages and carry a persisted `summary` message. The frontend hides any `name: "summary"` message, so the stray summary bubble no longer shows — but those already-discarded messages cannot be recovered.
 
@@ -326,7 +329,7 @@ Summarization runs after ThreadData and Sandbox initialization but before Title
 
 - Configuration is loaded once at startup
 - Summarization is **non-destructive**: the summary replaces older messages only in the transient model request (`wrap_model_call`), never in persisted state. The checkpointer therefore keeps the full original conversation
-- The per-thread summary cache lives in the middleware instance (in-memory, bounded). It is a cost optimization only — a cold/stale cache simply re-summarizes, which is always correct
+- The per-thread summary cache lives in a **process-wide, thread-safe LRU** (`_GLOBAL_SUMMARY_CACHE`, bounded), **not** on the middleware instance. `make_lead_agent` rebuilds a fresh middleware every run, so an instance cache would be cold each turn and re-run a blocking summary LLM call on every turn of a long conversation; the module-level cache keyed by `thread_id` lets a summary survive across turns. It is a cost optimization only — a cold/stale cache simply re-summarizes, which is always correct
 
 ## Example Configurations
 
diff --git a/offline-backend-20260512/backend/packages/harness/deerflow/agents/lead_agent/agent.py b/offline-backend-20260512/backend/packages/harness/deerflow/agents/lead_agent/agent.py
index 88824e72..00a2d1cb 100644
--- a/offline-backend-20260512/backend/packages/harness/deerflow/agents/lead_agent/agent.py
+++ b/offline-backend-20260512/backend/packages/harness/deerflow/agents/lead_agent/agent.py
@@ -54,12 +54,19 @@ def _resolve_model_name(requested_model_name: str | None = None, *, app_config:
     return default_model_name
 
 
-def _create_summarization_middleware(*, app_config: AppConfig | None = None) -> DeerFlowSummarizationMiddleware | None:
-    """Create and configure the summarization middleware from config."""
+def _create_summarization_middleware(*, app_config: AppConfig | None = None, force_enabled: bool = False) -> DeerFlowSummarizationMiddleware | None:
+    """Create and configure the summarization middleware from config.
+
+    Args:
+        force_enabled: When True, build the middleware regardless of the config
+            ``summarization.enabled`` flag. The per-request 上下文压缩 toggle
+            (chat input box, default off) drives this — the config file no longer
+            owns the on/off switch, only the tuning (trigger/keep/...).
+    """
     resolved_app_config = app_config or get_app_config()
     config = resolved_app_config.summarization
 
-    if not config.enabled:
+    if not (config.enabled or force_enabled):
         return None
 
     # Prepare trigger parameter
@@ -268,8 +275,13 @@ def _build_middlewares(
     if cfg.get("rag_mode_enabled", False):
         middlewares.append(RagCitationMiddleware())
 
-    # Add summarization middleware if enabled
-    summarization_middleware = _create_summarization_middleware(app_config=resolved_app_config)
+    # Add summarization middleware. The on/off switch now lives in the chat UI
+    # (上下文压缩 toggle, default off) and arrives as ``summarization_enabled`` in
+    # the runtime config; the config file only supplies tuning. force_enabled
+    # lets the per-request toggle turn it on even when config.enabled is false,
+    # while non-UI paths (channels/scheduled) still fall back to config.enabled.
+    summarization_requested = cfg.get("summarization_enabled", False)
+    summarization_middleware = _create_summarization_middleware(app_config=resolved_app_config, force_enabled=summarization_requested)
     if summarization_middleware is not None:
         middlewares.append(summarization_middleware)
 
diff --git a/offline-backend-20260512/backend/packages/harness/deerflow/agents/lead_agent/prompt.py b/offline-backend-20260512/backend/packages/harness/deerflow/agents/lead_agent/prompt.py
index 2555743d..839a83e0 100644
--- a/offline-backend-20260512/backend/packages/harness/deerflow/agents/lead_agent/prompt.py
+++ b/offline-backend-20260512/backend/packages/harness/deerflow/agents/lead_agent/prompt.py
@@ -86,8 +86,58 @@ def _drop_archived_skills(skills: list[Skill]) -> list[Skill]:
         return skills
 
 
+def _run_coro_blocking(coro):
+    """Run an async coroutine to completion from sync code.
+
+    Mirrors ``skill_tools._run_coro_blocking``: when already inside a running
+    event loop (e.g. prompt built during an async agent run), hand the coroutine
+    to a fresh loop in a worker thread; otherwise run it directly.
+    """
+    try:
+        asyncio.get_running_loop()
+        in_loop = True
+    except RuntimeError:
+        in_loop = False
+
+    if in_loop:
+        import concurrent.futures
+
+        with concurrent.futures.ThreadPoolExecutor(max_workers=1) as pool:
+            return pool.submit(asyncio.run, coro).result()
+    return asyncio.run(coro)
+
+
+def _load_always_on_names() -> set[str]:
+    """Names marked 常驻(免压缩), read from the **DB** (skills table).
+
+    Returns an empty set on any failure / no-DB deployment, so compression then
+    simply treats every skill as compressible (safe default).
+    """
+    try:
+        from deerflow.persistence.engine import get_session_factory
+        from deerflow.persistence.skills import make_skill_store
+
+        store = make_skill_store(get_session_factory())
+        return set(_run_coro_blocking(store.list_always_on_names()))
+    except Exception:
+        logger.debug("Failed to load always_on skill names from DB", exc_info=True)
+        return set()
+
+
+def _apply_always_on(skills: list[Skill]) -> list[Skill]:
+    """Stamp ``Skill.always_on`` from the DB-backed 常驻 set (in place)."""
+    try:
+        always_on = _load_always_on_names()
+        if always_on:
+            for skill in skills:
+                skill.always_on = skill.name in always_on
+    except Exception:
+        logger.debug("Failed to apply always_on flags to skills", exc_info=True)
+    return skills
+
+
 def _load_enabled_skills_sync() -> list[Skill]:
-    return _drop_archived_skills(list(get_or_new_skill_storage().load_skills(enabled_only=True)))
+    return _apply_always_on(_drop_archived_skills(list(get_or_new_skill_storage().load_skills(enabled_only=True))))
 
 
 def _start_enabled_skills_refresh_thread() -> None:
@@ -188,7 +238,7 @@ def _get_enabled_skills_for_config(app_config: AppConfig | None = None) -> list[
     """
     if app_config is None:
         return _get_enabled_skills()
-    return _drop_archived_skills(list(get_or_new_skill_storage(app_config=app_config).load_skills(enabled_only=True)))
+    return _apply_always_on(_drop_archived_skills(list(get_or_new_skill_storage(app_config=app_config).load_skills(enabled_only=True))))
 
 
 def _skill_mutability_label(category: SkillCategory | str) -> str:
@@ -758,51 +808,83 @@ def _get_memory_context(agent_name: str | None = None, *, app_config: AppConfig
         from deerflow.config.memory_config import get_effective_memory_config
         from deerflow.runtime.user_context import get_effective_user_id
 
-        user_id = get_effective_user_id()
-        config = get_effective_memory_config(user_id)
-
-        manager = build_memory_manager(
-            user_id=user_id,
-            agent_id=agent_name or "default",
-            config=config,
-            app_config=app_config,
-        )
+        user_id = get_effective_user_id()
+        config = get_effective_memory_config(user_id)
+
+        manager = build_memory_manager(
+            user_id=user_id,
+            agent_id=agent_name or "default",
+            config=config,
+            app_config=app_config,
+        )
         # 只取静态块 (builtin 的 frozen snapshot + Hindsight 的模式说明)。
         # Hindsight 的动态召回需要当前 query,走 MemoryMiddleware.before_model。
-        context_content = manager.build_system_prompt()
-
-        if not context_content.strip():
-            return ""
-
-        return f"{context_content}\n"
+        context_content = manager.build_system_prompt()
+
+        if not context_content.strip():
+            return ""
+
+        return f"{context_content}\n"
     except Exception:
         logger.exception("Failed to load memory context")
         return ""
 
 
+def _render_skill_block(name: str, description: str, category: str, location: str, display_hint: str) -> str:
+    return f"    <skill>\n        <name>{name}</name>\n        <description>{description} {_skill_mutability_label(category)}</description>\n        <location>{location}</location>{display_hint}\n    </skill>"
+
+
 @lru_cache(maxsize=32)
 def _get_cached_skills_prompt_section(
-    skill_signature: tuple[tuple[str, str, str, str], ...],
+    skill_signature: tuple[tuple[str, str, str, str, bool], ...],
     display_signature: tuple[tuple[str, str, str], ...],
     available_skills_key: tuple[str, ...] | None,
     container_base_path: str,
     skill_evolution_section: str,
+    compression_enabled: bool,
+    keep_index: bool,
 ) -> str:
     display_by_skill = {name: (mode, custom_instruction) for name, mode, custom_instruction in display_signature}
-    filtered = [(name, description, category, location) for name, description, category, location in skill_signature if available_skills_key is None or name in available_skills_key]
+    filtered = [(name, description, category, location, always_on) for name, description, category, location, always_on in skill_signature if available_skills_key is None or name in available_skills_key]
+
+    def _display_hint(name: str) -> str:
+        display_mode, custom_instruction = display_by_skill.get(name, ("", ""))
+        if display_mode:
+            return f"\n        <display mode=\"{display_mode}\">{custom_instruction}</display>"
+        return ""
+
     skills_list = ""
+    compression_note = ""
     if filtered:
-        items = []
-        for name, description, category, location in filtered:
-            display_mode, custom_instruction = display_by_skill.get(name, ("", ""))
-            display_hint = ""
-            if display_mode:
-                display_hint = f"\n        <display mode=\"{display_mode}\">{custom_instruction}</display>"
-            items.append(
-                f"    <skill>\n        <name>{name}</name>\n        <description>{description} {_skill_mutability_label(category)}</description>\n        <location>{location}</location>{display_hint}\n    </skill>"
-            )
-        skill_items = "\n".join(items)
-        skills_list = f"<available_skills>\n{skill_items}\n</available_skills>"
+        if compression_enabled:
+            # 压缩开启:常驻(always_on)技能全文注入;其余技能从提示词正文剔除,
+            # 仅保留可选的"名字索引",由 agent 通过 search_skills 工具按需检索。
+            resident = [s for s in filtered if s[4]]
+            compressed = [s for s in filtered if not s[4]]
+            blocks: list[str] = []
+            for name, description, category, location, _ in resident:
+                blocks.append(_render_skill_block(name, description, category, location, _display_hint(name)))
+            if blocks:
+                skills_list = "<available_skills>\n" + "\n".join(blocks) + "\n</available_skills>"
+            if compressed:
+                index_block = ""
+                if keep_index:
+                    index_items = "\n".join(f"    <skill><name>{name}</name></skill>" for name, _, _, _, _ in compressed)
+                    index_block = f"\n<compressed_skills>\n{index_items}\n</compressed_skills>"
+                compression_note = (
+                    "\n**Additional skills are available but not listed in full above to keep this prompt compact.**\n"
+                    "When a user query might match a skill that is NOT fully described above, call the `search_skills` "
+                    "tool with a short keyword query to discover matching skills (it returns each match's description and "
+                    f"`location`), then `read_file` the returned `location` and follow the skill. There are {len(compressed)} "
+                    "such compressed skill(s)."
+                    f"{index_block}\n"
+                )
+        else:
+            items = [
+                _render_skill_block(name, description, category, location, _display_hint(name))
+                for name, description, category, location, _ in filtered
+            ]
+            skills_list = "<available_skills>\n" + "\n".join(items) + "\n</available_skills>"
 
     # Restricted mode: agent has an explicit skill allowlist.
     # We deliberately omit the "Skills are located at: <path>" hint and the
@@ -842,7 +924,7 @@ You have access to skills that provide optimized workflows for specific tasks. E
 4. Load referenced resources only when needed during execution
 5. Follow the skill's instructions precisely
 {display_pattern}
-{location_hint}{evolution_block}
+{location_hint}{evolution_block}{compression_note}
 {skills_list}
 
 </skill_system>"""
@@ -877,13 +959,27 @@ def get_skills_prompt_section(
     is_restricted = available_skills is not None
     effective_skill_evolution_enabled = skill_evolution_enabled and not is_restricted
 
+    # 技能压缩仅作用于默认 lead agent 的全局技能集;自定义 agent 已有显式白名单
+    # (技能数量本就受控),不参与压缩。
+    compression_enabled = False
+    keep_index = True
+    if not is_restricted:
+        try:
+            from deerflow.config.system_settings import get_skill_compression_settings
+
+            compression_settings = get_skill_compression_settings()
+            compression_enabled = compression_settings.enabled
+            keep_index = compression_settings.keep_index_in_prompt
+        except Exception:
+            logger.debug("Failed to read skill compression settings for prompt injection", exc_info=True)
+
     if not skills and not effective_skill_evolution_enabled:
         return ""
 
     if available_skills is not None and not any(skill.name in available_skills for skill in skills):
         return ""
 
-    skill_signature = tuple((skill.name, skill.description, skill.category, skill.get_container_file_path(container_base_path)) for skill in skills)
+    skill_signature = tuple((skill.name, skill.description, skill.category, skill.get_container_file_path(container_base_path), skill.always_on) for skill in skills)
     display_signature: tuple[tuple[str, str, str], ...] = ()
     if display_enabled:
         try:
@@ -907,7 +1003,7 @@ def get_skills_prompt_section(
     if not skill_signature and available_key is not None:
         return ""
     skill_evolution_section = _build_skill_evolution_section(effective_skill_evolution_enabled)
-    return _get_cached_skills_prompt_section(skill_signature, display_signature, available_key, container_base_path, skill_evolution_section)
+    return _get_cached_skills_prompt_section(skill_signature, display_signature, available_key, container_base_path, skill_evolution_section, compression_enabled, keep_index)
 
 
 def get_agent_soul(agent_id: str | None) -> str:
diff --git a/offline-backend-20260512/backend/packages/harness/deerflow/agents/middlewares/summarization_middleware.py b/offline-backend-20260512/backend/packages/harness/deerflow/agents/middlewares/summarization_middleware.py
index 276f7adf..11e7d78c 100644
--- a/offline-backend-20260512/backend/packages/harness/deerflow/agents/middlewares/summarization_middleware.py
+++ b/offline-backend-20260512/backend/packages/harness/deerflow/agents/middlewares/summarization_middleware.py
@@ -18,6 +18,7 @@ post-compaction behavior).
 from __future__ import annotations
 
 import logging
+import threading
 from collections import OrderedDict
 from collections.abc import Awaitable, Callable, Collection
 from dataclasses import dataclass
@@ -36,6 +37,15 @@ logger = logging.getLogger(__name__)
 # many threads can't grow it without limit. Oldest entries are evicted first.
 _SUMMARY_CACHE_MAX_THREADS = 512
 
+# After compaction the preserved window must land *comfortably* below every
+# trigger threshold, otherwise the "compressed" context is still over the
+# trigger and re-summarizes on the very next turn (the sticky boundary never
+# sticks). We force the kept suffix down to this fraction of each trigger so a
+# re-compaction only happens once meaningfully *new* content accumulates — not
+# on every follow-up question. ``keep`` only sets a *target*; this is the hard
+# headroom guarantee on top of it.
+_KEEP_HEADROOM_FRACTION = 0.75
+
 
 @dataclass(frozen=True)
 class SummarizationEvent:
@@ -131,6 +141,27 @@ class _SummaryCacheEntry:
     reusable: bool
 
 
+# Process-wide summary cache shared across every middleware instance.
+#
+# ``make_lead_agent`` rebuilds the agent — and therefore a *fresh*
+# ``DeerFlowSummarizationMiddleware`` — on every run. If the cache lived on the
+# instance (``self._summary_cache``) it would be empty at the start of each
+# user turn, forcing a brand-new (blocking) summarization LLM call on *every*
+# turn of a long conversation. Hoisting it to module scope, keyed by
+# ``thread_id``, lets a summary computed on one turn be reused on the next,
+# so only the first threshold crossing (or a genuine re-compaction) pays for an
+# LLM call. A lock guards the non-atomic LRU eviction against concurrent
+# threads (subagents / parallel conversations share this process).
+_GLOBAL_SUMMARY_CACHE: "OrderedDict[str, _SummaryCacheEntry]" = OrderedDict()
+_GLOBAL_SUMMARY_CACHE_LOCK = threading.Lock()
+
+
+def _clear_global_summary_cache() -> None:
+    """Drop every cached summary. Intended for tests/diagnostics only."""
+    with _GLOBAL_SUMMARY_CACHE_LOCK:
+        _GLOBAL_SUMMARY_CACHE.clear()
+
+
 class DeerFlowSummarizationMiddleware(SummarizationMiddleware):
     """Summarization middleware with pre-compression hook dispatch and skill rescue."""
 
@@ -153,8 +184,9 @@ class DeerFlowSummarizationMiddleware(SummarizationMiddleware):
         self._preserve_recent_skill_tokens = max(0, preserve_recent_skill_tokens)
         self._preserve_recent_skill_tokens_per_skill = max(0, preserve_recent_skill_tokens_per_skill)
         # Per-thread cache of the most recent summary: thread_id -> (summarized message ids, summary text).
-        # Lets us reuse a summary across model calls instead of re-running the LLM every invocation.
-        self._summary_cache: OrderedDict[str, _SummaryCacheEntry] = OrderedDict()
+        # Points at the *process-wide* cache (not a fresh per-instance dict) so a
+        # summary survives across runs/turns — see ``_GLOBAL_SUMMARY_CACHE``.
+        self._summary_cache: OrderedDict[str, _SummaryCacheEntry] = _GLOBAL_SUMMARY_CACHE
 
     @override
     def before_model(self, state: AgentState, runtime: Runtime) -> dict | None:
@@ -208,6 +240,7 @@ class DeerFlowSummarizationMiddleware(SummarizationMiddleware):
             return payload  # type: ignore[return-value]
         # action == "summarize"
         messages_to_summarize, preserved, thread_id = payload  # type: ignore[misc]
+        self._emit_compacting_notice()
         self._fire_hooks(messages_to_summarize, preserved, getattr(request, "runtime", None))
         summary = self._create_summary(messages_to_summarize)
         return self._finish_summarize(messages_to_summarize, preserved, summary, thread_id)
@@ -221,10 +254,40 @@ class DeerFlowSummarizationMiddleware(SummarizationMiddleware):
         if action == "reuse":
             return payload  # type: ignore[return-value]
         messages_to_summarize, preserved, thread_id = payload  # type: ignore[misc]
+        self._emit_compacting_notice()
         self._fire_hooks(messages_to_summarize, preserved, getattr(request, "runtime", None))
         summary = await self._acreate_summary(messages_to_summarize)
         return self._finish_summarize(messages_to_summarize, preserved, summary, thread_id)
 
+    @staticmethod
+    def _emit_compacting_notice() -> None:
+        """Best-effort UI hint that a (blocking) summarization LLM call is starting.
+
+        Only the genuine summarize path calls this — the cheap cache-reuse path
+        does no LLM work and stays silent. Emitted on LangGraph's ``custom``
+        stream channel; the chat frontend renders it as a transient toast. A
+        no-op when there is no active stream writer (e.g. unit tests, non-stream
+        runs), so callers never branch.
+        """
+        # Always leave a backend trace of the genuine (blocking) compaction, even
+        # when there is no stream writer to render the frontend toast (scheduled
+        # runs, channels, tests). Grep ``上下文压缩`` / ``context compaction`` in
+        # the Gateway log to confirm a real summarization LLM call fired.
+        logger.info("summarization: 上下文压缩(真正摘要,触发一次 LLM 调用) context compaction firing")
+
+        try:
+            from langgraph.config import get_stream_writer
+
+            writer = get_stream_writer()
+        except Exception:
+            return
+        if writer is None:
+            return
+        try:
+            writer({"type": "context_compacting", "message": "正在压缩上下文以容纳更长对话…"})
+        except Exception:
+            logger.debug("summarization: failed to emit context_compacting notice", exc_info=True)
+
     def _plan_compression(
         self,
         request: ModelRequest,
@@ -243,27 +306,71 @@ class DeerFlowSummarizationMiddleware(SummarizationMiddleware):
         if not messages:
             return None
         self._ensure_message_ids(messages)
-        thread_id = self._resolve_request_thread_id(request)
 
-        cached = self._summary_cache.get(thread_id) if thread_id else None
+        # Threshold is the first gate: while the *full* current conversation is
+        # still under the configured trigger, there is nothing to compress —
+        # pass it through untouched. No cache lookup, no summary substitution.
+        # Compression only ever engages once the real context crosses the
+        # trigger, which is exactly when it is needed.
+        if not self._should_summarize(messages, self.token_counter(messages)):
+            return None
+
+        thread_id = self._resolve_request_thread_id(request)
+        if thread_id is None:
+            logger.debug("summarization: thread_id unresolved; cross-turn summary cache disabled for this call")
+        cached = self._get_cache_entry(thread_id) if thread_id else None
         present_ids = {m.id for m in messages}
-        if cached is not None and cached.reusable and cached.summarized_ids <= present_ids:
+        cache_usable = cached is not None and cached.reusable and cached.summarized_ids <= present_ids
+        if cache_usable:
             preserved = [m for m in messages if m.id not in cached.summarized_ids]
             effective = [self._summary_message(cached.summary_text), *preserved]
-            if not self._should_summarize(effective, self.token_counter(effective)):
+            # Decide re-summarization on the *counted* size of the compressed
+            # context only — NOT ``_should_summarize`` — because that folds in the
+            # last AIMessage's reported ``usage_metadata`` (system prompt + tool
+            # schemas, fixed overhead compression can't shrink). Using it here
+            # would re-summarize on every single turn once that overhead alone
+            # approaches the trigger.
+            if not self._exceeds_trigger_by_count(effective):
                 # Refresh LRU recency and reuse without an LLM call.
-                self._summary_cache.move_to_end(thread_id)
+                self._touch_cache_entry(thread_id)
+                logger.info(
+                    "summarization: 上下文压缩(复用缓存摘要,无 LLM 调用) context compaction reusing cached summary "
+                    "thread=%s full_tokens=%d compressed_tokens=%d preserved_msgs=%d",
+                    thread_id,
+                    self.token_counter(messages),
+                    self.token_counter(effective),
+                    len(preserved),
+                )
                 return ("reuse", effective)
             # New messages have pushed the compressed context back over the
             # threshold — fall through and re-summarize a larger prefix.
 
-        if not self._should_summarize(messages, self.token_counter(messages)):
-            return None
+        # Genuine (blocking) LLM summarize path. Log *why* we are paying for it so
+        # an "it compacts every turn" report is diagnosable straight from the
+        # backend log: cache=miss → thread_id/key problem; unusable → message-id
+        # instability; stale → preserved genuinely grew back over the trigger.
+        cache_state = "stale" if cache_usable else ("miss" if cached is None else "unusable")
+        logger.info(
+            "summarization: compacting thread=%s cache=%s counted_tokens=%d msgs=%d",
+            thread_id,
+            cache_state,
+            self.token_counter(messages),
+            len(messages),
+        )
 
         cutoff_index = self._determine_cutoff_index(messages)
         if cutoff_index <= 0:
             return None
 
+        # Force the preserved window below every trigger (with headroom) so the
+        # post-compaction context doesn't immediately re-trigger. Without this,
+        # a ``keep`` expressed in *messages* (e.g. 20) can itself blow past a
+        # *token* trigger (e.g. 60000) when recent messages are large (tool
+        # outputs, pasted docs) — and then every follow-up question re-summarizes.
+        cutoff_index = self._apply_stickiness_headroom(messages, cutoff_index)
+        if cutoff_index <= 0:
+            return None
+
         messages_to_summarize, preserved = self._partition_with_skill_rescue(messages, cutoff_index)
         if not messages_to_summarize:
             return None
@@ -294,12 +401,99 @@ class DeerFlowSummarizationMiddleware(SummarizationMiddleware):
 
         return compressed
 
+    def _get_cache_entry(self, thread_id: str) -> _SummaryCacheEntry | None:
+        """Return the cached entry for ``thread_id`` (thread-safe read)."""
+        with _GLOBAL_SUMMARY_CACHE_LOCK:
+            return self._summary_cache.get(thread_id)
+
+    def _touch_cache_entry(self, thread_id: str) -> None:
+        """Refresh LRU recency for ``thread_id`` if still present (thread-safe)."""
+        with _GLOBAL_SUMMARY_CACHE_LOCK:
+            if thread_id in self._summary_cache:
+                self._summary_cache.move_to_end(thread_id)
+
     def _store_cache_entry(self, thread_id: str, entry: _SummaryCacheEntry) -> None:
         """Insert/refresh a cache entry, evicting the oldest thread if over capacity."""
-        self._summary_cache[thread_id] = entry
-        self._summary_cache.move_to_end(thread_id)
-        while len(self._summary_cache) > _SUMMARY_CACHE_MAX_THREADS:
-            self._summary_cache.popitem(last=False)
+        with _GLOBAL_SUMMARY_CACHE_LOCK:
+            self._summary_cache[thread_id] = entry
+            self._summary_cache.move_to_end(thread_id)
+            while len(self._summary_cache) > _SUMMARY_CACHE_MAX_THREADS:
+                self._summary_cache.popitem(last=False)
+
+    def _exceeds_trigger_by_count(self, messages: list[AnyMessage]) -> bool:
+        """Trigger check using only the *counted* size of ``messages``.
+
+        Deliberately ignores the stock middleware's reported-``usage_metadata``
+        heuristic: that number includes fixed system-prompt + tool-schema
+        overhead which compression can never shrink, so relying on it to decide
+        *re*-summarization makes a long, tool-heavy thread re-compact on every
+        turn. The first-ever decision to compress still uses the full
+        ``_should_summarize`` (reported usage is a fair signal there); only the
+        cache-reuse-vs-recompact decision uses this stricter, content-only check.
+        """
+        total = self.token_counter(messages)
+        for kind, value in getattr(self, "_trigger_conditions", []):
+            if kind == "messages" and len(messages) >= value:
+                return True
+            if kind == "tokens" and total >= value:
+                return True
+        return False
+
+    def _apply_stickiness_headroom(self, messages: list[AnyMessage], cutoff_index: int) -> int:
+        """Advance ``cutoff_index`` (keep fewer recent messages) so the preserved
+        suffix lands below every trigger threshold with headroom.
+
+        Only ever *raises* the cutoff (summarizes more / keeps less); never keeps
+        more than the caller's ``keep`` policy already chose. This is what makes
+        the sticky boundary actually stick: after compaction the kept context is
+        guaranteed under ~``_KEEP_HEADROOM_FRACTION`` of each trigger, so a
+        re-compaction waits for genuinely new content instead of firing every turn.
+        """
+        n = len(messages)
+        if n <= 1:
+            return cutoff_index
+
+        for kind, value in getattr(self, "_trigger_conditions", []):
+            if kind == "messages":
+                keep_cap = max(1, int(int(value) * _KEEP_HEADROOM_FRACTION))
+                cutoff_index = max(cutoff_index, n - keep_cap)
+            elif kind == "tokens":
+                target = max(1, int(int(value) * _KEEP_HEADROOM_FRACTION))
+                cutoff_index = max(cutoff_index, self._suffix_token_cutoff(messages, target))
+            # "fraction" triggers depend on model profile limits; the message/token
+            # caps above (when present) already bound the window, so skip them here.
+
+        # Always keep at least the final message, and re-snap to a safe AI/Tool
+        # boundary so we never orphan a tool result.
+        cutoff_index = min(cutoff_index, n - 1)
+        return self._find_safe_cutoff_point(messages, cutoff_index)
+
+    def _suffix_token_cutoff(self, messages: list[AnyMessage], target_tokens: int) -> int:
+        """Smallest index ``i`` such that ``token_counter(messages[i:]) <= target_tokens``.
+
+        Binary search mirroring the stock token-based cutoff; returns 0 when the
+        whole list already fits, and never returns the full length (keeps ≥1).
+        """
+        n = len(messages)
+        if n == 0:
+            return 0
+        if self.token_counter(messages) <= target_tokens:
+            return 0
+
+        left, right = 0, n
+        cutoff = n
+        for _ in range(n.bit_length() + 1):
+            if left >= right:
+                break
+            mid = (left + right) // 2
+            if self.token_counter(messages[mid:]) <= target_tokens:
+                cutoff = mid
+                right = mid
+            else:
+                left = mid + 1
+        if cutoff == n:
+            cutoff = left
+        return min(cutoff, n - 1)
 
     @staticmethod
     def _resolve_request_thread_id(request: ModelRequest) -> str | None:
diff --git a/offline-backend-20260512/backend/packages/harness/deerflow/client.py b/offline-backend-20260512/backend/packages/harness/deerflow/client.py
index c970eb29..35374e24 100644
--- a/offline-backend-20260512/backend/packages/harness/deerflow/client.py
+++ b/offline-backend-20260512/backend/packages/harness/deerflow/client.py
@@ -840,13 +840,13 @@ class DeerFlowClient:
 
         current_config = get_extensions_config()
 
-        config_data = {
-            "mcpServers": mcp_servers,
-            "skills": {
-                name: skill.model_dump(by_alias=True, exclude_none=True)
-                for name, skill in current_config.skills.items()
-            },
-        }
+        config_data = {
+            "mcpServers": mcp_servers,
+            "skills": {
+                name: skill.model_dump(by_alias=True, exclude_none=True)
+                for name, skill in current_config.skills.items()
+            },
+        }
 
         self._atomic_write_json(config_path, config_data)
 
@@ -906,20 +906,20 @@ class DeerFlowClient:
         if config_path is None:
             raise FileNotFoundError("Cannot locate extensions_config.json. Set DEER_FLOW_EXTENSIONS_CONFIG_PATH or ensure it exists in the project root.")
 
-        extensions_config = get_extensions_config()
-        current_skill_config = extensions_config.skills.get(name)
-        extensions_config.skills[name] = SkillStateConfig(
-            enabled=enabled,
-            display=current_skill_config.display if current_skill_config else None,
-        )
-
-        config_data = {
-            "mcpServers": {n: s.model_dump(by_alias=True) for n, s in extensions_config.mcp_servers.items()},
-            "skills": {
-                n: sc.model_dump(by_alias=True, exclude_none=True)
-                for n, sc in extensions_config.skills.items()
-            },
-        }
+        extensions_config = get_extensions_config()
+        current_skill_config = extensions_config.skills.get(name)
+        extensions_config.skills[name] = SkillStateConfig(
+            enabled=enabled,
+            display=current_skill_config.display if current_skill_config else None,
+        )
+
+        config_data = {
+            "mcpServers": {n: s.model_dump(by_alias=True) for n, s in extensions_config.mcp_servers.items()},
+            "skills": {
+                n: sc.model_dump(by_alias=True, exclude_none=True)
+                for n, sc in extensions_config.skills.items()
+            },
+        }
 
         self._atomic_write_json(config_path, config_data)
 
diff --git a/offline-backend-20260512/backend/packages/harness/deerflow/config/system_settings.py b/offline-backend-20260512/backend/packages/harness/deerflow/config/system_settings.py
index 21e46372..e95183c1 100644
--- a/offline-backend-20260512/backend/packages/harness/deerflow/config/system_settings.py
+++ b/offline-backend-20260512/backend/packages/harness/deerflow/config/system_settings.py
@@ -170,6 +170,57 @@ class LeaderboardSettings(BaseModel):
     model_config = ConfigDict(extra="ignore")
 
 
+class SkillCompressionSettings(BaseModel):
+    """Admin-configurable skill compression (技能压缩).
+
+    When ``enabled``, only skills explicitly marked 常驻(免压缩) (``always_on``)
+    are injected full-text into the lead agent's system prompt. Every other
+    enabled skill is "compressed": kept out of the prompt body and surfaced to
+    the agent on demand via the ``search_skills`` keyword-retrieval tool (and a
+    compact name index when ``keep_index_in_prompt`` is true). When disabled the
+    behaviour is unchanged — every enabled skill is injected full-text.
+    """
+
+    enabled: bool = Field(default=False, description="Whether skill compression is active (master switch)")
+    search_top_n: int = Field(default=5, ge=1, le=50, description="Max skills returned by the search_skills tool")
+    keep_index_in_prompt: bool = Field(
+        default=True,
+        description="Whether compressed skills still appear as a compact name-only index in the prompt",
+    )
+    model_config = ConfigDict(extra="ignore")
+
+
+class AgentModuleSettings(BaseModel):
+    """Admin-configured「按类型聚合」module → selected agent-tag ids.
+
+    Keyed by module key (e.g. ``"zzjc"`` / ``"zzxd"``). Each value is the list
+    of agent **tag ids** whose agents that module's left sidebar shows. A missing
+    key / empty list means "not configured" — the frontend falls back to showing
+    all visible agents grouped by tag.
+    """
+
+    modules: dict[str, list[str]] = Field(
+        default_factory=dict,
+        description="module key → selected agent tag ids",
+    )
+    model_config = ConfigDict(extra="ignore")
+
+
+def get_module_tag_ids(settings: AgentModuleSettings, key: str) -> list[str]:
+    """Return the configured agent-tag ids for ``key`` (empty when unset)."""
+    return list(settings.modules.get(key, []))
+
+
+def set_module_tag_ids(
+    settings: AgentModuleSettings, key: str, tag_ids: list[str]
+) -> AgentModuleSettings:
+    """Return a copy with ``key`` set to a de-duplicated, stripped tag-id list."""
+    cleaned = list(dict.fromkeys([str(t).strip() for t in tag_ids if str(t).strip()]))
+    modules = dict(settings.modules)
+    modules[key] = cleaned
+    return settings.model_copy(update={"modules": modules})
+
+
 class SystemSettings(BaseModel):
     """Top-level system settings document."""
 
@@ -179,6 +230,8 @@ class SystemSettings(BaseModel):
     appearance: AppearanceSettings = Field(default_factory=AppearanceSettings)
     citation_display: CitationDisplaySettings = Field(default_factory=CitationDisplaySettings)
     leaderboard: LeaderboardSettings = Field(default_factory=LeaderboardSettings)
+    skill_compression: SkillCompressionSettings = Field(default_factory=SkillCompressionSettings)
+    agent_modules: AgentModuleSettings = Field(default_factory=AgentModuleSettings)
     model_config = ConfigDict(extra="ignore")
 
 
@@ -204,6 +257,15 @@ def load_system_settings() -> SystemSettings:
         return SystemSettings()
 
 
+def get_skill_compression_settings() -> SkillCompressionSettings:
+    """Convenience accessor for the skill-compression block (safe defaults on error)."""
+    try:
+        return load_system_settings().skill_compression
+    except Exception:
+        logger.debug("Failed to read skill compression settings; using defaults", exc_info=True)
+        return SkillCompressionSettings()
+
+
 def save_system_settings(settings: SystemSettings) -> SystemSettings:
     """Persist the settings document atomically, returning the saved value."""
     path = _settings_path()
diff --git a/offline-backend-20260512/backend/packages/harness/deerflow/knowledge/extractor.py b/offline-backend-20260512/backend/packages/harness/deerflow/knowledge/extractor.py
index a9dbbe63..13f673ea 100644
--- a/offline-backend-20260512/backend/packages/harness/deerflow/knowledge/extractor.py
+++ b/offline-backend-20260512/backend/packages/harness/deerflow/knowledge/extractor.py
@@ -120,14 +120,18 @@ def related_links(project_name: str | None) -> list[str]:  # noqa: ARG001
 _related_links = related_links
 
 
-def _related_section_md(related_lines: list[str], source_lines: list[str]) -> str:
-    """Render the ``## Related`` section, or ``""`` when there is nothing to show."""
-    if not related_lines and not source_lines:
+def _related_section_md(related_lines: list[str], source_lines: list[str] | None = None) -> str:  # noqa: ARG001
+    """Render the ``## Related`` section, or ``""`` when there is nothing to show.
+
+    Sources are intentionally **not** rendered into the body anymore. The note
+    detail page already shows every source once via the structured ``来源`` block
+    (``note.sources``); emitting a ``### Sources`` sub-section here duplicated that
+    list (and broke the frontend's clean ``## Related`` extraction). The
+    ``source_lines`` argument is kept for backward compatibility and ignored.
+    """
+    if not related_lines:
         return ""
     inner = "\n".join(related_lines)
-    if source_lines:
-        sources_block = "### Sources\n\n" + "\n".join(source_lines)
-        inner = (inner + "\n\n" + sources_block) if inner else sources_block
     return f"## Related\n\n{inner}\n"
 
 
diff --git a/offline-backend-20260512/backend/packages/harness/deerflow/knowledge/llm_extractor.py b/offline-backend-20260512/backend/packages/harness/deerflow/knowledge/llm_extractor.py
index e9d924a2..e57424a5 100644
--- a/offline-backend-20260512/backend/packages/harness/deerflow/knowledge/llm_extractor.py
+++ b/offline-backend-20260512/backend/packages/harness/deerflow/knowledge/llm_extractor.py
@@ -283,13 +283,16 @@ async def extract_thread_llm_multi(
     max_input_chars: int = 12000,
     allow_multi: bool = True,
     max_notes: int = 4,
+    system_prompt: str | None = None,
 ) -> list[NoteDraft]:
     """Distill a thread into one *or several* cross-linkable knowledge notes.
 
     The LLM may split a conversation that covers multiple independent topics into
     several declarative notes. Returns ``[]`` when there is nothing to sediment
     or the LLM path fails — callers should fall back to the rule-based extractor.
-    When ``allow_multi`` is false only the first note is kept.
+    When ``allow_multi`` is false only the first note is kept. ``system_prompt``
+    overrides the built-in distillation prompt with an operator-edited template
+    (used as-is, since the template is expected to be self-contained).
     """
     turns = extractor.split_turns(messages, max_messages=max_messages)
     if not turns:
@@ -303,8 +306,9 @@ async def extract_thread_llm_multi(
 
         model = create_chat_model(name=model_name or None, thinking_enabled=False)
         transcript = _build_transcript(turns, sources, max_chars=max_input_chars)
+        system_text = (system_prompt or "").strip() or (_SYSTEM_PROMPT + "\n\n" + WIKI_CAPTURE_GUIDANCE)
         prompt = [
-            SystemMessage(content=_SYSTEM_PROMPT + "\n\n" + WIKI_CAPTURE_GUIDANCE),
+            SystemMessage(content=system_text),
             HumanMessage(content=f"以下是一次对话，请按规范提炼为知识 JSON：\n\n{transcript}"),
         ]
         response = await model.ainvoke(prompt)
@@ -375,6 +379,7 @@ async def extract_document_llm_multi(
     max_input_chars: int = 12000,
     allow_multi: bool = True,
     max_notes: int = 4,
+    system_prompt: str | None = None,
 ) -> list[NoteDraft]:
     """Distill an imported **document** into one *or several* knowledge notes.
 
@@ -397,8 +402,9 @@ async def extract_document_llm_multi(
         model = create_chat_model(name=model_name or None, thinking_enabled=False)
         body = text if len(text) <= max_input_chars else text[:max_input_chars] + "\n…（已截断）"
         header = f"文档标题：{doc_title}\n\n" if doc_title else ""
+        system_text = (system_prompt or "").strip() or (_DOC_SYSTEM_PROMPT + "\n\n" + WIKI_CAPTURE_GUIDANCE)
         prompt = [
-            SystemMessage(content=_DOC_SYSTEM_PROMPT + "\n\n" + WIKI_CAPTURE_GUIDANCE),
+            SystemMessage(content=system_text),
             HumanMessage(content=f"以下是一篇文档，请按规范提炼为知识 JSON：\n\n{header}{body}"),
         ]
         response = await model.ainvoke(prompt)
diff --git a/offline-backend-20260512/backend/packages/harness/deerflow/knowledge/models.py b/offline-backend-20260512/backend/packages/harness/deerflow/knowledge/models.py
index bab17e68..480e5965 100644
--- a/offline-backend-20260512/backend/packages/harness/deerflow/knowledge/models.py
+++ b/offline-backend-20260512/backend/packages/harness/deerflow/knowledge/models.py
@@ -33,6 +33,7 @@ class KnowledgeNoteRow(Base):
     __table_args__ = (
         Index("ix_knowledge_notes_vault_path", "vault_path", mysql_length=191),
         Index("ix_knowledge_notes_source_id", "source_id", mysql_length=191),
+        Index("ix_knowledge_notes_folder", "folder", mysql_length=191),
     )
 
     id: Mapped[str] = mapped_column(String(64), primary_key=True)
@@ -40,6 +41,10 @@ class KnowledgeNoteRow(Base):
     summary: Mapped[str | None] = mapped_column(Text)
     content_md: Mapped[str] = mapped_column(PortableLongText())
     vault_path: Mapped[str | None] = mapped_column(String(1024))
+    # User-assigned directory (slash-joined path, e.g. "投研/行业"). Independent
+    # of the wiki-capture ``category`` and of ``vault_path``; drives the product
+    # directory tree. NULL = unfiled (shown under a fallback group).
+    folder: Mapped[str | None] = mapped_column(String(512))
     source_type: Mapped[str] = mapped_column(String(64), index=True)
     source_id: Mapped[str | None] = mapped_column(String(256))
     status: Mapped[str] = mapped_column(String(32), default="approved", index=True)
@@ -170,3 +175,49 @@ class KnowledgeRelationRow(Base):
     relation_type: Mapped[str] = mapped_column(String(128))
     weight: Mapped[float] = mapped_column(Float, default=1.0)
     created_at: Mapped[datetime] = mapped_column(BeijingDateTime(), default=lambda: datetime.now(UTC))
+
+
+class KnowledgeFolderRow(Base):
+    """User-managed knowledge directory (mirrors ``knowledge_folders``).
+
+    A folder is identified by its full slash-joined ``path`` (e.g. ``投研/行业``);
+    nesting is derived by splitting on ``/`` on the client. Folders exist
+    independently of notes so an operator can create an empty directory and file
+    notes into it later (``knowledge_notes.folder`` stores the chosen path).
+    Global/shared like the rest of the KB; ``created_by`` is audit-only.
+    """
+
+    __tablename__ = "knowledge_folders"
+    __table_args__ = (Index("uq_knowledge_folders_path", "path", unique=True, mysql_length=191),)
+
+    id: Mapped[str] = mapped_column(String(64), primary_key=True)
+    path: Mapped[str] = mapped_column(String(512))
+    name: Mapped[str] = mapped_column(String(256))
+    sort_order: Mapped[int] = mapped_column(Integer, default=0)
+    created_by: Mapped[str | None] = mapped_column(String(128))
+    created_at: Mapped[datetime] = mapped_column(BeijingDateTime(), default=lambda: datetime.now(UTC))
+
+
+class KnowledgeExtractTemplateRow(Base):
+    """Editable knowledge-extraction prompt template (mirrors ``knowledge_extract_templates``).
+
+    Lets an operator open up, edit and add multiple distillation prompts instead
+    of the hard-coded constants. ``scope`` is ``thread`` / ``document`` / ``both``;
+    at capture time the service resolves the chosen template (or the default for
+    that scope) and feeds its ``system_prompt`` to the extractor. Global/shared.
+    """
+
+    __tablename__ = "knowledge_extract_templates"
+    __table_args__ = (Index("uq_knowledge_extract_templates_name", "name", unique=True, mysql_length=191),)
+
+    id: Mapped[str] = mapped_column(String(64), primary_key=True)
+    name: Mapped[str] = mapped_column(String(256))
+    description: Mapped[str | None] = mapped_column(Text)
+    scope: Mapped[str] = mapped_column(String(16), default="both", index=True)
+    system_prompt: Mapped[str] = mapped_column(PortableLongText())
+    is_default: Mapped[bool] = mapped_column(default=False, index=True)
+    enabled: Mapped[bool] = mapped_column(default=True)
+    created_by: Mapped[str | None] = mapped_column(String(128))
+    updated_by: Mapped[str | None] = mapped_column(String(128))
+    created_at: Mapped[datetime] = mapped_column(BeijingDateTime(), default=lambda: datetime.now(UTC), index=True)
+    updated_at: Mapped[datetime] = mapped_column(BeijingDateTime(), default=lambda: datetime.now(UTC), onupdate=lambda: datetime.now(UTC))
diff --git a/offline-backend-20260512/backend/packages/harness/deerflow/knowledge/repository.py b/offline-backend-20260512/backend/packages/harness/deerflow/knowledge/repository.py
index 2cbe8f27..91f227f6 100644
--- a/offline-backend-20260512/backend/packages/harness/deerflow/knowledge/repository.py
+++ b/offline-backend-20260512/backend/packages/harness/deerflow/knowledge/repository.py
@@ -18,6 +18,8 @@ from sqlalchemy.ext.asyncio import AsyncSession, async_sessionmaker
 from deerflow.knowledge.models import (
     KnowledgeEmbeddingRow,
     KnowledgeEntityRow,
+    KnowledgeExtractTemplateRow,
+    KnowledgeFolderRow,
     KnowledgeGraphBlacklistRow,
     KnowledgeNoteRow,
     KnowledgeNoteTagRow,
@@ -127,6 +129,7 @@ class KnowledgeRepository:
         status: str,
         confidence: float = 0.0,
         vault_path: str | None = None,
+        folder: str | None = None,
         tags: list[str] | None = None,
         created_by: str | None = None,
     ) -> dict[str, Any]:
@@ -138,6 +141,7 @@ class KnowledgeRepository:
             summary=summary,
             content_md=content_md,
             vault_path=vault_path,
+            folder=(folder or None),
             source_type=source_type,
             source_id=source_id,
             status=status,
@@ -175,6 +179,8 @@ class KnowledgeRepository:
         status: str | None = None,
         tags: list[str] | None = None,
         vault_path: str | None = None,
+        folder: str | None = None,
+        set_folder: bool = False,
         updated_by: str | None = None,
     ) -> dict[str, Any] | None:
         async with self._sf() as session:
@@ -191,6 +197,10 @@ class KnowledgeRepository:
                 row.status = status
             if vault_path is not None:
                 row.vault_path = vault_path
+            # ``set_folder`` lets the caller assign *or clear* (folder=None) the
+            # directory; a bare ``folder=None`` without the flag leaves it as-is.
+            if set_folder:
+                row.folder = (folder or None)
             if updated_by is not None:
                 row.updated_by = updated_by
             row.updated_at = datetime.now(UTC)
@@ -279,6 +289,7 @@ class KnowledgeRepository:
         entity: str | None = None,
         source_type: str | None = None,
         status: str | None = None,
+        folder: str | None = None,
         include_archived: bool = False,
         limit: int = 20,
         offset: int = 0,
@@ -294,6 +305,12 @@ class KnowledgeRepository:
                 conditions.append(KnowledgeNoteRow.status != "archived")
             if source_type:
                 conditions.append(KnowledgeNoteRow.source_type == source_type)
+            if folder is not None:
+                # Exact folder, plus any nested sub-folders (``folder/…``).
+                if folder == "":
+                    conditions.append(KnowledgeNoteRow.folder.is_(None))
+                else:
+                    conditions.append(or_(KnowledgeNoteRow.folder == folder, KnowledgeNoteRow.folder.like(f"{folder}/%")))
             if keyword:
                 like = f"%{keyword}%"
                 conditions.append(or_(KnowledgeNoteRow.title.like(like), KnowledgeNoteRow.summary.like(like), KnowledgeNoteRow.content_md.like(like)))
@@ -581,3 +598,184 @@ class KnowledgeRepository:
                     )
                 )
             await session.commit()
+
+    # -- folders ----------------------------------------------------------
+    async def list_folders(self) -> list[dict[str, Any]]:
+        """Return every managed directory, ordered by sort_order then path."""
+        async with self._sf() as session:
+            stmt = select(KnowledgeFolderRow).order_by(KnowledgeFolderRow.sort_order, KnowledgeFolderRow.path)
+            return [{"id": r.id, "path": r.path, "name": r.name, "sort_order": r.sort_order} for r in (await session.execute(stmt)).scalars()]
+
+    async def create_folder(self, path: str, *, created_by: str | None = None) -> dict[str, Any]:
+        """Create a directory by its full slash-joined path (idempotent on path).
+
+        Parent directories are auto-created so ``投研/行业/寿险`` materializes all
+        three levels. Returns the leaf folder row.
+        """
+        segments = [s.strip() for s in (path or "").split("/") if s.strip()]
+        if not segments:
+            raise ValueError("folder path is empty")
+        async with self._sf() as session:
+            acc: list[str] = []
+            leaf: KnowledgeFolderRow | None = None
+            for i, seg in enumerate(segments):
+                acc.append(seg)
+                full = "/".join(acc)
+                row = await self._folder_by_path(session, full)
+                if row is None:
+                    row = KnowledgeFolderRow(id=_new_id("kbdir"), path=full, name=seg, sort_order=i, created_by=created_by, created_at=datetime.now(UTC))
+                    session.add(row)
+                    await session.flush()
+                leaf = row
+            # Capture before commit: the async session expires attributes on
+            # commit, and re-reading them outside an awaited load would error.
+            result = {"id": leaf.id, "path": leaf.path, "name": leaf.name, "sort_order": leaf.sort_order}
+            await session.commit()
+            return result
+
+    @staticmethod
+    async def _folder_by_path(session: AsyncSession, path: str) -> KnowledgeFolderRow | None:
+        return (await session.execute(select(KnowledgeFolderRow).where(KnowledgeFolderRow.path == path))).scalars().first()
+
+    async def rename_folder(self, folder_id: str, new_path: str) -> dict[str, Any] | None:
+        """Rename/move a folder and re-point every descendant folder + note.
+
+        Both the folder rows under ``old/…`` and notes whose ``folder`` is ``old``
+        or ``old/…`` are rewritten to the new prefix, so the move is atomic.
+        """
+        new_path = "/".join(s.strip() for s in (new_path or "").split("/") if s.strip())
+        if not new_path:
+            raise ValueError("folder path is empty")
+        async with self._sf() as session:
+            row = await session.get(KnowledgeFolderRow, folder_id)
+            if row is None:
+                return None
+            old = row.path
+            if new_path == old:
+                return {"id": row.id, "path": row.path, "name": row.name, "sort_order": row.sort_order}
+            for f in (await session.execute(select(KnowledgeFolderRow).where(or_(KnowledgeFolderRow.path == old, KnowledgeFolderRow.path.like(f"{old}/%"))))).scalars():
+                f.path = new_path + f.path[len(old):]
+                f.name = f.path.split("/")[-1]
+            for n in (await session.execute(select(KnowledgeNoteRow).where(or_(KnowledgeNoteRow.folder == old, KnowledgeNoteRow.folder.like(f"{old}/%"))))).scalars():
+                n.folder = new_path + (n.folder[len(old):] if n.folder else "")
+            await session.commit()
+            refreshed = await session.get(KnowledgeFolderRow, folder_id)
+            return {"id": refreshed.id, "path": refreshed.path, "name": refreshed.name, "sort_order": refreshed.sort_order}
+
+    async def delete_folder(self, folder_id: str, *, reassign_to: str | None = None) -> bool:
+        """Delete a folder (and its sub-folders); notes inside move to ``reassign_to`` (or unfiled)."""
+        async with self._sf() as session:
+            row = await session.get(KnowledgeFolderRow, folder_id)
+            if row is None:
+                return False
+            old = row.path
+            target = (reassign_to or "").strip() or None
+            for n in (await session.execute(select(KnowledgeNoteRow).where(or_(KnowledgeNoteRow.folder == old, KnowledgeNoteRow.folder.like(f"{old}/%"))))).scalars():
+                n.folder = target
+            await session.execute(delete(KnowledgeFolderRow).where(or_(KnowledgeFolderRow.path == old, KnowledgeFolderRow.path.like(f"{old}/%"))))
+            await session.commit()
+            return True
+
+    # -- extraction templates ---------------------------------------------
+    @staticmethod
+    def _template_to_dict(row: KnowledgeExtractTemplateRow) -> dict[str, Any]:
+        d = row.to_dict()
+        d["created_at"] = _iso(d.get("created_at"))
+        d["updated_at"] = _iso(d.get("updated_at"))
+        return d
+
+    async def list_templates(self, *, scope: str | None = None, enabled_only: bool = False) -> list[dict[str, Any]]:
+        async with self._sf() as session:
+            stmt = select(KnowledgeExtractTemplateRow)
+            if scope:
+                stmt = stmt.where(or_(KnowledgeExtractTemplateRow.scope == scope, KnowledgeExtractTemplateRow.scope == "both"))
+            if enabled_only:
+                stmt = stmt.where(KnowledgeExtractTemplateRow.enabled.is_(True))
+            stmt = stmt.order_by(KnowledgeExtractTemplateRow.is_default.desc(), KnowledgeExtractTemplateRow.created_at)
+            return [self._template_to_dict(r) for r in (await session.execute(stmt)).scalars()]
+
+    async def get_template(self, template_id: str) -> dict[str, Any] | None:
+        async with self._sf() as session:
+            row = await session.get(KnowledgeExtractTemplateRow, template_id)
+            return self._template_to_dict(row) if row else None
+
+    async def get_default_template(self, scope: str) -> dict[str, Any] | None:
+        """Return the default enabled template applicable to ``scope`` (thread/document)."""
+        async with self._sf() as session:
+            stmt = (
+                select(KnowledgeExtractTemplateRow)
+                .where(KnowledgeExtractTemplateRow.enabled.is_(True))
+                .where(KnowledgeExtractTemplateRow.is_default.is_(True))
+                .where(or_(KnowledgeExtractTemplateRow.scope == scope, KnowledgeExtractTemplateRow.scope == "both"))
+                .limit(1)
+            )
+            row = (await session.execute(stmt)).scalars().first()
+            return self._template_to_dict(row) if row else None
+
+    async def template_name_taken(self, name: str, *, exclude_id: str | None = None) -> bool:
+        async with self._sf() as session:
+            stmt = select(KnowledgeExtractTemplateRow.id).where(func.lower(KnowledgeExtractTemplateRow.name) == (name or "").strip().lower())
+            if exclude_id:
+                stmt = stmt.where(KnowledgeExtractTemplateRow.id != exclude_id)
+            return (await session.execute(stmt)).first() is not None
+
+    async def create_template(self, *, name: str, system_prompt: str, scope: str = "both", description: str | None = None, is_default: bool = False, enabled: bool = True, created_by: str | None = None) -> dict[str, Any]:
+        async with self._sf() as session:
+            if is_default:
+                await self._clear_default(session, scope)
+            row = KnowledgeExtractTemplateRow(
+                id=_new_id("kbtpl"),
+                name=name.strip()[:256],
+                description=description,
+                scope=scope,
+                system_prompt=system_prompt,
+                is_default=is_default,
+                enabled=enabled,
+                created_by=created_by,
+                updated_by=created_by,
+                created_at=datetime.now(UTC),
+                updated_at=datetime.now(UTC),
+            )
+            session.add(row)
+            await session.commit()
+            await session.refresh(row)
+            return self._template_to_dict(row)
+
+    async def update_template(self, template_id: str, *, name: str | None = None, system_prompt: str | None = None, scope: str | None = None, description: str | None = None, is_default: bool | None = None, enabled: bool | None = None, updated_by: str | None = None) -> dict[str, Any] | None:
+        async with self._sf() as session:
+            row = await session.get(KnowledgeExtractTemplateRow, template_id)
+            if row is None:
+                return None
+            if name is not None:
+                row.name = name.strip()[:256]
+            if system_prompt is not None:
+                row.system_prompt = system_prompt
+            if scope is not None:
+                row.scope = scope
+            if description is not None:
+                row.description = description
+            if enabled is not None:
+                row.enabled = enabled
+            if is_default:
+                await self._clear_default(session, scope or row.scope)
+                row.is_default = True
+            elif is_default is False:
+                row.is_default = False
+            row.updated_by = updated_by
+            row.updated_at = datetime.now(UTC)
+            await session.commit()
+            await session.refresh(row)
+            return self._template_to_dict(row)
+
+    async def delete_template(self, template_id: str) -> bool:
+        async with self._sf() as session:
+            result = await session.execute(delete(KnowledgeExtractTemplateRow).where(KnowledgeExtractTemplateRow.id == template_id))
+            await session.commit()
+            return bool(result.rowcount)
+
+    @staticmethod
+    async def _clear_default(session: AsyncSession, scope: str) -> None:
+        """Unset is_default on any template that would collide with a new default for ``scope``."""
+        scopes = {scope, "both"} if scope != "both" else {"thread", "document", "both"}
+        for row in (await session.execute(select(KnowledgeExtractTemplateRow).where(KnowledgeExtractTemplateRow.is_default.is_(True)).where(KnowledgeExtractTemplateRow.scope.in_(scopes)))).scalars():
+            row.is_default = False
diff --git a/offline-backend-20260512/backend/packages/harness/deerflow/knowledge/schemas.py b/offline-backend-20260512/backend/packages/harness/deerflow/knowledge/schemas.py
index c46cfdec..cd1c0dfb 100644
--- a/offline-backend-20260512/backend/packages/harness/deerflow/knowledge/schemas.py
+++ b/offline-backend-20260512/backend/packages/harness/deerflow/knowledge/schemas.py
@@ -48,6 +48,8 @@ class NoteDraft:
     source_id: str | None = None
     status: str = "approved"
     confidence: float = 0.0
+    # User-assigned directory (slash-joined path); None = unfiled.
+    folder: str | None = None
     tags: list[str] = field(default_factory=list)
     key_points: list[str] = field(default_factory=list)
     actions: list[str] = field(default_factory=list)
diff --git a/offline-backend-20260512/backend/packages/harness/deerflow/knowledge/service.py b/offline-backend-20260512/backend/packages/harness/deerflow/knowledge/service.py
index b83ca128..4c951bd1 100644
--- a/offline-backend-20260512/backend/packages/harness/deerflow/knowledge/service.py
+++ b/offline-backend-20260512/backend/packages/harness/deerflow/knowledge/service.py
@@ -145,6 +145,7 @@ class KnowledgeService:
             source_id=draft.source_id,
             status=draft.status,
             confidence=draft.confidence,
+            folder=draft.folder,
             tags=draft.tags,
             created_by=created_by,
         )
@@ -223,15 +224,21 @@ class KnowledgeService:
         created_by: str | None = None,
         skip_existing: bool = False,
         model_name: str | None = None,
+        template_id: str | None = None,
+        folder: str | None = None,
     ) -> dict[str, Any]:
         """Sediment a thread into a knowledge note.
 
         ``model_name`` optionally overrides the configured extraction model for
         this one capture (falls back to ``config.extract_model_name``).
+        ``template_id`` selects an operator-edited extraction prompt template
+        (falls back to the ``thread`` default, then the built-in prompt).
+        ``folder`` files the resulting note(s) under a user directory.
 
         Returns ``{"note": <note>, "created": bool}``. Raises
         :class:`EmptyThreadError` when there is nothing worth keeping.
         """
+        extract_prompt = await self._resolve_extract_prompt(template_id, "thread")
         # Always-on entry log: proves a capture was triggered for this thread,
         # regardless of how it ends (success / skip / fallback). Lets you tell
         # "capture never ran" apart from "logs not visible" via one grep.
@@ -276,6 +283,7 @@ class KnowledgeService:
                     model_name=model_name or self.config.extract_model_name,
                     max_input_chars=self.config.extract_max_input_chars,
                     allow_multi=self.config.multi_note_enabled,
+                    system_prompt=extract_prompt,
                 )
             except llm_extractor.LLMRejectedIngest as exc:
                 # The model judged this conversation worthless (greeting / meta
@@ -324,6 +332,8 @@ class KnowledgeService:
         normalized = normalize_status(effective_status)
         for draft in drafts:
             draft.status = normalized
+            if folder:
+                draft.folder = folder
         # Cross-link sibling notes from the same conversation by title, so the
         # split pages reference each other (resolved once all are persisted).
         if len(drafts) > 1:
@@ -370,6 +380,7 @@ class KnowledgeService:
         tags: list[str] | None = None,
         status: str = "approved",
         summary: str | None = None,
+        folder: str | None = None,
         created_by: str | None = None,
     ) -> dict[str, Any]:
         """Create a note directly from product-edited Markdown."""
@@ -388,6 +399,7 @@ class KnowledgeService:
             source_type="manual",
             source_id=None,
             status=normalize_status(status),
+            folder=(folder or None),
             tags=base_tags,
             confidence=0.5,
             related=extractor.related_links(self.project_name),
@@ -431,6 +443,8 @@ class KnowledgeService:
         created_by: str | None = None,
         model_name: str | None = None,
         skip_existing: bool = False,
+        template_id: str | None = None,
+        folder: str | None = None,
     ) -> dict[str, Any]:
         """Import a file (already converted to Markdown/text) into knowledge notes.
 
@@ -468,6 +482,7 @@ class KnowledgeService:
                 logger.info("[knowledge] file '%s' already imported (note %s); skipping", filename, existing.get("id"))
                 return {"note": existing, "created": False, "notes": [existing], "llm_fallback": False}
 
+        extract_prompt = await self._resolve_extract_prompt(template_id, "document")
         drafts: list[NoteDraft] = []
         llm_fallback = False
         if self.config.llm_extract_enabled:
@@ -482,6 +497,7 @@ class KnowledgeService:
                     model_name=model_name or self.config.extract_model_name,
                     max_input_chars=self.config.extract_max_input_chars,
                     allow_multi=self.config.multi_note_enabled,
+                    system_prompt=extract_prompt,
                 )
             except llm_extractor.LLMRejectedIngest as exc:
                 raise EmptyThreadError(f"File {filename} judged not worth ingesting: {exc}") from exc
@@ -496,6 +512,8 @@ class KnowledgeService:
         normalized = normalize_status(effective_status)
         for draft in drafts:
             draft.status = normalized
+            if folder:
+                draft.folder = folder
 
         # Cross-link sibling notes split from the same document (resolved by title
         # once all are persisted), mirroring multi-note thread capture.
@@ -526,6 +544,7 @@ class KnowledgeService:
         entity: str | None = None,
         source_type: str | None = None,
         status: str | None = None,
+        folder: str | None = None,
         limit: int = 20,
         offset: int = 0,
     ) -> dict[str, Any]:
@@ -535,6 +554,7 @@ class KnowledgeService:
             entity=entity,
             source_type=normalize_source_type(source_type) if source_type else None,
             status=status,
+            folder=folder,
             limit=limit,
             offset=offset,
         )
@@ -555,6 +575,8 @@ class KnowledgeService:
         tags: list[str] | None = None,
         status: str | None = None,
         summary: str | None = None,
+        folder: str | None = None,
+        set_folder: bool = False,
         updated_by: str | None = None,
     ) -> dict[str, Any] | None:
         current = await self.repo.get_note(note_id, with_sources=False)
@@ -576,6 +598,8 @@ class KnowledgeService:
             content_md=body,
             status=normalize_status(status) if status else None,
             tags=tags,
+            folder=folder,
+            set_folder=set_folder,
             updated_by=updated_by,
         )
         if note is None:
@@ -604,6 +628,84 @@ class KnowledgeService:
     async def list_versions(self, note_id: str) -> list[dict[str, Any]]:
         return await self.repo.list_versions(note_id)
 
+    # -- folders ----------------------------------------------------------
+    async def list_folders(self) -> list[dict[str, Any]]:
+        return await self.repo.list_folders()
+
+    async def create_folder(self, path: str, *, created_by: str | None = None) -> dict[str, Any]:
+        return await self.repo.create_folder(path, created_by=created_by)
+
+    async def rename_folder(self, folder_id: str, new_path: str) -> dict[str, Any] | None:
+        return await self.repo.rename_folder(folder_id, new_path)
+
+    async def delete_folder(self, folder_id: str, *, reassign_to: str | None = None) -> bool:
+        return await self.repo.delete_folder(folder_id, reassign_to=reassign_to)
+
+    # -- extraction templates ---------------------------------------------
+    async def _resolve_extract_prompt(self, template_id: str | None, scope: str) -> str | None:
+        """Return the system prompt for a capture, or ``None`` for the built-in default.
+
+        Resolution: explicit ``template_id`` → the scope's default template →
+        ``None`` (the extractor then uses its hard-coded prompt). Best-effort: any
+        lookup failure degrades to the built-in prompt so capture never breaks.
+        """
+        try:
+            if template_id:
+                tpl = await self.repo.get_template(template_id)
+                if tpl and tpl.get("enabled") and (tpl.get("system_prompt") or "").strip():
+                    return tpl["system_prompt"]
+            default = await self.repo.get_default_template(scope)
+            if default and (default.get("system_prompt") or "").strip():
+                return default["system_prompt"]
+        except Exception:
+            logger.exception("Failed to resolve knowledge extraction template (id=%s, scope=%s)", template_id, scope)
+        return None
+
+    async def list_extract_templates(self, *, scope: str | None = None, enabled_only: bool = False) -> list[dict[str, Any]]:
+        return await self.repo.list_templates(scope=scope, enabled_only=enabled_only)
+
+    async def get_extract_template(self, template_id: str) -> dict[str, Any] | None:
+        return await self.repo.get_template(template_id)
+
+    async def create_extract_template(self, **kwargs: Any) -> dict[str, Any]:
+        return await self.repo.create_template(**kwargs)
+
+    async def update_extract_template(self, template_id: str, **kwargs: Any) -> dict[str, Any] | None:
+        return await self.repo.update_template(template_id, **kwargs)
+
+    async def delete_extract_template(self, template_id: str) -> bool:
+        return await self.repo.delete_template(template_id)
+
+    async def template_name_taken(self, name: str, *, exclude_id: str | None = None) -> bool:
+        return await self.repo.template_name_taken(name, exclude_id=exclude_id)
+
+    async def seed_default_templates(self, *, created_by: str | None = None) -> int:
+        """Seed the two built-in distillation prompts as editable templates (idempotent).
+
+        Runs once on startup so an operator immediately sees — and can edit — the
+        prompts that drive 对话 / 文档 extraction. Existing templates with the same
+        name are left untouched. Returns the number of templates created.
+        """
+        from deerflow.knowledge import llm_extractor
+
+        guidance = llm_extractor.WIKI_CAPTURE_GUIDANCE
+        seeds = [
+            ("默认 · 对话提炼", "thread", llm_extractor._SYSTEM_PROMPT + "\n\n" + guidance, "对话沉淀的默认提炼提示词(wiki-capture 规范)。"),
+            ("默认 · 文档提炼", "document", llm_extractor._DOC_SYSTEM_PROMPT + "\n\n" + guidance, "文档导入的默认提炼提示词(wiki-capture 规范)。"),
+        ]
+        created = 0
+        for name, scope, prompt, desc in seeds:
+            try:
+                if await self.repo.template_name_taken(name):
+                    continue
+                await self.repo.create_template(name=name, system_prompt=prompt, scope=scope, description=desc, is_default=True, enabled=True, created_by=created_by)
+                created += 1
+            except Exception:
+                logger.exception("Failed to seed knowledge extraction template '%s'", name)
+        if created:
+            logger.info("[knowledge] seeded %d default extraction template(s)", created)
+        return created
+
     async def delete_note(self, note_id: str, *, updated_by: str | None = None, hard: bool = False) -> bool:
         """Soft delete (archive) by default; hard delete removes the vault file."""
         current = await self.repo.get_note(note_id, with_sources=False)
diff --git a/offline-backend-20260512/backend/packages/harness/deerflow/persistence/agents/model.py b/offline-backend-20260512/backend/packages/harness/deerflow/persistence/agents/model.py
index ebb46d2f..51670e33 100644
--- a/offline-backend-20260512/backend/packages/harness/deerflow/persistence/agents/model.py
+++ b/offline-backend-20260512/backend/packages/harness/deerflow/persistence/agents/model.py
@@ -32,7 +32,7 @@ class AgentRow(Base):
     # ``is_private``. Empty string = "not assigned" → the listing layer treats
     # it as the admin-configured default square. Each agent belongs to at most
     # one square (single column), so it can never be published to two squares.
-    square_id: Mapped[str] = mapped_column(String(64), nullable=False, default="", server_default="", index=True)
+    square_id: Mapped[str] = mapped_column(String(64), nullable=False, default="", server_default="", index=True)
     # NULL = not featured on the homepage selector. Non-NULL = admin-curated
     # featured rank (lower sorts first). Only ``published=True`` agents are
     # eligible; the router enforces that invariant on writes. BIGINT so the
@@ -57,6 +57,13 @@ class AgentFavoriteRow(Base):
     id: Mapped[str] = mapped_column(String(64), primary_key=True)
     user_id: Mapped[str] = mapped_column(String(64), nullable=False, index=True)
     agent_id: Mapped[str] = mapped_column(String(64), nullable=False, index=True)
+    # Where this favorite came from: ``"user"`` = the user bookmarked it
+    # themselves; ``"position"`` = it was auto-granted by the user's 岗位
+    # (position) via the subscription sync engine. Sync only ever adds/removes
+    # ``"position"`` rows, so a user's own bookmarks are never clobbered.
+    origin: Mapped[str] = mapped_column(String(16), nullable=False, default="user", server_default="user")
+    # The position that granted this favorite (only set when ``origin='position'``).
+    position_id: Mapped[str | None] = mapped_column(String(64), nullable=True, index=True)
     created_at: Mapped[datetime] = mapped_column(BeijingDateTime(), nullable=False, default=lambda: datetime.now(UTC))
 
 
diff --git a/offline-backend-20260512/backend/packages/harness/deerflow/persistence/migrations/versions/20260616_01_positions.py b/offline-backend-20260512/backend/packages/harness/deerflow/persistence/migrations/versions/20260616_01_positions.py
new file mode 100644
index 00000000..11a198f7
--- /dev/null
+++ b/offline-backend-20260512/backend/packages/harness/deerflow/persistence/migrations/versions/20260616_01_positions.py
@@ -0,0 +1,91 @@
+"""岗位 (position) subscription system.
+
+New tables ``positions`` + ``position_tag_bindings``; new column
+``users.position_id``; ``origin`` / ``position_id`` on ``agent_favorites`` and
+``scheduled_task_subscriptions``.
+
+Note: the running app also auto-adds these via ``_ensure_orm_columns_sync`` and
+``create_all`` on startup; this migration keeps the alembic history consistent
+per the repo convention (see backend/CLAUDE.md → Schema Migrations).
+
+Revision ID: 20260616_01
+Revises: 20260612_01
+Create Date: 2026-06-16
+"""
+
+from __future__ import annotations
+
+import sqlalchemy as sa
+from alembic import op
+from sqlalchemy import inspect
+
+
+revision = "20260616_01"
+down_revision = "20260612_01"
+branch_labels = None
+depends_on = None
+
+
+def _has_table(table_name: str) -> bool:
+    return table_name in inspect(op.get_bind()).get_table_names()
+
+
+def _has_column(table_name: str, column_name: str) -> bool:
+    if not _has_table(table_name):
+        return False
+    return any(col["name"] == column_name for col in inspect(op.get_bind()).get_columns(table_name))
+
+
+def upgrade() -> None:
+    if not _has_table("positions"):
+        op.create_table(
+            "positions",
+            sa.Column("id", sa.String(64), primary_key=True),
+            sa.Column("name", sa.String(191), nullable=False),
+            sa.Column("description", sa.String(1024), nullable=False, server_default=""),
+            sa.Column("is_default", sa.Boolean(), nullable=False, server_default=sa.text("0")),
+            sa.Column("created_at", sa.DateTime(timezone=True), nullable=False),
+            sa.Column("updated_at", sa.DateTime(timezone=True), nullable=False),
+            sa.UniqueConstraint("name", name="uq_positions_name"),
+        )
+        op.create_index("ix_positions_is_default", "positions", ["is_default"])
+
+    if not _has_table("position_tag_bindings"):
+        op.create_table(
+            "position_tag_bindings",
+            sa.Column("id", sa.String(64), primary_key=True),
+            sa.Column("position_id", sa.String(64), nullable=False),
+            sa.Column("resource_type", sa.String(32), nullable=False),
+            sa.Column("tag_id", sa.String(64), nullable=False),
+            sa.Column("created_at", sa.DateTime(timezone=True), nullable=False),
+            sa.UniqueConstraint("position_id", "resource_type", "tag_id", name="uq_position_tag_bindings"),
+        )
+        op.create_index("ix_position_tag_bindings_position_id", "position_tag_bindings", ["position_id"])
+        op.create_index("ix_position_tag_bindings_resource_type", "position_tag_bindings", ["resource_type"])
+        op.create_index("ix_position_tag_bindings_tag_id", "position_tag_bindings", ["tag_id"])
+        op.create_index("ix_position_tag_bindings_position_type", "position_tag_bindings", ["position_id", "resource_type"])
+
+    if _has_table("users") and not _has_column("users", "position_id"):
+        op.add_column("users", sa.Column("position_id", sa.String(64), nullable=True))
+        op.create_index("ix_users_position_id", "users", ["position_id"])
+
+    for table in ("agent_favorites", "scheduled_task_subscriptions"):
+        if _has_table(table) and not _has_column(table, "origin"):
+            op.add_column(table, sa.Column("origin", sa.String(16), nullable=False, server_default="user"))
+        if _has_table(table) and not _has_column(table, "position_id"):
+            op.add_column(table, sa.Column("position_id", sa.String(64), nullable=True))
+            op.create_index(f"ix_{table}_position_id", table, ["position_id"])
+
+
+def downgrade() -> None:
+    for table in ("agent_favorites", "scheduled_task_subscriptions"):
+        if _has_column(table, "position_id"):
+            op.drop_column(table, "position_id")
+        if _has_column(table, "origin"):
+            op.drop_column(table, "origin")
+    if _has_column("users", "position_id"):
+        op.drop_column("users", "position_id")
+    if _has_table("position_tag_bindings"):
+        op.drop_table("position_tag_bindings")
+    if _has_table("positions"):
+        op.drop_table("positions")
diff --git a/offline-backend-20260512/backend/packages/harness/deerflow/persistence/migrations/versions/20260616_02_knowledge_folders_templates.py b/offline-backend-20260512/backend/packages/harness/deerflow/persistence/migrations/versions/20260616_02_knowledge_folders_templates.py
new file mode 100644
index 00000000..8f0b9232
--- /dev/null
+++ b/offline-backend-20260512/backend/packages/harness/deerflow/persistence/migrations/versions/20260616_02_knowledge_folders_templates.py
@@ -0,0 +1,84 @@
+"""Knowledge directories + editable extraction templates.
+
+New column ``knowledge_notes.folder``; new tables ``knowledge_folders`` and
+``knowledge_extract_templates``.
+
+Note: the running app also auto-adds the column + tables via
+``_ensure_orm_columns_sync`` / ``create_all`` on startup; this migration keeps
+the alembic history consistent per the repo convention (see backend/CLAUDE.md →
+Schema Migrations).
+
+Revision ID: 20260616_02
+Revises: 20260616_01
+Create Date: 2026-06-16
+"""
+
+from __future__ import annotations
+
+import sqlalchemy as sa
+from alembic import op
+from sqlalchemy import inspect
+from sqlalchemy.dialects import mysql
+
+
+revision = "20260616_02"
+down_revision = "20260616_01"
+branch_labels = None
+depends_on = None
+
+
+def _has_table(table_name: str) -> bool:
+    return table_name in inspect(op.get_bind()).get_table_names()
+
+
+def _has_column(table_name: str, column_name: str) -> bool:
+    if not _has_table(table_name):
+        return False
+    return any(col["name"] == column_name for col in inspect(op.get_bind()).get_columns(table_name))
+
+
+def upgrade() -> None:
+    if _has_table("knowledge_notes") and not _has_column("knowledge_notes", "folder"):
+        op.add_column("knowledge_notes", sa.Column("folder", sa.String(512), nullable=True))
+        op.create_index("ix_knowledge_notes_folder", "knowledge_notes", ["folder"], mysql_length=191)
+
+    if not _has_table("knowledge_folders"):
+        op.create_table(
+            "knowledge_folders",
+            sa.Column("id", sa.String(64), primary_key=True),
+            sa.Column("path", sa.String(512), nullable=False),
+            sa.Column("name", sa.String(256), nullable=False),
+            sa.Column("sort_order", sa.Integer(), nullable=False, server_default="0"),
+            sa.Column("created_by", sa.String(128), nullable=True),
+            sa.Column("created_at", sa.DateTime(timezone=True), nullable=False),
+        )
+        op.create_index("uq_knowledge_folders_path", "knowledge_folders", ["path"], unique=True, mysql_length=191)
+
+    if not _has_table("knowledge_extract_templates"):
+        op.create_table(
+            "knowledge_extract_templates",
+            sa.Column("id", sa.String(64), primary_key=True),
+            sa.Column("name", sa.String(256), nullable=False),
+            sa.Column("description", sa.Text(), nullable=True),
+            sa.Column("scope", sa.String(16), nullable=False, server_default="both"),
+            sa.Column("system_prompt", sa.Text().with_variant(mysql.LONGTEXT(), "mysql"), nullable=False),
+            sa.Column("is_default", sa.Boolean(), nullable=False, server_default=sa.text("0")),
+            sa.Column("enabled", sa.Boolean(), nullable=False, server_default=sa.text("1")),
+            sa.Column("created_by", sa.String(128), nullable=True),
+            sa.Column("updated_by", sa.String(128), nullable=True),
+            sa.Column("created_at", sa.DateTime(timezone=True), nullable=False),
+            sa.Column("updated_at", sa.DateTime(timezone=True), nullable=False),
+        )
+        op.create_index("ix_knowledge_extract_templates_scope", "knowledge_extract_templates", ["scope"])
+        op.create_index("ix_knowledge_extract_templates_is_default", "knowledge_extract_templates", ["is_default"])
+        op.create_index("ix_knowledge_extract_templates_created_at", "knowledge_extract_templates", ["created_at"])
+        op.create_index("uq_knowledge_extract_templates_name", "knowledge_extract_templates", ["name"], unique=True, mysql_length=191)
+
+
+def downgrade() -> None:
+    if _has_table("knowledge_extract_templates"):
+        op.drop_table("knowledge_extract_templates")
+    if _has_table("knowledge_folders"):
+        op.drop_table("knowledge_folders")
+    if _has_column("knowledge_notes", "folder"):
+        op.drop_column("knowledge_notes", "folder")
diff --git a/offline-backend-20260512/backend/packages/harness/deerflow/persistence/migrations/versions/20260616_02_skill_favorites.py b/offline-backend-20260512/backend/packages/harness/deerflow/persistence/migrations/versions/20260616_02_skill_favorites.py
new file mode 100644
index 00000000..0a223a3b
--- /dev/null
+++ b/offline-backend-20260512/backend/packages/harness/deerflow/persistence/migrations/versions/20260616_02_skill_favorites.py
@@ -0,0 +1,54 @@
+"""我的技能 (skill favorites) table.
+
+New table ``skill_favorites`` — the per-user "我的技能" collection, mirroring
+``agent_favorites``. A 岗位 (position) materializes its skill tag bindings into
+this table (``origin='position'``) so granted skills land in the member's 我的
+list while the browse-all 广场 stays complete.
+
+Note: the running app also auto-creates this table via ``create_all`` on
+startup (new tables are always created, even on existing DBs); this migration
+keeps the alembic history consistent per the repo convention (see
+backend/CLAUDE.md → Schema Migrations).
+
+Revision ID: 20260616_02
+Revises: 20260616_01
+Create Date: 2026-06-16
+"""
+
+from __future__ import annotations
+
+import sqlalchemy as sa
+from alembic import op
+from sqlalchemy import inspect
+
+
+revision = "20260616_02"
+down_revision = "20260616_01"
+branch_labels = None
+depends_on = None
+
+
+def _has_table(table_name: str) -> bool:
+    return table_name in inspect(op.get_bind()).get_table_names()
+
+
+def upgrade() -> None:
+    if not _has_table("skill_favorites"):
+        op.create_table(
+            "skill_favorites",
+            sa.Column("id", sa.String(64), primary_key=True),
+            sa.Column("user_id", sa.String(64), nullable=False),
+            sa.Column("skill_name", sa.String(64), nullable=False),
+            sa.Column("origin", sa.String(16), nullable=False, server_default="user"),
+            sa.Column("position_id", sa.String(64), nullable=True),
+            sa.Column("created_at", sa.DateTime(timezone=True), nullable=False),
+            sa.UniqueConstraint("user_id", "skill_name", name="uq_skill_favorites_user_skill"),
+        )
+        op.create_index("ix_skill_favorites_user_id", "skill_favorites", ["user_id"])
+        op.create_index("ix_skill_favorites_skill_name", "skill_favorites", ["skill_name"])
+        op.create_index("ix_skill_favorites_position_id", "skill_favorites", ["position_id"])
+
+
+def downgrade() -> None:
+    if _has_table("skill_favorites"):
+        op.drop_table("skill_favorites")
diff --git a/offline-backend-20260512/backend/packages/harness/deerflow/persistence/migrations/versions/20260616_03_skill_always_on.py b/offline-backend-20260512/backend/packages/harness/deerflow/persistence/migrations/versions/20260616_03_skill_always_on.py
new file mode 100644
index 00000000..68d458bd
--- /dev/null
+++ b/offline-backend-20260512/backend/packages/harness/deerflow/persistence/migrations/versions/20260616_03_skill_always_on.py
@@ -0,0 +1,45 @@
+"""Add always_on (常驻/免压缩) column to skills table.
+
+Revision ID: 20260616_03
+Revises: 20260616_02
+Create Date: 2026-06-16
+"""
+
+from __future__ import annotations
+
+import sqlalchemy as sa
+from alembic import op
+from sqlalchemy import inspect
+
+
+revision = "20260616_03"
+down_revision = "20260616_02"
+branch_labels = None
+depends_on = None
+
+
+def _has_table(table_name: str) -> bool:
+    inspector = inspect(op.get_bind())
+    return table_name in inspector.get_table_names()
+
+
+def _has_column(table_name: str, column_name: str) -> bool:
+    if not _has_table(table_name):
+        return False
+    inspector = inspect(op.get_bind())
+    return any(col["name"] == column_name for col in inspector.get_columns(table_name))
+
+
+def upgrade() -> None:
+    # ``skills`` may not exist yet on a brand-new DB — create_all handles that
+    # case with the column already present, so only patch existing tables.
+    if _has_table("skills") and not _has_column("skills", "always_on"):
+        op.add_column(
+            "skills",
+            sa.Column("always_on", sa.Boolean(), nullable=False, server_default="0"),
+        )
+
+
+def downgrade() -> None:
+    if _has_column("skills", "always_on"):
+        op.drop_column("skills", "always_on")
diff --git a/offline-backend-20260512/backend/packages/harness/deerflow/persistence/models/__init__.py b/offline-backend-20260512/backend/packages/harness/deerflow/persistence/models/__init__.py
index 04aa993c..3870e1f5 100644
--- a/offline-backend-20260512/backend/packages/harness/deerflow/persistence/models/__init__.py
+++ b/offline-backend-20260512/backend/packages/harness/deerflow/persistence/models/__init__.py
@@ -28,6 +28,8 @@ from deerflow.persistence.feedback.model import FeedbackRow
 from deerflow.knowledge.models import (
     KnowledgeEmbeddingRow,
     KnowledgeEntityRow,
+    KnowledgeExtractTemplateRow,
+    KnowledgeFolderRow,
     KnowledgeGraphBlacklistRow,
     KnowledgeNoteRow,
     KnowledgeNoteTagRow,
@@ -41,7 +43,9 @@ from deerflow.persistence.llm_metrics.model import LlmCallMetricRow
 from deerflow.persistence.models.run_event import RunEventRow
 from deerflow.persistence.notifications.model import NotificationRow
 from deerflow.persistence.parallel_agent_tasks.model import ParallelAgentTaskRow
+from deerflow.persistence.positions.model import PositionRow, PositionTagBindingRow
 from deerflow.persistence.recommended_questions.model import RecommendedQuestionRow
+from deerflow.persistence.research_templates.model import ResearchTemplateRow
 from deerflow.persistence.roundtable_chains.model import RoundtableChainRow
 from deerflow.persistence.roundtable_draft_shares.model import RoundtableDraftShareRow
 from deerflow.persistence.roundtable_drafts.model import (
@@ -57,7 +61,7 @@ from deerflow.persistence.scheduled_tasks.model import (
     ScheduledTaskSubscriptionRow,
     SchedulerThreadRow,
 )
-from deerflow.persistence.skills.model import SkillDisplayAdapterRow, SkillRow
+from deerflow.persistence.skills.model import SkillDisplayAdapterRow, SkillFavoriteRow, SkillRow
 from deerflow.persistence.tags.model import TagAssignmentRow, TagRow
 from deerflow.persistence.thread_meta.model import ThreadMetaRow
 from deerflow.persistence.embed_sessions.model import EmbedSessionRow
@@ -77,6 +81,8 @@ __all__ = [
     "BusinessMappingRow",
     "KnowledgeEmbeddingRow",
     "KnowledgeEntityRow",
+    "KnowledgeExtractTemplateRow",
+    "KnowledgeFolderRow",
     "KnowledgeGraphBlacklistRow",
     "KnowledgeNoteRow",
     "KnowledgeNoteTagRow",
@@ -88,7 +94,10 @@ __all__ = [
     "LlmCallMetricRow",
     "NotificationRow",
     "ParallelAgentTaskRow",
+    "PositionRow",
+    "PositionTagBindingRow",
     "RecommendedQuestionRow",
+    "ResearchTemplateRow",
     "RoundtableChainRow",
     "RoundtableDraftRow",
     "RoundtableDraftShareRow",
@@ -103,6 +112,7 @@ __all__ = [
     "SchedulerThreadRow",
     "SkillRow",
     "SkillDisplayAdapterRow",
+    "SkillFavoriteRow",
     "TagAssignmentRow",
     "TagRow",
     "EmbedSessionRow",
diff --git a/offline-backend-20260512/backend/packages/harness/deerflow/persistence/positions/__init__.py b/offline-backend-20260512/backend/packages/harness/deerflow/persistence/positions/__init__.py
new file mode 100644
index 00000000..e1508770
--- /dev/null
+++ b/offline-backend-20260512/backend/packages/harness/deerflow/persistence/positions/__init__.py
@@ -0,0 +1,26 @@
+"""Persistence layer for 岗位 (positions) and per-resource tag bindings."""
+
+from deerflow.persistence.positions.base import PositionStore
+from deerflow.persistence.positions.memory import MemoryPositionStore
+from deerflow.persistence.positions.model import (
+    POSITION_RESOURCE_TYPES,
+    PositionRow,
+    PositionTagBindingRow,
+)
+from deerflow.persistence.positions.sql import PositionRepository
+
+__all__ = [
+    "PositionStore",
+    "MemoryPositionStore",
+    "PositionRepository",
+    "PositionRow",
+    "PositionTagBindingRow",
+    "POSITION_RESOURCE_TYPES",
+    "make_position_store",
+]
+
+
+def make_position_store(session_factory):
+    if session_factory is None:
+        return MemoryPositionStore()
+    return PositionRepository(session_factory)
diff --git a/offline-backend-20260512/backend/packages/harness/deerflow/persistence/positions/base.py b/offline-backend-20260512/backend/packages/harness/deerflow/persistence/positions/base.py
new file mode 100644
index 00000000..ca78f429
--- /dev/null
+++ b/offline-backend-20260512/backend/packages/harness/deerflow/persistence/positions/base.py
@@ -0,0 +1,112 @@
+"""Abstract persistence API for 岗位 (positions) and their tag bindings."""
+
+from __future__ import annotations
+
+from abc import ABC, abstractmethod
+from typing import Any
+
+
+class PositionStore(ABC):
+    # -------------------------------------------------------------- positions
+
+    @abstractmethod
+    async def list_positions(self) -> list[dict[str, Any]]:
+        """List all positions. Each record carries a derived ``member_count``."""
+        raise NotImplementedError
+
+    @abstractmethod
+    async def get_position(self, position_id: str) -> dict[str, Any] | None:
+        raise NotImplementedError
+
+    @abstractmethod
+    async def create_position(self, data: dict[str, Any]) -> dict[str, Any]:
+        """Create a position. ``data`` must carry ``id`` and ``name``.
+
+        Raises ``ValueError`` if ``name`` is not unique. When ``is_default`` is
+        truthy, any other default position is cleared first.
+        """
+        raise NotImplementedError
+
+    @abstractmethod
+    async def update_position(self, position_id: str, data: dict[str, Any]) -> dict[str, Any] | None:
+        """Update ``name`` / ``description`` / ``is_default``.
+
+        Returns ``None`` if the position does not exist. Raises ``ValueError``
+        on a name collision. Setting ``is_default=True`` clears the flag on the
+        other positions.
+        """
+        raise NotImplementedError
+
+    @abstractmethod
+    async def delete_position(self, position_id: str) -> bool:
+        """Delete a position, its tag bindings, and unassign its members."""
+        raise NotImplementedError
+
+    @abstractmethod
+    async def get_default_position_id(self) -> str | None:
+        """Return the id of the ``is_default`` position, or ``None``."""
+        raise NotImplementedError
+
+    # ------------------------------------------------------------ tag bindings
+
+    @abstractmethod
+    async def get_tag_bindings(self, position_id: str) -> dict[str, list[str]]:
+        """Return ``{resource_type: [tag_id, ...]}`` for one position.
+
+        Resource types with no bound tags are omitted (callers treat a missing
+        key as "use the default set").
+        """
+        raise NotImplementedError
+
+    @abstractmethod
+    async def set_tag_bindings(self, position_id: str, bindings: dict[str, list[str]]) -> dict[str, list[str]]:
+        """Replace the full set of tag bindings for a position.
+
+        ``bindings`` maps each resource_type to its desired tag-id list. A
+        resource_type omitted from ``bindings`` (or mapped to ``[]``) ends up
+        with no rows — i.e. it falls back to the default set.
+        """
+        raise NotImplementedError
+
+    # --------------------------------------------------------------- members
+
+    @abstractmethod
+    async def get_user_position_id(self, user_id: str) -> str | None:
+        """Return the position id explicitly assigned to a user, or ``None``."""
+        raise NotImplementedError
+
+    @abstractmethod
+    async def list_members(self, position_id: str) -> list[dict[str, Any]]:
+        """Return ``[{id, email}, ...]`` of users assigned to this position."""
+        raise NotImplementedError
+
+    @abstractmethod
+    async def assign_users(self, position_id: str, user_ids: list[str]) -> int:
+        """Assign a position to many users (1:1 — overwrites any prior position).
+
+        Returns the number of users updated.
+        """
+        raise NotImplementedError
+
+    @abstractmethod
+    async def unassign_users(self, user_ids: list[str]) -> int:
+        """Clear the position of many users. Returns the number updated."""
+        raise NotImplementedError
+
+    @abstractmethod
+    async def list_member_ids(self, position_id: str) -> list[str]:
+        """Return just the user ids assigned to this position (sync fan-out)."""
+        raise NotImplementedError
+
+    @abstractmethod
+    async def list_assigned_user_ids(self) -> list[str]:
+        """Return every user id that has an explicit position (for bulk resync)."""
+        raise NotImplementedError
+
+    @abstractmethod
+    async def list_all_users(self) -> list[dict[str, Any]]:
+        """Return ``[{id, email, system_role, position_id}, ...]`` for every user.
+
+        Powers the admin member-picker and the per-user position editor.
+        """
+        raise NotImplementedError
diff --git a/offline-backend-20260512/backend/packages/harness/deerflow/persistence/positions/memory.py b/offline-backend-20260512/backend/packages/harness/deerflow/persistence/positions/memory.py
new file mode 100644
index 00000000..913a1840
--- /dev/null
+++ b/offline-backend-20260512/backend/packages/harness/deerflow/persistence/positions/memory.py
@@ -0,0 +1,143 @@
+"""In-memory ``PositionStore`` implementation for tests."""
+
+from __future__ import annotations
+
+import copy
+from datetime import UTC, datetime
+from typing import Any
+
+
+def _now() -> str:
+    return datetime.now(UTC).isoformat()
+
+
+class MemoryPositionStore:
+    def __init__(self) -> None:
+        self._positions: dict[str, dict[str, Any]] = {}
+        # position_id -> {resource_type -> [tag_id]}
+        self._bindings: dict[str, dict[str, list[str]]] = {}
+        # user_id -> position_id
+        self._user_position: dict[str, str] = {}
+        # user_id -> email (optional; populated via register_user for nicer member lists)
+        self._user_email: dict[str, str] = {}
+
+    # test helper — not part of the abstract API
+    def register_user(self, user_id: str, email: str = "") -> None:
+        self._user_email[user_id] = email
+
+    def _member_count(self, position_id: str) -> int:
+        return sum(1 for pid in self._user_position.values() if pid == position_id)
+
+    def _to_dict(self, record: dict[str, Any]) -> dict[str, Any]:
+        out = copy.deepcopy(record)
+        out["member_count"] = self._member_count(record["id"])
+        return out
+
+    # -------------------------------------------------------------- positions
+
+    async def list_positions(self) -> list[dict[str, Any]]:
+        rows = sorted(self._positions.values(), key=lambda r: r["created_at"])
+        return [self._to_dict(r) for r in rows]
+
+    async def get_position(self, position_id: str) -> dict[str, Any] | None:
+        record = self._positions.get(position_id)
+        return self._to_dict(record) if record is not None else None
+
+    def _clear_default(self, keep_id: str | None) -> None:
+        for r in self._positions.values():
+            if r["id"] != keep_id:
+                r["is_default"] = False
+
+    async def create_position(self, data: dict[str, Any]) -> dict[str, Any]:
+        name = data["name"]
+        if any(r["name"] == name for r in self._positions.values()):
+            raise ValueError(f"Position '{name}' already exists")
+        is_default = bool(data.get("is_default"))
+        if is_default:
+            self._clear_default(keep_id=data["id"])
+        record = {
+            "id": data["id"],
+            "name": name,
+            "description": data.get("description") or "",
+            "is_default": is_default,
+            "created_at": _now(),
+            "updated_at": _now(),
+        }
+        self._positions[data["id"]] = record
+        return self._to_dict(record)
+
+    async def update_position(self, position_id: str, data: dict[str, Any]) -> dict[str, Any] | None:
+        record = self._positions.get(position_id)
+        if record is None:
+            return None
+        new_name = data.get("name", record["name"])
+        if new_name is not None and any(r["id"] != position_id and r["name"] == new_name for r in self._positions.values()):
+            raise ValueError(f"Position '{new_name}' already exists")
+        if "name" in data and data["name"] is not None:
+            record["name"] = data["name"]
+        if "description" in data and data["description"] is not None:
+            record["description"] = data["description"]
+        if "is_default" in data and data["is_default"] is not None:
+            record["is_default"] = bool(data["is_default"])
+            if record["is_default"]:
+                self._clear_default(keep_id=position_id)
+        record["updated_at"] = _now()
+        return self._to_dict(record)
+
+    async def delete_position(self, position_id: str) -> bool:
+        if position_id not in self._positions:
+            return False
+        del self._positions[position_id]
+        self._bindings.pop(position_id, None)
+        for uid in [uid for uid, pid in self._user_position.items() if pid == position_id]:
+            del self._user_position[uid]
+        return True
+
+    async def get_default_position_id(self) -> str | None:
+        for r in self._positions.values():
+            if r["is_default"]:
+                return r["id"]
+        return None
+
+    # ------------------------------------------------------------ tag bindings
+
+    async def get_tag_bindings(self, position_id: str) -> dict[str, list[str]]:
+        return copy.deepcopy(self._bindings.get(position_id, {}))
+
+    async def set_tag_bindings(self, position_id: str, bindings: dict[str, list[str]]) -> dict[str, list[str]]:
+        cleaned = {rt: list(dict.fromkeys(tag_ids)) for rt, tag_ids in bindings.items() if tag_ids}
+        self._bindings[position_id] = cleaned
+        return copy.deepcopy(cleaned)
+
+    # --------------------------------------------------------------- members
+
+    async def get_user_position_id(self, user_id: str) -> str | None:
+        return self._user_position.get(user_id)
+
+    async def list_members(self, position_id: str) -> list[dict[str, Any]]:
+        return [{"id": uid, "email": self._user_email.get(uid, "")} for uid, pid in self._user_position.items() if pid == position_id]
+
+    async def assign_users(self, position_id: str, user_ids: list[str]) -> int:
+        for uid in user_ids:
+            self._user_position[uid] = position_id
+        return len(user_ids)
+
+    async def unassign_users(self, user_ids: list[str]) -> int:
+        count = 0
+        for uid in user_ids:
+            if uid in self._user_position:
+                del self._user_position[uid]
+                count += 1
+        return count
+
+    async def list_member_ids(self, position_id: str) -> list[str]:
+        return [uid for uid, pid in self._user_position.items() if pid == position_id]
+
+    async def list_assigned_user_ids(self) -> list[str]:
+        return list(self._user_position.keys())
+
+    async def list_all_users(self) -> list[dict[str, Any]]:
+        return [
+            {"id": uid, "email": email, "system_role": "user", "position_id": self._user_position.get(uid)}
+            for uid, email in sorted(self._user_email.items(), key=lambda kv: kv[1])
+        ]
diff --git a/offline-backend-20260512/backend/packages/harness/deerflow/persistence/positions/model.py b/offline-backend-20260512/backend/packages/harness/deerflow/persistence/positions/model.py
new file mode 100644
index 00000000..eb475f21
--- /dev/null
+++ b/offline-backend-20260512/backend/packages/harness/deerflow/persistence/positions/model.py
@@ -0,0 +1,73 @@
+"""ORM models for 岗位 (positions) and their per-resource tag bindings.
+
+A **position** (岗位) is an admin-managed aggregation layer that decides which
+agents / skills / scheduled tasks / recommended questions a user is subscribed
+to. Each user has at most one position (the link lives on
+``users.position_id``, a 1:1 relation — see ``UserRow``).
+
+Two tables here:
+
+- ``positions`` — the position itself. ``is_default`` marks the fallback
+  position inherited by users who have not been assigned one explicitly.
+- ``position_tag_bindings`` — for each ``(position, resource_type)`` the set of
+  tags whose resources are granted to the position's members. **No binding for
+  a resource_type means "use the default set"** (see ``deerflow.services.
+  position_sync`` for what "default" resolves to per type).
+"""
+
+from __future__ import annotations
+
+from datetime import UTC, datetime
+
+from sqlalchemy import Boolean, Index, String, UniqueConstraint
+from sqlalchemy.orm import Mapped, mapped_column
+
+from deerflow.persistence.base import Base
+from deerflow.persistence.types import BeijingDateTime
+
+# Resource types a position can grant. The first three reuse the existing
+# typed-tag namespace (``deerflow.persistence.tags``); ``recommended_question``
+# was added alongside (see TAG_TYPES). The removed ``user_prompt`` type is no
+# longer accepted — ``get_tag_bindings`` drops any stale rows of that type.
+POSITION_RESOURCE_TYPES: tuple[str, ...] = (
+    "agent",
+    "skill",
+    "scheduled_task",
+    "recommended_question",
+)
+
+
+class PositionRow(Base):
+    __tablename__ = "positions"
+    __table_args__ = (UniqueConstraint("name", name="uq_positions_name"),)
+
+    id: Mapped[str] = mapped_column(String(64), primary_key=True)
+    name: Mapped[str] = mapped_column(String(191), nullable=False)
+    description: Mapped[str] = mapped_column(String(1024), nullable=False, default="")
+    # Exactly one position should carry ``is_default=True`` at a time (the
+    # router enforces "only one default" by clearing the flag on others). Users
+    # with no explicit ``users.position_id`` inherit this position's grants.
+    is_default: Mapped[bool] = mapped_column(Boolean, nullable=False, default=False, index=True)
+    created_at: Mapped[datetime] = mapped_column(BeijingDateTime(), nullable=False, default=lambda: datetime.now(UTC))
+    updated_at: Mapped[datetime] = mapped_column(BeijingDateTime(), nullable=False, default=lambda: datetime.now(UTC), onupdate=lambda: datetime.now(UTC))
+
+
+class PositionTagBindingRow(Base):
+    """One bound tag for a ``(position, resource_type)`` pair.
+
+    The absence of *any* row for a ``(position, resource_type)`` means the
+    position uses the **default set** for that resource type. One or more rows
+    mean the position grants the *union* of the resources carrying those tags.
+    """
+
+    __tablename__ = "position_tag_bindings"
+    __table_args__ = (
+        UniqueConstraint("position_id", "resource_type", "tag_id", name="uq_position_tag_bindings"),
+        Index("ix_position_tag_bindings_position_type", "position_id", "resource_type"),
+    )
+
+    id: Mapped[str] = mapped_column(String(64), primary_key=True)
+    position_id: Mapped[str] = mapped_column(String(64), nullable=False, index=True)
+    resource_type: Mapped[str] = mapped_column(String(32), nullable=False, index=True)
+    tag_id: Mapped[str] = mapped_column(String(64), nullable=False, index=True)
+    created_at: Mapped[datetime] = mapped_column(BeijingDateTime(), nullable=False, default=lambda: datetime.now(UTC))
diff --git a/offline-backend-20260512/backend/packages/harness/deerflow/persistence/positions/sql.py b/offline-backend-20260512/backend/packages/harness/deerflow/persistence/positions/sql.py
new file mode 100644
index 00000000..4b23ac8b
--- /dev/null
+++ b/offline-backend-20260512/backend/packages/harness/deerflow/persistence/positions/sql.py
@@ -0,0 +1,201 @@
+"""SQLAlchemy-backed position + tag-binding repository."""
+
+from __future__ import annotations
+
+import uuid
+from datetime import UTC, datetime
+from typing import Any
+
+from sqlalchemy import delete, func, select, update
+from sqlalchemy.exc import IntegrityError
+from sqlalchemy.ext.asyncio import AsyncSession, async_sessionmaker
+
+from deerflow.persistence.positions.base import PositionStore
+from deerflow.persistence.positions.model import POSITION_RESOURCE_TYPES, PositionRow, PositionTagBindingRow
+from deerflow.persistence.user.model import UserRow
+
+
+def _position_to_dict(row: PositionRow, member_count: int = 0) -> dict[str, Any]:
+    data = row.to_dict()
+    for key in ("created_at", "updated_at"):
+        value = data.get(key)
+        if isinstance(value, datetime):
+            data[key] = value.isoformat()
+    data["member_count"] = member_count
+    return data
+
+
+class PositionRepository(PositionStore):
+    def __init__(self, session_factory: async_sessionmaker[AsyncSession]) -> None:
+        self._sf = session_factory
+
+    @staticmethod
+    async def _member_counts(session: AsyncSession, position_ids: list[str]) -> dict[str, int]:
+        if not position_ids:
+            return {}
+        stmt = select(UserRow.position_id, func.count()).where(UserRow.position_id.in_(position_ids)).group_by(UserRow.position_id)
+        result = await session.execute(stmt)
+        return {pid: count for pid, count in result.all() if pid is not None}
+
+    @staticmethod
+    async def _clear_default(session: AsyncSession, keep_id: str | None) -> None:
+        stmt = update(PositionRow).where(PositionRow.is_default.is_(True))
+        if keep_id is not None:
+            stmt = stmt.where(PositionRow.id != keep_id)
+        await session.execute(stmt.values(is_default=False))
+
+    # -------------------------------------------------------------- positions
+
+    async def list_positions(self) -> list[dict[str, Any]]:
+        async with self._sf() as session:
+            rows = list((await session.execute(select(PositionRow).order_by(PositionRow.created_at.asc()))).scalars())
+            counts = await self._member_counts(session, [r.id for r in rows])
+            return [_position_to_dict(r, counts.get(r.id, 0)) for r in rows]
+
+    async def get_position(self, position_id: str) -> dict[str, Any] | None:
+        async with self._sf() as session:
+            row = await session.get(PositionRow, position_id)
+            if row is None:
+                return None
+            counts = await self._member_counts(session, [position_id])
+            return _position_to_dict(row, counts.get(position_id, 0))
+
+    async def create_position(self, data: dict[str, Any]) -> dict[str, Any]:
+        now = datetime.now(UTC)
+        is_default = bool(data.get("is_default"))
+        row = PositionRow(
+            id=data["id"],
+            name=data["name"],
+            description=data.get("description") or "",
+            is_default=is_default,
+            created_at=now,
+            updated_at=now,
+        )
+        async with self._sf() as session:
+            if is_default:
+                await self._clear_default(session, keep_id=row.id)
+            session.add(row)
+            try:
+                await session.commit()
+            except IntegrityError as exc:
+                await session.rollback()
+                raise ValueError(f"Position '{data['name']}' already exists") from exc
+            await session.refresh(row)
+            return _position_to_dict(row, 0)
+
+    async def update_position(self, position_id: str, data: dict[str, Any]) -> dict[str, Any] | None:
+        async with self._sf() as session:
+            row = await session.get(PositionRow, position_id)
+            if row is None:
+                return None
+            if "name" in data and data["name"] is not None:
+                row.name = data["name"]
+            if "description" in data and data["description"] is not None:
+                row.description = data["description"]
+            if "is_default" in data and data["is_default"] is not None:
+                row.is_default = bool(data["is_default"])
+                if row.is_default:
+                    await self._clear_default(session, keep_id=position_id)
+            row.updated_at = datetime.now(UTC)
+            try:
+                await session.commit()
+            except IntegrityError as exc:
+                await session.rollback()
+                raise ValueError(f"Position '{data.get('name')}' already exists") from exc
+            await session.refresh(row)
+            counts = await self._member_counts(session, [position_id])
+            return _position_to_dict(row, counts.get(position_id, 0))
+
+    async def delete_position(self, position_id: str) -> bool:
+        async with self._sf() as session:
+            row = await session.get(PositionRow, position_id)
+            if row is None:
+                return False
+            await session.execute(delete(PositionTagBindingRow).where(PositionTagBindingRow.position_id == position_id))
+            await session.execute(update(UserRow).where(UserRow.position_id == position_id).values(position_id=None))
+            await session.delete(row)
+            await session.commit()
+            return True
+
+    async def get_default_position_id(self) -> str | None:
+        async with self._sf() as session:
+            return (await session.execute(select(PositionRow.id).where(PositionRow.is_default.is_(True)).limit(1))).scalar_one_or_none()
+
+    # ------------------------------------------------------------ tag bindings
+
+    async def get_tag_bindings(self, position_id: str) -> dict[str, list[str]]:
+        async with self._sf() as session:
+            rows = list(
+                (
+                    await session.execute(
+                        select(PositionTagBindingRow).where(PositionTagBindingRow.position_id == position_id).order_by(PositionTagBindingRow.created_at.asc())
+                    )
+                ).scalars()
+            )
+            out: dict[str, list[str]] = {}
+            for r in rows:
+                # Drop bindings for resource types no longer supported (e.g. the
+                # removed ``user_prompt`` type) so stale rows can't poison the
+                # editor or be echoed back into a 422 on save.
+                if r.resource_type not in POSITION_RESOURCE_TYPES:
+                    continue
+                out.setdefault(r.resource_type, []).append(r.tag_id)
+            return out
+
+    async def set_tag_bindings(self, position_id: str, bindings: dict[str, list[str]]) -> dict[str, list[str]]:
+        now = datetime.now(UTC)
+        async with self._sf() as session:
+            await session.execute(delete(PositionTagBindingRow).where(PositionTagBindingRow.position_id == position_id))
+            for resource_type, tag_ids in bindings.items():
+                for tag_id in dict.fromkeys(tag_ids):
+                    session.add(
+                        PositionTagBindingRow(
+                            id=uuid.uuid4().hex,
+                            position_id=position_id,
+                            resource_type=resource_type,
+                            tag_id=tag_id,
+                            created_at=now,
+                        )
+                    )
+            await session.commit()
+            return await self.get_tag_bindings(position_id)
+
+    # --------------------------------------------------------------- members
+
+    async def get_user_position_id(self, user_id: str) -> str | None:
+        async with self._sf() as session:
+            return (await session.execute(select(UserRow.position_id).where(UserRow.id == user_id))).scalar_one_or_none()
+
+    async def list_members(self, position_id: str) -> list[dict[str, Any]]:
+        async with self._sf() as session:
+            rows = (await session.execute(select(UserRow.id, UserRow.email).where(UserRow.position_id == position_id).order_by(UserRow.email.asc()))).all()
+            return [{"id": uid, "email": email} for uid, email in rows]
+
+    async def assign_users(self, position_id: str, user_ids: list[str]) -> int:
+        if not user_ids:
+            return 0
+        async with self._sf() as session:
+            result = await session.execute(update(UserRow).where(UserRow.id.in_(user_ids)).values(position_id=position_id))
+            await session.commit()
+            return result.rowcount or 0
+
+    async def unassign_users(self, user_ids: list[str]) -> int:
+        if not user_ids:
+            return 0
+        async with self._sf() as session:
+            result = await session.execute(update(UserRow).where(UserRow.id.in_(user_ids)).values(position_id=None))
+            await session.commit()
+            return result.rowcount or 0
+
+    async def list_member_ids(self, position_id: str) -> list[str]:
+        async with self._sf() as session:
+            return [uid for (uid,) in (await session.execute(select(UserRow.id).where(UserRow.position_id == position_id))).all()]
+
+    async def list_assigned_user_ids(self) -> list[str]:
+        async with self._sf() as session:
+            return [uid for (uid,) in (await session.execute(select(UserRow.id).where(UserRow.position_id.isnot(None)))).all()]
+
+    async def list_all_users(self) -> list[dict[str, Any]]:
+        async with self._sf() as session:
+            rows = (await session.execute(select(UserRow.id, UserRow.email, UserRow.system_role, UserRow.position_id).order_by(UserRow.email.asc()))).all()
+            return [{"id": uid, "email": email, "system_role": role, "position_id": pid} for uid, email, role, pid in rows]
diff --git a/offline-backend-20260512/backend/packages/harness/deerflow/persistence/research_templates/__init__.py b/offline-backend-20260512/backend/packages/harness/deerflow/persistence/research_templates/__init__.py
new file mode 100644
index 00000000..7d236327
--- /dev/null
+++ b/offline-backend-20260512/backend/packages/harness/deerflow/persistence/research_templates/__init__.py
@@ -0,0 +1,15 @@
+from deerflow.persistence.research_templates.model import ResearchTemplateRow
+from deerflow.persistence.research_templates.sql import ResearchTemplateRepository
+
+
+def make_research_template_store(session_factory) -> ResearchTemplateRepository | None:
+    if session_factory is None:
+        return None
+    return ResearchTemplateRepository(session_factory)
+
+
+__all__ = [
+    "ResearchTemplateRow",
+    "ResearchTemplateRepository",
+    "make_research_template_store",
+]
diff --git a/offline-backend-20260512/backend/packages/harness/deerflow/persistence/research_templates/model.py b/offline-backend-20260512/backend/packages/harness/deerflow/persistence/research_templates/model.py
new file mode 100644
index 00000000..00e5896c
--- /dev/null
+++ b/offline-backend-20260512/backend/packages/harness/deerflow/persistence/research_templates/model.py
@@ -0,0 +1,35 @@
+"""ORM model for AI writing research templates (课题研究模板).
+
+A research template bundles a set of writing-form presets — article type,
+outline, material source, word count, audience, strict mode and a topic/intent —
+so a user can one-click pre-fill the whole 课题研究 form. Maintained from the
+research-template config page (需求 0616 #5).
+"""
+
+from __future__ import annotations
+
+from datetime import UTC, datetime
+
+from sqlalchemy import Boolean, DateTime, Integer, String, Text
+from sqlalchemy.orm import Mapped, mapped_column
+
+from deerflow.persistence.base import Base
+
+
+class ResearchTemplateRow(Base):
+    __tablename__ = "ai_writing_research_templates"
+
+    id: Mapped[str] = mapped_column(String(64), primary_key=True)
+    name: Mapped[str] = mapped_column(String(128), nullable=False)
+    description: Mapped[str | None] = mapped_column(Text, nullable=True)
+    # 预设字段（点击模板时回填表单）。
+    article_type_key: Mapped[str | None] = mapped_column(String(64), nullable=True)
+    outline: Mapped[str | None] = mapped_column(Text, nullable=True)
+    material_source: Mapped[str] = mapped_column(String(32), nullable=False, default="general")
+    word_count: Mapped[int] = mapped_column(Integer, nullable=False, default=800)
+    audience: Mapped[str | None] = mapped_column(String(128), nullable=True)
+    strict_mode: Mapped[bool] = mapped_column(Boolean, nullable=False, default=False)
+    topic: Mapped[str | None] = mapped_column(Text, nullable=True)
+    sort_order: Mapped[int] = mapped_column(Integer, nullable=False, default=0)
+    created_at: Mapped[datetime] = mapped_column(DateTime(timezone=True), default=lambda: datetime.now(UTC))
+    updated_at: Mapped[datetime] = mapped_column(DateTime(timezone=True), default=lambda: datetime.now(UTC))
diff --git a/offline-backend-20260512/backend/packages/harness/deerflow/persistence/research_templates/sql.py b/offline-backend-20260512/backend/packages/harness/deerflow/persistence/research_templates/sql.py
new file mode 100644
index 00000000..1d73b8a9
--- /dev/null
+++ b/offline-backend-20260512/backend/packages/harness/deerflow/persistence/research_templates/sql.py
@@ -0,0 +1,221 @@
+"""SQLAlchemy-backed research template storage."""
+
+from __future__ import annotations
+
+import uuid
+from datetime import UTC, datetime
+
+from sqlalchemy import select
+from sqlalchemy.ext.asyncio import AsyncSession, async_sessionmaker
+
+from deerflow.persistence.research_templates.model import ResearchTemplateRow
+
+_UNSET = object()
+
+
+# 新机器内置范例模板（表为空时由 seed_defaults() 写入；作为给用户的范例，内容务求规范好用）。
+# article_type_key 取自默认文章类型（news/analysis/feature/report，见 20260517_02 迁移）。
+DEFAULT_RESEARCH_TEMPLATES: list[dict] = [
+    {
+        "name": "网摘",
+        "description": "汇编近期网络热点信息，分条摘录核心要点并标注来源，快速掌握全貌",
+        "article_type_key": "news",
+        "material_source": "general",
+        "word_count": 1000,
+        "audience": "关注热点动态的一般读者",
+        "strict_mode": False,
+        "topic": (
+            "围绕【主题】检索并汇编近一周网络上的重要信息与热点动态，"
+            "逐条摘录核心事实、关键数据与原文出处，帮助读者快速掌握全貌。"
+        ),
+        "outline": (
+            "一、热点速览（用一句话清单列出本期要点）\n"
+            "二、重点摘录\n"
+            "　1.【事件/话题标题】——核心事实、关键数据、信息来源\n"
+            "　2.【事件/话题标题】——……\n"
+            "　3.（按重要性依次罗列）\n"
+            "三、关联动态与背景补充（串联相关信息、补足来龙去脉）\n"
+            "四、小结与观察（提炼趋势或值得关注的看点）"
+        ),
+        "sort_order": 0,
+    },
+    {
+        "name": "周报",
+        "description": "按周梳理工作进展、数据指标与问题风险，输出结构化周度汇报",
+        "article_type_key": "report",
+        "material_source": "general",
+        "word_count": 1200,
+        "audience": "团队成员与上级主管",
+        "strict_mode": True,
+        "topic": (
+            "汇总本周【项目/团队】的工作进展，撰写周报："
+            "覆盖关键成果、数据指标、问题风险与下周计划，简明、客观、可量化。"
+        ),
+        "outline": (
+            "一、本周概览（一句话总结 + 关键指标对比上周）\n"
+            "二、重点进展\n"
+            "　1. 已完成事项与成果（结论先行，附支撑数据）\n"
+            "　2. 关键数据 / 里程碑达成情况\n"
+            "三、问题与风险（逐条列出问题、影响面及应对措施）\n"
+            "四、下周计划与资源需求（目标、负责人、所需支持）"
+        ),
+        "sort_order": 1,
+    },
+    {
+        "name": "要讯",
+        "description": "提炼最重要的资讯，精炼成简报式要点，突出核心事实与潜在影响",
+        "article_type_key": "analysis",
+        "material_source": "general",
+        "word_count": 600,
+        "audience": "需快速掌握要点的决策者",
+        "strict_mode": True,
+        "topic": (
+            "围绕【主题/领域】提炼近期最重要的资讯，编写要讯："
+            "每条聚焦核心事实、关键背景与潜在影响，文字精炼、判断审慎。"
+        ),
+        "outline": (
+            "一、头条要讯（1–3 条，每条用一句话点明核心）\n"
+            "二、要点详述（逐条展开）\n"
+            "　· 事实：发生了什么、何时何地、涉及谁\n"
+            "　· 背景：为何重要、与既有态势的关系\n"
+            "　· 影响：可能带来的后果与连锁反应\n"
+            "三、研判与建议（给出倾向性判断与可行动建议）"
+        ),
+        "sort_order": 2,
+    },
+]
+
+
+class ResearchTemplateRepository:
+    def __init__(self, session_factory: async_sessionmaker[AsyncSession]) -> None:
+        self._sf = session_factory
+
+    @staticmethod
+    def _row_to_dict(row: ResearchTemplateRow) -> dict:
+        d = row.to_dict()
+        for k in ("created_at", "updated_at"):
+            if isinstance(d.get(k), datetime):
+                d[k] = d[k].isoformat()
+        return d
+
+    async def create(
+        self,
+        *,
+        name: str,
+        description: str | None = None,
+        article_type_key: str | None = None,
+        outline: str | None = None,
+        material_source: str = "general",
+        word_count: int = 800,
+        audience: str | None = None,
+        strict_mode: bool = False,
+        topic: str | None = None,
+        sort_order: int = 0,
+    ) -> dict:
+        now = datetime.now(UTC)
+        row = ResearchTemplateRow(
+            id=str(uuid.uuid4()),
+            name=name,
+            description=description,
+            article_type_key=article_type_key,
+            outline=outline,
+            material_source=material_source,
+            word_count=word_count,
+            audience=audience,
+            strict_mode=strict_mode,
+            topic=topic,
+            sort_order=sort_order,
+            created_at=now,
+            updated_at=now,
+        )
+        async with self._sf() as session:
+            session.add(row)
+            await session.commit()
+            await session.refresh(row)
+            return self._row_to_dict(row)
+
+    async def update(
+        self,
+        template_id: str,
+        *,
+        name=_UNSET,
+        description=_UNSET,
+        article_type_key=_UNSET,
+        outline=_UNSET,
+        material_source=_UNSET,
+        word_count=_UNSET,
+        audience=_UNSET,
+        strict_mode=_UNSET,
+        topic=_UNSET,
+        sort_order=_UNSET,
+    ) -> dict | None:
+        async with self._sf() as session:
+            row = await session.get(ResearchTemplateRow, template_id)
+            if row is None:
+                return None
+            if name is not _UNSET:
+                row.name = name
+            if description is not _UNSET:
+                row.description = description
+            if article_type_key is not _UNSET:
+                row.article_type_key = article_type_key
+            if outline is not _UNSET:
+                row.outline = outline
+            if material_source is not _UNSET:
+                row.material_source = material_source
+            if word_count is not _UNSET:
+                row.word_count = word_count
+            if audience is not _UNSET:
+                row.audience = audience
+            if strict_mode is not _UNSET:
+                row.strict_mode = strict_mode
+            if topic is not _UNSET:
+                row.topic = topic
+            if sort_order is not _UNSET:
+                row.sort_order = sort_order
+            row.updated_at = datetime.now(UTC)
+            await session.commit()
+            await session.refresh(row)
+            return self._row_to_dict(row)
+
+    async def get(self, template_id: str) -> dict | None:
+        async with self._sf() as session:
+            row = await session.get(ResearchTemplateRow, template_id)
+            if row is None:
+                return None
+            return self._row_to_dict(row)
+
+    async def list(self) -> list[dict]:
+        stmt = select(ResearchTemplateRow).order_by(
+            ResearchTemplateRow.sort_order.asc(), ResearchTemplateRow.created_at.asc()
+        )
+        async with self._sf() as session:
+            result = await session.execute(stmt)
+            return [self._row_to_dict(r) for r in result.scalars()]
+
+    async def delete(self, template_id: str) -> bool:
+        async with self._sf() as session:
+            row = await session.get(ResearchTemplateRow, template_id)
+            if row is None:
+                return False
+            await session.delete(row)
+            await session.commit()
+            return True
+
+    async def seed_defaults(self) -> int:
+        """新机器/空表时写入内置范例模板（网摘/周报/要讯）。
+
+        幂等：仅当表为空时插入；已有任何模板则跳过（避免覆盖用户配置或重复播种）。
+        返回插入条数。
+        """
+        async with self._sf() as session:
+            existing = await session.execute(select(ResearchTemplateRow.id).limit(1))
+            if existing.first() is not None:
+                return 0
+            now = datetime.now(UTC)
+            for t in DEFAULT_RESEARCH_TEMPLATES:
+                session.add(
+                    ResearchTemplateRow(id=str(uuid.uuid4()), created_at=now, updated_at=now, **t)
+                )
+            await session.commit()
+            return len(DEFAULT_RESEARCH_TEMPLATES)
diff --git a/offline-backend-20260512/backend/packages/harness/deerflow/persistence/scheduled_tasks/model.py b/offline-backend-20260512/backend/packages/harness/deerflow/persistence/scheduled_tasks/model.py
index d32d7843..e17a1c90 100644
--- a/offline-backend-20260512/backend/packages/harness/deerflow/persistence/scheduled_tasks/model.py
+++ b/offline-backend-20260512/backend/packages/harness/deerflow/persistence/scheduled_tasks/model.py
@@ -55,6 +55,11 @@ class ScheduledTaskSubscriptionRow(Base):
     task_id: Mapped[str] = mapped_column(String(64), index=True)
     user_id: Mapped[str] = mapped_column(String(64), index=True)
     notify_element: Mapped[bool] = mapped_column(Boolean, default=False)
+    # Source of this subscription: ``"user"`` (subscribed manually from the
+    # 广场) or ``"position"`` (auto-granted by the user's 岗位 via the sync
+    # engine). Sync only ever touches ``"position"`` rows.
+    origin: Mapped[str] = mapped_column(String(16), nullable=False, default="user", server_default="user")
+    position_id: Mapped[str | None] = mapped_column(String(64), nullable=True, index=True)
     created_at: Mapped[datetime] = mapped_column(BeijingDateTime(), default=lambda: datetime.now(UTC))
     updated_at: Mapped[datetime] = mapped_column(BeijingDateTime(), default=lambda: datetime.now(UTC), onupdate=lambda: datetime.now(UTC))
 
diff --git a/offline-backend-20260512/backend/packages/harness/deerflow/persistence/skills/base.py b/offline-backend-20260512/backend/packages/harness/deerflow/persistence/skills/base.py
index f84ea8a5..69390001 100644
--- a/offline-backend-20260512/backend/packages/harness/deerflow/persistence/skills/base.py
+++ b/offline-backend-20260512/backend/packages/harness/deerflow/persistence/skills/base.py
@@ -85,3 +85,23 @@ class SkillStore(ABC):
         DB round-trip per candidate skill.
         """
         raise NotImplementedError
+
+    @abstractmethod
+    async def list_always_on_names(self) -> set[str]:
+        """Return the set of skill names marked 常驻(免压缩) (``always_on = True``).
+
+        Batch query used by the prompt builder to decide which skills bypass
+        skill compression — one query instead of a per-skill round-trip.
+        """
+        raise NotImplementedError
+
+    @abstractmethod
+    async def list_favorite_skill_names(self, user_id: str) -> set[str]:
+        """Return the skill names in ``user_id``'s "我的技能" collection.
+
+        These are skills the user does not own but that landed in their list —
+        currently only via their 岗位 (position) sync (``origin='position'``).
+        Owned skills are NOT included here (the caller detects those via
+        ``owner_user_id``).
+        """
+        raise NotImplementedError
diff --git a/offline-backend-20260512/backend/packages/harness/deerflow/persistence/skills/memory.py b/offline-backend-20260512/backend/packages/harness/deerflow/persistence/skills/memory.py
index 04f29409..e5f1adb6 100644
--- a/offline-backend-20260512/backend/packages/harness/deerflow/persistence/skills/memory.py
+++ b/offline-backend-20260512/backend/packages/harness/deerflow/persistence/skills/memory.py
@@ -20,12 +20,15 @@ def _copy(record: dict[str, Any] | None) -> dict[str, Any] | None:
 def _with_derived(record: dict[str, Any]) -> dict[str, Any]:
     out = dict(record)
     out["builtin"] = out.get("owner_user_id") is None
+    out.setdefault("always_on", False)
     return out
 
 
 class MemorySkillStore(SkillStore):
     def __init__(self) -> None:
         self._skills: dict[str, dict[str, Any]] = {}
+        # user_id -> {skill_name}
+        self._favorites: dict[str, set[str]] = {}
 
     @staticmethod
     def _is_visible(record: dict[str, Any], user_id: str) -> bool:
@@ -71,6 +74,7 @@ class MemorySkillStore(SkillStore):
             "published": bool(data.get("published", False)),
             "detail": data.get("detail"),
             "name_zh": data.get("name_zh"),
+            "always_on": bool(data.get("always_on", False)),
             "square_id": str(data.get("square_id") or ""),
             "created_at": _now(),
             "updated_at": _now(),
@@ -107,6 +111,8 @@ class MemorySkillStore(SkillStore):
             record["detail"] = data["detail"]
         if "name_zh" in data:
             record["name_zh"] = data["name_zh"]
+        if "always_on" in data:
+            record["always_on"] = bool(data["always_on"])
         if "square_id" in data:
             record["square_id"] = str(data["square_id"] or "")
         record["updated_at"] = _now()
@@ -124,6 +130,8 @@ class MemorySkillStore(SkillStore):
             record["detail"] = data["detail"]
         if "name_zh" in data:
             record["name_zh"] = data["name_zh"]
+        if "always_on" in data:
+            record["always_on"] = bool(data["always_on"])
         if "square_id" in data:
             record["square_id"] = str(data["square_id"] or "")
         record["updated_at"] = _now()
@@ -153,3 +161,13 @@ class MemorySkillStore(SkillStore):
 
     async def list_published_names(self) -> set[str]:
         return {name for name, record in self._skills.items() if record.get("published")}
+
+    async def list_always_on_names(self) -> set[str]:
+        return {name for name, record in self._skills.items() if record.get("always_on")}
+
+    async def list_favorite_skill_names(self, user_id: str) -> set[str]:
+        return set(self._favorites.get(user_id, set()))
+
+    # test helper — not part of the abstract API
+    def add_favorite(self, user_id: str, skill_name: str) -> None:
+        self._favorites.setdefault(user_id, set()).add(skill_name)
diff --git a/offline-backend-20260512/backend/packages/harness/deerflow/persistence/skills/model.py b/offline-backend-20260512/backend/packages/harness/deerflow/persistence/skills/model.py
index d631f112..50827767 100644
--- a/offline-backend-20260512/backend/packages/harness/deerflow/persistence/skills/model.py
+++ b/offline-backend-20260512/backend/packages/harness/deerflow/persistence/skills/model.py
@@ -10,10 +10,11 @@ from __future__ import annotations
 
 from datetime import UTC, datetime
 
-from sqlalchemy import Boolean, DateTime, String, Text
+from sqlalchemy import Boolean, DateTime, String, Text, UniqueConstraint
 from sqlalchemy.orm import Mapped, mapped_column
 
 from deerflow.persistence.base import Base
+from deerflow.persistence.types import BeijingDateTime
 
 
 class SkillRow(Base):
@@ -33,6 +34,10 @@ class SkillRow(Base):
     # the on-disk SKILL.md) so it can be set for built-in skills too, and so the
     # shared source file is never rewritten.
     name_zh: Mapped[str | None] = mapped_column(String(128), nullable=True, default=None)
+    # 常驻(免压缩):开启技能压缩后,标记此项的技能仍以全文形式注入系统提示词,而非
+    # 降级为索引 + search_skills 检索。存 DB(而非 extensions_config.json),内置技能
+    # 也能标记,且避免在某些部署里回写源码目录下的配置文件导致报错。
+    always_on: Mapped[bool] = mapped_column(Boolean, nullable=False, default=False, server_default="0", index=True)
     # Publish square (发布广场) this skill is filed under once published. Pure
     # organizational category, no access control. Empty string = unassigned →
     # the listing layer resolves it to the default square. One square per skill.
@@ -41,6 +46,32 @@ class SkillRow(Base):
     updated_at: Mapped[datetime] = mapped_column(DateTime(timezone=True), nullable=False, default=lambda: datetime.now(UTC), onupdate=lambda: datetime.now(UTC))
 
 
+class SkillFavoriteRow(Base):
+    """A user's "我的技能" membership for a skill they don't own.
+
+    Mirrors :class:`deerflow.persistence.agents.model.AgentFavoriteRow`: lets a
+    skill land in a user's "我的" list without the user owning it. Skills are
+    keyed by ``name`` (the on-disk folder / DB primary key), so this row
+    references ``skill_name`` rather than an opaque id. One row per
+    (user, skill); the unique constraint makes the grant idempotent.
+    """
+
+    __tablename__ = "skill_favorites"
+    __table_args__ = (UniqueConstraint("user_id", "skill_name", name="uq_skill_favorites_user_skill"),)
+
+    id: Mapped[str] = mapped_column(String(64), primary_key=True)
+    user_id: Mapped[str] = mapped_column(String(64), nullable=False, index=True)
+    skill_name: Mapped[str] = mapped_column(String(64), nullable=False, index=True)
+    # ``"user"`` = the user added it themselves (reserved for a future favorite
+    # button); ``"position"`` = auto-granted by the user's 岗位 (position) via
+    # the subscription sync engine. Sync only ever adds/removes ``"position"``
+    # rows, so a user's own picks are never clobbered.
+    origin: Mapped[str] = mapped_column(String(16), nullable=False, default="user", server_default="user")
+    # The position that granted this favorite (only set when ``origin='position'``).
+    position_id: Mapped[str | None] = mapped_column(String(64), nullable=True, index=True)
+    created_at: Mapped[datetime] = mapped_column(BeijingDateTime(), nullable=False, default=lambda: datetime.now(UTC))
+
+
 class SkillDisplayAdapterRow(Base):
     __tablename__ = "skill_display_adapters"
 
diff --git a/offline-backend-20260512/backend/packages/harness/deerflow/persistence/skills/sql.py b/offline-backend-20260512/backend/packages/harness/deerflow/persistence/skills/sql.py
index 30c1d90c..78004c07 100644
--- a/offline-backend-20260512/backend/packages/harness/deerflow/persistence/skills/sql.py
+++ b/offline-backend-20260512/backend/packages/harness/deerflow/persistence/skills/sql.py
@@ -9,7 +9,7 @@ from sqlalchemy import func, or_, select
 from sqlalchemy.ext.asyncio import AsyncSession, async_sessionmaker
 
 from deerflow.persistence.skills.base import SkillStore
-from deerflow.persistence.skills.model import SkillRow
+from deerflow.persistence.skills.model import SkillFavoriteRow, SkillRow
 
 
 def _row_to_dict(row: SkillRow) -> dict[str, Any]:
@@ -65,6 +65,7 @@ class SkillRepository(SkillStore):
             published=bool(data.get("published", False)),
             detail=data.get("detail"),
             name_zh=data.get("name_zh"),
+            always_on=bool(data.get("always_on", False)),
             square_id=str(data.get("square_id") or ""),
             created_at=now,
             updated_at=now,
@@ -110,6 +111,8 @@ class SkillRepository(SkillStore):
                 row.detail = data["detail"]
             if "name_zh" in data:
                 row.name_zh = data["name_zh"]
+            if "always_on" in data:
+                row.always_on = bool(data["always_on"])
             if "square_id" in data:
                 row.square_id = str(data["square_id"] or "")
             row.updated_at = datetime.now(UTC)
@@ -130,6 +133,8 @@ class SkillRepository(SkillStore):
                 row.detail = data["detail"]
             if "name_zh" in data:
                 row.name_zh = data["name_zh"]
+            if "always_on" in data:
+                row.always_on = bool(data["always_on"])
             if "square_id" in data:
                 row.square_id = str(data["square_id"] or "")
             row.updated_at = datetime.now(UTC)
@@ -171,3 +176,15 @@ class SkillRepository(SkillStore):
         async with self._sf() as session:
             result = await session.execute(stmt)
             return {row[0] for row in result.all()}
+
+    async def list_always_on_names(self) -> set[str]:
+        stmt = select(SkillRow.name).where(SkillRow.always_on.is_(True))
+        async with self._sf() as session:
+            result = await session.execute(stmt)
+            return {row[0] for row in result.all()}
+
+    async def list_favorite_skill_names(self, user_id: str) -> set[str]:
+        stmt = select(SkillFavoriteRow.skill_name).where(SkillFavoriteRow.user_id == user_id)
+        async with self._sf() as session:
+            result = await session.execute(stmt)
+            return {row[0] for row in result.all()}
diff --git a/offline-backend-20260512/backend/packages/harness/deerflow/persistence/tags/base.py b/offline-backend-20260512/backend/packages/harness/deerflow/persistence/tags/base.py
index e9b3a961..8ebfb296 100644
--- a/offline-backend-20260512/backend/packages/harness/deerflow/persistence/tags/base.py
+++ b/offline-backend-20260512/backend/packages/harness/deerflow/persistence/tags/base.py
@@ -71,3 +71,14 @@ class TagStore(ABC):
     async def unassign_all(self, target_type: str, target_id: str) -> int:
         """Drop every tag assignment for a resource. Returns the count removed."""
         raise NotImplementedError
+
+    @abstractmethod
+    async def list_targets_for_tags(self, target_type: str, tag_ids: list[str]) -> list[str]:
+        """Return the ids of ``target_type`` resources carrying ANY of ``tag_ids`` (OR).
+
+        Reverse of :meth:`list_assignments`. Powers the 岗位 (position) sync
+        engine and the per-position visibility filters: "which resources does
+        this set of tags grant?". Returns a de-duplicated list; empty when
+        ``tag_ids`` is empty.
+        """
+        raise NotImplementedError
diff --git a/offline-backend-20260512/backend/packages/harness/deerflow/persistence/tags/memory.py b/offline-backend-20260512/backend/packages/harness/deerflow/persistence/tags/memory.py
index 592618e1..4ff3a02e 100644
--- a/offline-backend-20260512/backend/packages/harness/deerflow/persistence/tags/memory.py
+++ b/offline-backend-20260512/backend/packages/harness/deerflow/persistence/tags/memory.py
@@ -136,3 +136,15 @@ class MemoryTagStore(TagStore):
         for aid in victims:
             del self._assignments[aid]
         return len(victims)
+
+    async def list_targets_for_tags(self, target_type: str, tag_ids: list[str]) -> list[str]:
+        wanted = set(tag_ids)
+        if not wanted:
+            return []
+        out: list[str] = []
+        seen: set[str] = set()
+        for a in self._assignments.values():
+            if a["target_type"] == target_type and a["tag_id"] in wanted and a["target_id"] not in seen:
+                seen.add(a["target_id"])
+                out.append(a["target_id"])
+        return out
diff --git a/offline-backend-20260512/backend/packages/harness/deerflow/persistence/tags/model.py b/offline-backend-20260512/backend/packages/harness/deerflow/persistence/tags/model.py
index d83bba99..3deb6f63 100644
--- a/offline-backend-20260512/backend/packages/harness/deerflow/persistence/tags/model.py
+++ b/offline-backend-20260512/backend/packages/harness/deerflow/persistence/tags/model.py
@@ -21,7 +21,9 @@ from deerflow.persistence.base import Base
 from deerflow.persistence.types import BeijingDateTime
 
 # Allowed tag/target types. Kept here so persistence + API share one source.
-TAG_TYPES: tuple[str, ...] = ("agent", "skill", "scheduled_task")
+# ``recommended_question`` was added for the 岗位 (position) subscription system
+# so admin-configured recommended questions can be tag-targeted like agents.
+TAG_TYPES: tuple[str, ...] = ("agent", "skill", "scheduled_task", "recommended_question")
 
 
 class TagRow(Base):
diff --git a/offline-backend-20260512/backend/packages/harness/deerflow/persistence/tags/sql.py b/offline-backend-20260512/backend/packages/harness/deerflow/persistence/tags/sql.py
index e7ff73d9..8652cc8f 100644
--- a/offline-backend-20260512/backend/packages/harness/deerflow/persistence/tags/sql.py
+++ b/offline-backend-20260512/backend/packages/harness/deerflow/persistence/tags/sql.py
@@ -185,3 +185,15 @@ class TagRepository(TagStore):
             )
             await session.commit()
             return result.rowcount or 0
+
+    async def list_targets_for_tags(self, target_type: str, tag_ids: list[str]) -> list[str]:
+        unique_ids = list(dict.fromkeys(tag_ids))
+        if not unique_ids:
+            return []
+        stmt = (
+            select(TagAssignmentRow.target_id)
+            .where(TagAssignmentRow.target_type == target_type, TagAssignmentRow.tag_id.in_(unique_ids))
+            .distinct()
+        )
+        async with self._sf() as session:
+            return [tid for (tid,) in (await session.execute(stmt)).all()]
diff --git a/offline-backend-20260512/backend/packages/harness/deerflow/persistence/user/model.py b/offline-backend-20260512/backend/packages/harness/deerflow/persistence/user/model.py
index c079bdfd..2c065b65 100644
--- a/offline-backend-20260512/backend/packages/harness/deerflow/persistence/user/model.py
+++ b/offline-backend-20260512/backend/packages/harness/deerflow/persistence/user/model.py
@@ -49,6 +49,11 @@ class UserRow(Base):
     needs_setup: Mapped[bool] = mapped_column(Boolean, nullable=False, default=False)
     token_version: Mapped[int] = mapped_column(nullable=False, default=0)
 
+    # 岗位 (position) the user is assigned to (1:1). NULL = no explicit
+    # position → the user inherits the ``is_default`` position's grants. Drives
+    # the subscription sync engine (see ``deerflow.services.position_sync``).
+    position_id: Mapped[str | None] = mapped_column(String(64), nullable=True, index=True)
+
     __table_args__ = (
         Index(
             "idx_users_oauth_identity",
diff --git a/offline-backend-20260512/backend/packages/harness/deerflow/services/__init__.py b/offline-backend-20260512/backend/packages/harness/deerflow/services/__init__.py
new file mode 100644
index 00000000..67324884
--- /dev/null
+++ b/offline-backend-20260512/backend/packages/harness/deerflow/services/__init__.py
@@ -0,0 +1 @@
+"""Cross-cutting harness services (no ``app.*`` imports)."""
diff --git a/offline-backend-20260512/backend/packages/harness/deerflow/services/position_sync.py b/offline-backend-20260512/backend/packages/harness/deerflow/services/position_sync.py
new file mode 100644
index 00000000..bc3ee1e1
--- /dev/null
+++ b/offline-backend-20260512/backend/packages/harness/deerflow/services/position_sync.py
@@ -0,0 +1,183 @@
+"""岗位 (position) → resource subscription **sync engine**.
+
+A user's position decides which agents / skills / scheduled tasks they are
+subscribed to, and which recommended questions they can see. This service turns
+a position's *tag bindings* into concrete, materialized rows in the existing
+"我的" tables (idempotent, re-runnable).
+
+Resolution rule (per resource type):
+
+* The user's position is ``users.position_id``; if unset, the ``is_default``
+  position is used; if there is none, every type falls back to its default set.
+* For a resource type **with** bound tags → the union of resources carrying any
+  of those tags.
+* For a resource type **without** bound tags → the *default set*:
+  - ``agent``               → all built-in agents (``user_id IS NULL``)
+  - ``scheduled_task``      → empty (no built-in tasks)
+  - ``skill``               → empty (the 广场 still shows every visible skill;
+    a position only *adds* its tagged skills to the member's "我的技能")
+  - ``recommended_question`` → *unrestricted* (handled at query time, never
+    materialized — see :meth:`resolve_visible`)
+
+Materialization (``sync_user``) only ever touches rows it created itself
+(``origin='position'``), so a user's own favorites / subscriptions are never
+clobbered. ``skill`` is materialized into ``skill_favorites`` exactly like
+``agent`` → ``agent_favorites``: the tagged skills are *added* to the user's
+"我的技能" while the browse-all 广场 stays complete (it is **not** restricted).
+Only recommended-question visibility is computed live by :meth:`resolve_visible`
+and is *not* materialized. (The admin-configured prompts shown to users ARE the
+recommended questions — there is no separate prompt-template store.)
+"""
+
+from __future__ import annotations
+
+import logging
+import uuid
+from datetime import UTC, datetime
+
+from sqlalchemy import select
+from sqlalchemy.ext.asyncio import AsyncSession, async_sessionmaker
+
+from deerflow.persistence.agents.model import AgentFavoriteRow, AgentRow
+from deerflow.persistence.positions.base import PositionStore
+from deerflow.persistence.scheduled_tasks.model import ScheduledTaskSubscriptionRow
+from deerflow.persistence.skills.model import SkillFavoriteRow
+from deerflow.persistence.tags.base import TagStore
+
+logger = logging.getLogger(__name__)
+
+# Materialized into "我的" tables by sync_user.
+_MATERIALIZED_TYPES = ("agent", "skill", "scheduled_task")
+# Resolved live at query time (visibility filters), never materialized.
+_VISIBILITY_TYPES = ("recommended_question",)
+
+
+class PositionSyncService:
+    def __init__(self, session_factory: async_sessionmaker[AsyncSession], position_store: PositionStore, tag_store: TagStore) -> None:
+        self._sf = session_factory
+        self._positions = position_store
+        self._tags = tag_store
+
+    # ----------------------------------------------------------- resolution
+
+    async def _effective_position_id(self, user_id: str) -> str | None:
+        pid = await self._positions.get_user_position_id(user_id)
+        if pid:
+            return pid
+        return await self._positions.get_default_position_id()
+
+    async def resolve_visible(self, user_id: str, resource_type: str) -> set[str] | None:
+        """Return the set of visible ``resource_type`` ids, or ``None`` = unrestricted.
+
+        Used by the recommended-questions filter (the only remaining query-time
+        visibility type; skills are now *materialized*, not filtered). ``None``
+        means "no position restriction → show everything"; a set means "restrict
+        to exactly these ids".
+        """
+        pid = await self._effective_position_id(user_id)
+        if not pid:
+            return None
+        bindings = await self._positions.get_tag_bindings(pid)
+        tag_ids = bindings.get(resource_type) or []
+        if not tag_ids:
+            return None
+        return set(await self._tags.list_targets_for_tags(resource_type, tag_ids))
+
+    async def _resolve_materialized(self, user_id: str, resource_type: str, session: AsyncSession) -> set[str]:
+        pid = await self._effective_position_id(user_id)
+        bindings = await self._positions.get_tag_bindings(pid) if pid else {}
+        tag_ids = bindings.get(resource_type) or []
+        if tag_ids:
+            return set(await self._tags.list_targets_for_tags(resource_type, tag_ids))
+        return await self._default_set(resource_type, session)
+
+    async def _default_set(self, resource_type: str, session: AsyncSession) -> set[str]:
+        if resource_type == "agent":
+            rows = (await session.execute(select(AgentRow.id).where(AgentRow.user_id.is_(None)))).all()
+            return {r[0] for r in rows}
+        # scheduled_task (and any other materialized type) defaults to empty.
+        return set()
+
+    # --------------------------------------------------------- materialize
+
+    async def sync_user(self, user_id: str) -> None:
+        """Reconcile all materialized subscriptions for one user (idempotent)."""
+        if self._sf is None:
+            return
+        pid = await self._effective_position_id(user_id)
+        async with self._sf() as session:
+            await self._sync_agents(session, user_id, pid)
+            await self._sync_skills(session, user_id, pid)
+            await self._sync_tasks(session, user_id, pid)
+            await session.commit()
+
+    async def _sync_agents(self, session: AsyncSession, user_id: str, position_id: str | None) -> None:
+        target = await self._resolve_materialized(user_id, "agent", session)
+        rows = list((await session.execute(select(AgentFavoriteRow).where(AgentFavoriteRow.user_id == user_id))).scalars())
+        owned_any = {r.agent_id for r in rows}
+        position_rows = {r.agent_id: r for r in rows if r.origin == "position"}
+        now = datetime.now(UTC)
+        for agent_id in target - owned_any:
+            session.add(AgentFavoriteRow(id=uuid.uuid4().hex, user_id=user_id, agent_id=agent_id, origin="position", position_id=position_id, created_at=now))
+        for agent_id, row in position_rows.items():
+            if agent_id not in target:
+                await session.delete(row)
+
+    async def _sync_skills(self, session: AsyncSession, user_id: str, position_id: str | None) -> None:
+        target = await self._resolve_materialized(user_id, "skill", session)
+        rows = list((await session.execute(select(SkillFavoriteRow).where(SkillFavoriteRow.user_id == user_id))).scalars())
+        owned_any = {r.skill_name for r in rows}
+        position_rows = {r.skill_name: r for r in rows if r.origin == "position"}
+        now = datetime.now(UTC)
+        for skill_name in target - owned_any:
+            session.add(SkillFavoriteRow(id=uuid.uuid4().hex, user_id=user_id, skill_name=skill_name, origin="position", position_id=position_id, created_at=now))
+        for skill_name, row in position_rows.items():
+            if skill_name not in target:
+                await session.delete(row)
+
+    async def _sync_tasks(self, session: AsyncSession, user_id: str, position_id: str | None) -> None:
+        target = await self._resolve_materialized(user_id, "scheduled_task", session)
+        rows = list((await session.execute(select(ScheduledTaskSubscriptionRow).where(ScheduledTaskSubscriptionRow.user_id == user_id))).scalars())
+        owned_any = {r.task_id for r in rows}
+        position_rows = {r.task_id: r for r in rows if r.origin == "position"}
+        now = datetime.now(UTC)
+        for task_id in target - owned_any:
+            session.add(ScheduledTaskSubscriptionRow(id=uuid.uuid4().hex, task_id=task_id, user_id=user_id, notify_element=False, origin="position", position_id=position_id, created_at=now, updated_at=now))
+        for task_id, row in position_rows.items():
+            if task_id not in target:
+                await session.delete(row)
+
+    # ------------------------------------------------------------- fan-out
+
+    async def sync_position(self, position_id: str) -> int:
+        """Re-sync every member of a position. Returns the member count."""
+        member_ids = await self._positions.list_member_ids(position_id)
+        for uid in member_ids:
+            try:
+                await self.sync_user(uid)
+            except Exception:
+                logger.exception("position sync failed for user %s (position %s)", uid, position_id)
+        return len(member_ids)
+
+    async def sync_default_members(self) -> int:
+        """Re-sync users with NO explicit position (they inherit the default)."""
+        default_id = await self._positions.get_default_position_id()
+        if not default_id:
+            return 0
+        assigned = set(await self._positions.list_assigned_user_ids())
+        count = 0
+        if self._sf is None:
+            return 0
+        from deerflow.persistence.user.model import UserRow
+
+        async with self._sf() as session:
+            all_ids = [uid for (uid,) in (await session.execute(select(UserRow.id))).all()]
+        for uid in all_ids:
+            if uid in assigned:
+                continue
+            try:
+                await self.sync_user(uid)
+                count += 1
+            except Exception:
+                logger.exception("default position sync failed for user %s", uid)
+        return count
diff --git a/offline-backend-20260512/backend/packages/harness/deerflow/skills/storage/local_skill_storage.py b/offline-backend-20260512/backend/packages/harness/deerflow/skills/storage/local_skill_storage.py
index 0491c7c8..23aa8ead 100644
--- a/offline-backend-20260512/backend/packages/harness/deerflow/skills/storage/local_skill_storage.py
+++ b/offline-backend-20260512/backend/packages/harness/deerflow/skills/storage/local_skill_storage.py
@@ -54,11 +54,42 @@ class LocalSkillStorage(SkillStorage):
         return self._host_root
 
     def custom_skill_exists(self, name: str) -> bool:
-        return self.get_custom_skill_file(name).exists()
+        return self._resolve_skill_dir(name, SkillCategory.CUSTOM) is not None
 
     def public_skill_exists(self, name: str) -> bool:
+        return self._resolve_skill_dir(name, SkillCategory.PUBLIC) is not None
+
+    def _resolve_skill_dir(self, name: str, category: SkillCategory) -> Path | None:
+        """Return the on-disk directory of the skill identified by *name*.
+
+        Discovery (``_iter_skill_files``/``load_skills``) keys a skill by the
+        ``name`` in its SKILL.md frontmatter, but a hand-dropped or mis-extracted
+        skill can live at a directory whose name differs from that frontmatter
+        name. Locating a skill by the canonical ``<category>/<name>`` path alone
+        would then make it *listable but not deletable/readable*.
+
+        Fast path: the canonical ``<category>/<name>`` layout. Fallback: scan the
+        discovered SKILL.md files and match on the parsed frontmatter ``name`` so
+        delete/read/editability stay consistent with what discovery surfaces.
+        Returns ``None`` when no such skill is discoverable.
+        """
+        from deerflow.skills.parser import parse_skill_file
+
         normalized_name = self.validate_skill_name(name)
-        return (self._host_root / SkillCategory.PUBLIC.value / normalized_name / SKILL_MD_FILE).exists()
+        canonical = self._host_root / category.value / normalized_name
+        if (canonical / SKILL_MD_FILE).is_file():
+            return canonical
+        for found_category, category_root, md_path in self._iter_skill_files():
+            if found_category != category:
+                continue
+            skill = parse_skill_file(
+                md_path,
+                category=found_category,
+                relative_path=md_path.parent.relative_to(category_root),
+            )
+            if skill and skill.name == normalized_name:
+                return md_path.parent
+        return None
 
     def _iter_skill_files(self) -> Iterable[tuple[SkillCategory, Path, Path]]:
         if not self._host_root.exists():
@@ -71,12 +102,19 @@ class LocalSkillStorage(SkillStorage):
                 dir_names[:] = sorted(name for name in dir_names if not name.startswith("."))
                 if SKILL_MD_FILE not in file_names:
                     continue
+                # A directory containing SKILL.md *is* a self-contained skill;
+                # treat it as a leaf and do not descend into its subdirectories.
+                # This keeps bundled assets that ship their own SKILL.md (e.g.
+                # ``examples/`` or ``references/`` inside a skill package) from
+                # being surfaced as separate, undeletable top-level skills.
+                dir_names[:] = []
                 yield category, category_path, Path(current_root) / SKILL_MD_FILE
 
     def read_custom_skill(self, name: str) -> str:
-        if not self.custom_skill_exists(name):
+        skill_dir = self._resolve_skill_dir(name, SkillCategory.CUSTOM)
+        if skill_dir is None:
             raise FileNotFoundError(f"Custom skill '{name}' not found.")
-        return (self.get_custom_skill_dir(name) / SKILL_MD_FILE).read_text(encoding="utf-8")
+        return (skill_dir / SKILL_MD_FILE).read_text(encoding="utf-8")
 
     def write_custom_skill(self, name: str, relative_path: str, content: str) -> None:
         target = self.validate_relative_path(relative_path, self.get_custom_skill_dir(name))
@@ -251,7 +289,17 @@ class LocalSkillStorage(SkillStorage):
     def delete_custom_skill(self, name: str, *, history_meta: dict | None = None) -> None:
         self.validate_skill_name(name)
         self.ensure_custom_skill_is_editable(name)
-        target = self.get_custom_skill_dir(name)
+        # Delete the skill's *discovered* directory, not the assumed
+        # ``custom/<name>`` path — a skill placed at a non-canonical or nested
+        # location must still be removable now that discovery surfaced it.
+        target = self._resolve_skill_dir(name, SkillCategory.CUSTOM)
+        if target is None:
+            raise FileNotFoundError(f"Custom skill '{name}' not found.")
+        # Safety: never escape the custom root, and never remove the root itself.
+        custom_root = (self._host_root / SkillCategory.CUSTOM.value).resolve()
+        resolved_target = target.resolve()
+        if resolved_target == custom_root or custom_root not in resolved_target.parents:
+            raise ValueError(f"Refusing to delete skill '{name}': resolved path '{resolved_target}' is outside the custom skills directory.")
         if history_meta is not None:
             prev_content = self.read_custom_skill(name)
             try:
diff --git a/offline-backend-20260512/backend/packages/harness/deerflow/skills/types.py b/offline-backend-20260512/backend/packages/harness/deerflow/skills/types.py
index a3f3c154..0e5b9baf 100644
--- a/offline-backend-20260512/backend/packages/harness/deerflow/skills/types.py
+++ b/offline-backend-20260512/backend/packages/harness/deerflow/skills/types.py
@@ -28,6 +28,7 @@ class Skill:
     relative_path: Path  # Relative path from category root to skill directory
     category: SkillCategory  # 'public' or 'custom'
     enabled: bool = False  # Whether this skill is enabled
+    always_on: bool = False  # 常驻(免压缩):开启压缩后仍把本技能全文注入系统提示词
 
     @property
     def skill_path(self) -> str:
diff --git a/offline-backend-20260512/backend/packages/harness/deerflow/tools/builtins/skill_tools.py b/offline-backend-20260512/backend/packages/harness/deerflow/tools/builtins/skill_tools.py
index 50fae12a..c4ebb28b 100644
--- a/offline-backend-20260512/backend/packages/harness/deerflow/tools/builtins/skill_tools.py
+++ b/offline-backend-20260512/backend/packages/harness/deerflow/tools/builtins/skill_tools.py
@@ -176,6 +176,113 @@ def skill_view_tool(name: str, runtime: ToolRuntime) -> str:
     return skill_view_impl(name, _resolve_active_agent_id(runtime))
 
 
+def _resolve_skills_container_base_path() -> str:
+    """容器内技能挂载根路径(用于回传 location,供 agent read_file)。"""
+    try:
+        from deerflow.config import get_app_config
+
+        return get_app_config().skills.container_path
+    except Exception:
+        return "/mnt/skills"
+
+
+def _score_skill(query_terms: list[str], haystack: str) -> int:
+    """对单个技能按关键词命中计分(纯关键词、离线、确定性)。"""
+    score = 0
+    for term in query_terms:
+        if term and term in haystack:
+            score += 1
+    return score
+
+
+def search_skills_impl(query: str, agent_id: str | None, top_n: int = 5) -> str:
+    """按关键词检索可用技能,返回 top-N 的 name/description/category/location。
+
+    与 ``skill_list`` 共享可见性过滤(自定义 agent 走白名单;默认 lead agent 只
+    暴露未归档且内置或属于当前用户的技能),并只检索**已启用**的技能 —— 它是
+    "技能压缩"开启后,agent 调出被压缩(非常驻)技能的主要入口。匹配字段为技能
+    名与描述。
+    """
+    import re
+
+    from deerflow.skills.storage import get_or_new_skill_storage
+    from deerflow.skills.types import SkillCategory
+    from deerflow.skills.usage import STATE_ARCHIVED, all_entries
+
+    query = (query or "").strip()
+    if not query:
+        return "search_skills: empty query. Provide a short keyword query describing the task."
+
+    terms = [t for t in re.findall(r"\w+", query.lower()) if t]
+    # 整串也作为一个匹配项,兼顾中文等不被 \w+ 切分出有意义 token 的查询。
+    whole = query.lower()
+    if whole and whole not in terms:
+        terms.append(whole)
+
+    allowed = _load_agent_skill_whitelist(agent_id)
+    owned: set[str] | None = None
+    if allowed is None:
+        owned = _load_owned_skill_names(_resolve_effective_user_id())
+
+    try:
+        top_n = max(1, min(int(top_n or 5), 50))
+    except (TypeError, ValueError):
+        top_n = 5
+
+    try:
+        storage = get_or_new_skill_storage()
+        skills = storage.load_skills(enabled_only=True)
+        usage = all_entries()
+        base_path = _resolve_skills_container_base_path()
+        scored: list[tuple[int, dict]] = []
+        for skill in skills:
+            entry = usage.get(skill.name, {})
+            if allowed is not None:
+                if skill.name not in allowed:
+                    continue
+            else:
+                if entry.get("state") == STATE_ARCHIVED:
+                    continue
+                if skill.category == SkillCategory.CUSTOM and owned is not None and skill.name not in owned:
+                    continue
+            haystack = f"{skill.name} {skill.description}".lower()
+            score = _score_skill(terms, haystack)
+            if score <= 0:
+                continue
+            scored.append((
+                score,
+                {
+                    "name": skill.name,
+                    "description": skill.description,
+                    "category": skill.category.value if hasattr(skill.category, "value") else str(skill.category),
+                    "location": skill.get_container_file_path(base_path),
+                },
+            ))
+        scored.sort(key=lambda x: x[0], reverse=True)
+        results = [item for _, item in scored[:top_n]]
+        if not results:
+            return f"No skills matched '{query}'. Try a broader keyword, or call skill_list to browse everything available."
+        return json.dumps(results, ensure_ascii=False, indent=2)
+    except Exception as e:
+        logger.debug("search_skills failed: %s", e)
+        return f"Error searching skills: {e}"
+
+
+@tool("search_skills")
+def search_skills_tool(query: str, runtime: ToolRuntime) -> str:
+    """Search available skills by keyword and return the best matches.
+
+    Use this to discover skills that are not already described in full in your
+    system prompt. Returns a JSON array of matches, each with `name`,
+    `description`, `category`, and `location` — then `read_file` the `location`
+    of the most relevant match and follow its SKILL.md instructions.
+
+    Args:
+        query: A short keyword query describing the task (e.g. 'pdf table extract').
+    """
+    return search_skills_impl(query, _resolve_active_agent_id(runtime))
+
+
 @tool("skill_list")
 def skill_list_tool(runtime: ToolRuntime) -> str:
     """List skills available to this agent, with metadata and usage stats.
diff --git a/offline-backend-20260512/backend/packages/harness/deerflow/tools/tools.py b/offline-backend-20260512/backend/packages/harness/deerflow/tools/tools.py
index 2bf592ee..fa104905 100644
--- a/offline-backend-20260512/backend/packages/harness/deerflow/tools/tools.py
+++ b/offline-backend-20260512/backend/packages/harness/deerflow/tools/tools.py
@@ -139,6 +139,18 @@ def get_available_tools(
 
     builtin_tools.extend([skill_list_tool, skill_view_tool])
 
+    # search_skills —— 仅当管理员开启"技能压缩"时提供。压缩开启后,非常驻技能不再
+    # 全文注入系统提示词,agent 通过该关键词检索工具按需调出(纯关键词、离线、确定性)。
+    try:
+        from deerflow.config.system_settings import get_skill_compression_settings
+
+        if get_skill_compression_settings().enabled:
+            from deerflow.tools.builtins.skill_tools import search_skills_tool
+
+            builtin_tools.append(search_skills_tool)
+    except Exception:  # noqa: BLE001 - 读取设置失败不能阻断 agent 构建
+        logger.debug("Failed to evaluate skill compression setting for search_skills tool", exc_info=True)
+
     # read_peer_delivery —— 圆桌席位「按需读取前序席位完整交付」的只读工具(从 run context
     # 取网关注入的全文,非圆桌场景为空 → 工具自述「无」)。始终提供(只读、无副作用);圆桌
     # 总控/报告角色由 run policy 显式排除,避免干扰它们。
diff --git a/offline-backend-20260512/backend/scripts/cleanup_knowledge_related.py b/offline-backend-20260512/backend/scripts/cleanup_knowledge_related.py
new file mode 100644
index 00000000..928ace25
--- /dev/null
+++ b/offline-backend-20260512/backend/scripts/cleanup_knowledge_related.py
@@ -0,0 +1,182 @@
+"""一次性清洗:修正历史知识笔记里遗留的 `## Related` 结构链接与 `### Sources`。
+
+背景
+====
+早期沉淀的每条笔记都会在 `## Related` 段自动塞入“结构性” wikilink
+(`[[index]]` / `[[_meta/taxonomy]]` / `[[projects/<project>/…]]` / 项目 hub),
+并在 `## Related` 下嵌套一个 `### Sources` 子段。现在:
+
+- 真·相关链接只用裸标题 `[[Title]]`(由 service 层按共享标签注入);
+- 来源统一由结构化的「来源」区块(`note.sources`)展示,不再写进正文。
+
+但历史笔记的 `content_md`(数据库 + vault 文件)里仍焊死着这些遗留内容,
+导致详情页出现噪声链接、以及 sources/来源 重复。本脚本把它们清理掉。
+
+清洗规则(保守,只删明确匹配项)
+================================
+对每条笔记的正文:
+1. 删除 `## Related` 段内的 `### Sources` 子段(标题 + 其下条目);
+2. 删除 `## Related` 段内的“结构性”链接条目 —— 即 wikilink 目标包含 `/`、
+   或等于 `index` / `taxonomy` 之类的脚手架页(真·相关链接是裸标题,予以保留);
+3. 若清理后 `## Related` 段已无任何内容条目,则连同 `## Related` 标题一并删除;
+4. 折叠多余空行。
+
+裸标题的 `[[Title]]` 相关链接、其它正文内容一律保留(无数据丢失)。
+
+用法
+====
+::
+
+    # 预览(不写盘),打印每条将要发生的改动
+    PYTHONPATH=. python scripts/cleanup_knowledge_related.py --dry-run
+
+    # 执行(写回数据库 + 重写 vault 文件)
+    PYTHONPATH=. python scripts/cleanup_knowledge_related.py
+
+    # 只处理单条
+    PYTHONPATH=. python scripts/cleanup_knowledge_related.py --note-id kb_xxx
+
+脚本幂等:已经干净的笔记会被跳过。写回走 `KnowledgeService.update_note`,
+所以每次改动都会落一份版本快照(可在「编辑历史」回滚)。
+"""
+
+from __future__ import annotations
+
+import argparse
+import asyncio
+import logging
+import re
+import sys
+
+logger = logging.getLogger("cleanup_knowledge_related")
+
+_RELATED_HEADER_RE = re.compile(r"^##\s+(Related|相关知识|相关文档|相关)\s*$", re.IGNORECASE)
+_SOURCES_HEADER_RE = re.compile(r"^###\s+(Sources|来源)\s*$", re.IGNORECASE)
+_H12_RE = re.compile(r"^#{1,2}\s+")
+_H123_RE = re.compile(r"^#{1,3}\s+")
+_BULLET_LINK_RE = re.compile(r"^[-*]\s*\[\[([^\]\n|]+)(?:\|[^\]\n]+)?\]\]\s*$")
+
+
+def _is_structural_target(target: str) -> bool:
+    """A `## Related` wikilink that points at vault scaffolding (not a real note).
+
+    Real related cross-links are bare note titles (`[[标题]]`) with no slashes.
+    The legacy structural links were full vault paths / meta pages.
+    """
+    t = (target or "").strip().strip("/").lower()
+    if not t:
+        return True
+    if "/" in t:  # projects/<p>/…, _meta/taxonomy, … — all structural
+        return True
+    return t in {"index", "taxonomy", "hot", "log", "_meta"}
+
+
+def clean_note_body(md: str) -> str:
+    """Strip the legacy `### Sources` subsection + structural `## Related` links.
+
+    Pure function (no I/O) so it can be unit-tested. Returns the cleaned body;
+    identical to the input when there is nothing to clean.
+    """
+    lines = md.split("\n")
+    n = len(lines)
+    out: list[str] = []
+    i = 0
+    in_fence = False
+    while i < n:
+        line = lines[i]
+        if re.match(r"^\s*(```|~~~)", line):
+            in_fence = not in_fence
+            out.append(line)
+            i += 1
+            continue
+        if in_fence or not _RELATED_HEADER_RE.match(line.strip()):
+            out.append(line)
+            i += 1
+            continue
+
+        # --- inside a `## Related` section -------------------------------
+        header = line
+        j = i + 1
+        kept: list[str] = []  # surviving content lines of this section
+        while j < n:
+            cur = lines[j]
+            stripped = cur.strip()
+            if _H12_RE.match(cur):  # next h1/h2 ends the Related section
+                break
+            if _SOURCES_HEADER_RE.match(stripped):
+                # Drop the `### Sources` header and everything until the next header.
+                j += 1
+                while j < n and not _H123_RE.match(lines[j]):
+                    j += 1
+                continue
+            m = _BULLET_LINK_RE.match(stripped)
+            if m and _is_structural_target(m.group(1)):
+                j += 1  # drop the structural link bullet
+                continue
+            kept.append(cur)
+            j += 1
+
+        meaningful = [ln for ln in kept if ln.strip()]
+        if meaningful:
+            out.append(header)
+            out.extend(kept)
+        # else: drop the now-empty `## Related` section entirely.
+        i = j
+
+    cleaned = "\n".join(out)
+    cleaned = re.sub(r"\n{3,}", "\n\n", cleaned).rstrip() + "\n"
+    return cleaned
+
+
+async def _run(dry_run: bool, only_note_id: str | None) -> int:
+    from deerflow.config import get_app_config
+    from deerflow.knowledge import make_knowledge_service
+    from deerflow.persistence.engine import close_engine, get_session_factory, init_engine_from_config
+
+    config = get_app_config()
+    await init_engine_from_config(config.database)
+    sf = get_session_factory()
+    if sf is None:
+        logger.error("No database session factory — is the database backend configured?")
+        return 1
+    service = make_knowledge_service(sf, config.knowledge)
+    if service is None:
+        logger.error("Knowledge base disabled (knowledge.enabled=false) or memory backend; nothing to do.")
+        return 1
+
+    notes = await service.repo.list_all(include_archived=True)
+    if only_note_id:
+        notes = [n for n in notes if n.get("id") == only_note_id]
+
+    scanned = changed = 0
+    for note in notes:
+        scanned += 1
+        body = note.get("content_md") or ""
+        cleaned = clean_note_body(body)
+        if cleaned == body:
+            continue
+        changed += 1
+        nid = note.get("id")
+        title = note.get("title") or nid
+        if dry_run:
+            logger.info("[dry-run] would clean %s — %s (%d → %d chars)", nid, title, len(body), len(cleaned))
+        else:
+            await service.update_note(nid, content_md=cleaned, updated_by="cleanup_script")
+            logger.info("cleaned %s — %s", nid, title)
+
+    await close_engine()
+    logger.info("Done. scanned=%d, %s=%d", scanned, "would-change" if dry_run else "changed", changed)
+    return 0
+
+
+def main() -> int:
+    logging.basicConfig(level=logging.INFO, format="%(message)s")
+    parser = argparse.ArgumentParser(description="Clean legacy `## Related` structural links + `### Sources` from knowledge notes.")
+    parser.add_argument("--dry-run", action="store_true", help="Preview changes without writing.")
+    parser.add_argument("--note-id", default=None, help="Only process this note id.")
+    args = parser.parse_args()
+    return asyncio.run(_run(args.dry_run, args.note_id))
+
+
+if __name__ == "__main__":
+    sys.exit(main())
diff --git a/offline-backend-20260512/backend/scripts/sql/column_additions_mysql.sql b/offline-backend-20260512/backend/scripts/sql/column_additions_mysql.sql
index ae02f476..a858af21 100644
--- a/offline-backend-20260512/backend/scripts/sql/column_additions_mysql.sql
+++ b/offline-backend-20260512/backend/scripts/sql/column_additions_mysql.sql
@@ -65,6 +65,9 @@ CALL _add_column_if_missing('skills', 'detail', 'TEXT NULL DEFAULT NULL');
 -- square_id: 技能所属的发布广场(发布广场功能),对应 migration 20260611_01。
 --          ORM: SkillRow.square_id = String(64), nullable=False, default=''。
 CALL _add_column_if_missing('skills', 'square_id', "VARCHAR(64) NOT NULL DEFAULT ''");
+-- always_on: 常驻(免压缩) —— 开启技能压缩后仍全文注入系统提示词,对应 migration 20260616_03。
+--          ORM: SkillRow.always_on = Boolean, nullable=False, default=False。
+CALL _add_column_if_missing('skills', 'always_on', 'TINYINT(1) NOT NULL DEFAULT 0');
 
 -- ---- agents ------------------------------------------------------------------
 -- square_id: 智能体所属的发布广场(发布广场功能),对应 migration 20260611_01。
@@ -86,6 +89,22 @@ CALL _add_column_if_missing('roundtable_jobs', 'orchestration_plan', 'LONGTEXT N
 --            选入研讨时弹窗默认用它(仍可临时改)。ORM: RoundtableChainRow.seat_mode = String(16), nullable。
 CALL _add_column_if_missing('roundtable_chains', 'seat_mode', 'VARCHAR(16) NULL DEFAULT NULL');
 
+-- ---- 岗位 (positions) 订阅系统,对应 migration 20260616_01 -------------------
+-- positions / position_tag_bindings 是【新表】,由 create_all 自动建,无需补列。
+-- 以下是给【已存在的表】补的列:
+-- users.position_id: 用户所属岗位(1:1)。ORM: UserRow.position_id = String(64) NULL。
+CALL _add_column_if_missing('users', 'position_id', 'VARCHAR(64) NULL DEFAULT NULL');
+-- agent_favorites: origin 区分用户自收藏 vs 岗位授予;position_id 记授予岗位。
+CALL _add_column_if_missing('agent_favorites', 'origin', "VARCHAR(16) NOT NULL DEFAULT 'user'");
+CALL _add_column_if_missing('agent_favorites', 'position_id', 'VARCHAR(64) NULL DEFAULT NULL');
+-- scheduled_task_subscriptions: 同上。
+CALL _add_column_if_missing('scheduled_task_subscriptions', 'origin', "VARCHAR(16) NOT NULL DEFAULT 'user'");
+CALL _add_column_if_missing('scheduled_task_subscriptions', 'position_id', 'VARCHAR(64) NULL DEFAULT NULL');
+
+-- knowledge_notes.folder: 用户自建目录(斜杠分隔路径),驱动产品目录树;NULL=未归类。
+-- (knowledge_folders / knowledge_extract_templates 为【新表】,由 create_all 自动建,无需在此补列。)
+CALL _add_column_if_missing('knowledge_notes', 'folder', 'VARCHAR(512) NULL DEFAULT NULL');
+
 -- ---- 清理临时存储过程 --------------------------------------------------------
 DROP PROCEDURE IF EXISTS _add_column_if_missing;
 
@@ -103,11 +122,25 @@ PREPARE stmt FROM @idx_ddl;
 EXECUTE stmt;
 DEALLOCATE PREPARE stmt;
 
+-- ---- knowledge_notes.folder 索引(幂等) --------------------------------------
+SET @idx_exists = (
+    SELECT COUNT(*) FROM information_schema.statistics
+     WHERE table_schema = DATABASE()
+       AND table_name = 'knowledge_notes'
+       AND index_name = 'ix_knowledge_notes_folder'
+);
+SET @idx_ddl = IF(@idx_exists = 0,
+    'CREATE INDEX `ix_knowledge_notes_folder` ON `knowledge_notes` (`folder`(191))',
+    'SELECT ''SKIPPED  ix_knowledge_notes_folder (already exists)'' AS result');
+PREPARE stmt FROM @idx_ddl;
+EXECUTE stmt;
+DEALLOCATE PREPARE stmt;
+
 -- ---- 执行后自检:确认本批次涉及的列都已存在 ---------------------------------
 SELECT table_name, column_name
   FROM information_schema.columns
  WHERE table_schema = DATABASE()
-   AND ((table_name = 'skills' AND column_name IN ('name_zh', 'detail', 'square_id'))
+   AND ((table_name = 'skills' AND column_name IN ('name_zh', 'detail', 'square_id', 'always_on'))
         OR (table_name = 'agents' AND column_name IN ('square_id'))
         OR (table_name = 'roundtable_drafts' AND column_name = 'task_id'))
  ORDER BY table_name, column_name;
diff --git a/offline-backend-20260512/backend/tests/test_agent_modules.py b/offline-backend-20260512/backend/tests/test_agent_modules.py
new file mode 100644
index 00000000..3b8e8fff
--- /dev/null
+++ b/offline-backend-20260512/backend/tests/test_agent_modules.py
@@ -0,0 +1,60 @@
+"""Tests for the admin-configured「按类型聚合」module → agent-tag config.
+
+Backs the zzJC / zzXD module left sidebar (需求 0616 #3/#4): an admin picks
+which agent tags belong to the module, persisted in ``system_settings.json``.
+"""
+
+from __future__ import annotations
+
+import json
+from pathlib import Path
+
+import pytest
+
+from deerflow.config import system_settings as ss
+
+
+@pytest.fixture
+def tmp_runtime_home(monkeypatch: pytest.MonkeyPatch, tmp_path: Path) -> Path:
+    """Redirect ``runtime_home()`` to a per-test tmp dir."""
+
+    def fake_home() -> Path:
+        return tmp_path
+
+    monkeypatch.setattr(ss, "runtime_home", fake_home)
+    return tmp_path
+
+
+def test_load_returns_empty_modules_when_missing(tmp_runtime_home: Path) -> None:
+    settings = ss.load_system_settings()
+    assert settings.agent_modules.modules == {}
+    assert ss.get_module_tag_ids(settings.agent_modules, "zzjc") == []
+
+
+def test_set_module_tag_ids_dedupes_and_strips() -> None:
+    base = ss.AgentModuleSettings()
+    updated = ss.set_module_tag_ids(base, "zzjc", ["  t1 ", "t2", "t1", "", "  "])
+    assert updated.modules["zzjc"] == ["t1", "t2"]
+    # Original is untouched (returns a copy).
+    assert base.modules == {}
+
+
+def test_set_module_tag_ids_keeps_other_modules() -> None:
+    base = ss.set_module_tag_ids(ss.AgentModuleSettings(), "zzjc", ["t1"])
+    both = ss.set_module_tag_ids(base, "zzxd", ["t2", "t3"])
+    assert ss.get_module_tag_ids(both, "zzjc") == ["t1"]
+    assert ss.get_module_tag_ids(both, "zzxd") == ["t2", "t3"]
+
+
+def test_save_then_load_roundtrips(tmp_runtime_home: Path) -> None:
+    current = ss.load_system_settings()
+    updated = ss.set_module_tag_ids(current.agent_modules, "zzjc", ["tag-a", "tag-b"])
+    ss.save_system_settings(current.model_copy(update={"agent_modules": updated}))
+
+    file_path = tmp_runtime_home / "system_settings.json"
+    assert file_path.is_file()
+    saved = json.loads(file_path.read_text(encoding="utf-8"))
+    assert saved["agent_modules"]["modules"]["zzjc"] == ["tag-a", "tag-b"]
+
+    reloaded = ss.load_system_settings()
+    assert ss.get_module_tag_ids(reloaded.agent_modules, "zzjc") == ["tag-a", "tag-b"]
diff --git a/offline-backend-20260512/backend/tests/test_recommended_questions_tags.py b/offline-backend-20260512/backend/tests/test_recommended_questions_tags.py
new file mode 100644
index 00000000..f4d19e78
--- /dev/null
+++ b/offline-backend-20260512/backend/tests/test_recommended_questions_tags.py
@@ -0,0 +1,79 @@
+"""Router-level tests for recommended-question tag embedding + filtering.
+
+The list endpoint coroutine is invoked directly with a stub request whose
+``app.state`` carries a fake question store and an in-memory tag store, mirroring
+the lightweight router tests already in this suite (see ``test_tags_api.py``).
+"""
+
+from __future__ import annotations
+
+from types import SimpleNamespace
+
+import pytest
+
+from app.gateway.routers import recommended_questions as rq_router
+from deerflow.persistence.tags import MemoryTagStore
+
+
+class _FakeQuestionStore:
+    def __init__(self, rows: list[dict]) -> None:
+        self._rows = rows
+
+    async def list_all(self) -> list[dict]:
+        return list(self._rows)
+
+
+def _request(question_store, tag_store) -> SimpleNamespace:
+    state = SimpleNamespace(
+        user=None,
+        recommended_question_store=question_store,
+        tag_store=tag_store,
+        position_sync=None,
+    )
+    return SimpleNamespace(state=state, app=SimpleNamespace(state=state))
+
+
+def _rows() -> list[dict]:
+    return [
+        {"id": "q1", "title": "市场分析", "content": "分析一下市场", "sort_order": 0},
+        {"id": "q2", "title": "理赔报告", "content": "生成理赔报告", "sort_order": 1},
+    ]
+
+
+@pytest.mark.asyncio
+async def test_list_embeds_assigned_tags() -> None:
+    tag_store = MemoryTagStore()
+    tag = await tag_store.create_tag(
+        {"id": "t1", "name": "理赔", "tag_type": "recommended_question"}
+    )
+    await tag_store.set_target_tags("recommended_question", "q2", [tag["id"]])
+
+    req = _request(_FakeQuestionStore(_rows()), tag_store)
+    resp = await rq_router.list_recommended_questions(req, all=True)
+
+    by_id = {q.id: q for q in resp.questions}
+    assert by_id["q1"].tags == []
+    assert [t["name"] for t in by_id["q2"].tags] == ["理赔"]
+
+
+@pytest.mark.asyncio
+async def test_list_filters_by_tag_ids() -> None:
+    tag_store = MemoryTagStore()
+    tag = await tag_store.create_tag(
+        {"id": "t1", "name": "理赔", "tag_type": "recommended_question"}
+    )
+    await tag_store.set_target_tags("recommended_question", "q2", [tag["id"]])
+
+    req = _request(_FakeQuestionStore(_rows()), tag_store)
+    resp = await rq_router.list_recommended_questions(req, all=True, tag_ids=[tag["id"]])
+
+    assert [q.id for q in resp.questions] == ["q2"]
+
+
+@pytest.mark.asyncio
+async def test_list_without_tag_store_does_not_break() -> None:
+    req = _request(_FakeQuestionStore(_rows()), None)
+    resp = await rq_router.list_recommended_questions(req, all=True)
+
+    assert [q.id for q in resp.questions] == ["q1", "q2"]
+    assert all(q.tags == [] for q in resp.questions)
diff --git a/offline-backend-20260512/backend/tests/test_research_templates.py b/offline-backend-20260512/backend/tests/test_research_templates.py
new file mode 100644
index 00000000..6d93df2e
--- /dev/null
+++ b/offline-backend-20260512/backend/tests/test_research_templates.py
@@ -0,0 +1,124 @@
+"""Tests for the research-template store (课题研究模板，需求 0616 #5).
+
+A research template bundles writing-form presets (article type / outline /
+material source / word count / audience / strict mode / topic) for one-click
+pre-fill. Covers create (with defaults) / list ordering / get / partial update /
+delete.
+"""
+
+from __future__ import annotations
+
+import pytest
+import pytest_asyncio
+from sqlalchemy.ext.asyncio import async_sessionmaker, create_async_engine
+
+from deerflow.persistence.base import Base
+from deerflow.persistence.research_templates import ResearchTemplateRepository
+
+
+@pytest_asyncio.fixture
+async def store(tmp_path):
+    engine = create_async_engine(f"sqlite+aiosqlite:///{tmp_path / 'rt.db'}")
+    async with engine.begin() as conn:
+        import deerflow.persistence.research_templates.model  # noqa: F401
+
+        await conn.run_sync(Base.metadata.create_all)
+    sf = async_sessionmaker(engine, expire_on_commit=False)
+    return ResearchTemplateRepository(sf)
+
+
+@pytest.mark.asyncio
+async def test_create_applies_defaults(store: ResearchTemplateRepository):
+    row = await store.create(name="深度分析模板")
+    assert row["name"] == "深度分析模板"
+    assert row["material_source"] == "general"
+    assert row["word_count"] == 800
+    assert row["strict_mode"] is False
+    assert row["article_type_key"] is None
+    assert row["id"]
+
+
+@pytest.mark.asyncio
+async def test_create_full_then_get_roundtrips(store: ResearchTemplateRepository):
+    created = await store.create(
+        name="舆情研究",
+        description="围绕舆情的研究模板",
+        article_type_key="analysis",
+        outline="一、背景\n二、分析\n三、结论",
+        material_source="notebook",
+        word_count=1500,
+        audience="决策者",
+        strict_mode=True,
+        topic="某品牌近期舆情",
+        sort_order=2,
+    )
+    fetched = await store.get(created["id"])
+    assert fetched is not None
+    assert fetched["article_type_key"] == "analysis"
+    assert fetched["outline"].startswith("一、背景")
+    assert fetched["material_source"] == "notebook"
+    assert fetched["word_count"] == 1500
+    assert fetched["audience"] == "决策者"
+    assert fetched["strict_mode"] is True
+    assert fetched["topic"] == "某品牌近期舆情"
+
+
+@pytest.mark.asyncio
+async def test_list_orders_by_sort_order(store: ResearchTemplateRepository):
+    await store.create(name="B", sort_order=2)
+    await store.create(name="A", sort_order=1)
+    rows = await store.list()
+    assert [r["name"] for r in rows] == ["A", "B"]
+
+
+@pytest.mark.asyncio
+async def test_update_partial_keeps_other_fields(store: ResearchTemplateRepository):
+    created = await store.create(name="模板", word_count=800, strict_mode=False)
+    updated = await store.update(created["id"], word_count=2000, strict_mode=True)
+    assert updated is not None
+    assert updated["word_count"] == 2000
+    assert updated["strict_mode"] is True
+    assert updated["name"] == "模板"  # untouched
+
+
+@pytest.mark.asyncio
+async def test_update_missing_returns_none(store: ResearchTemplateRepository):
+    assert await store.update("nope", name="x") is None
+
+
+@pytest.mark.asyncio
+async def test_delete(store: ResearchTemplateRepository):
+    created = await store.create(name="待删")
+    assert await store.delete(created["id"]) is True
+    assert await store.get(created["id"]) is None
+    assert await store.delete(created["id"]) is False
+
+
+@pytest.mark.asyncio
+async def test_seed_defaults_inserts_builtins_when_empty(store: ResearchTemplateRepository):
+    inserted = await store.seed_defaults()
+    assert inserted == 3
+    rows = await store.list()
+    names = [r["name"] for r in rows]
+    assert names == ["网摘", "周报", "要讯"]  # sort_order 0/1/2
+    # 内置模板内容完整（范例质量）。
+    by_name = {r["name"]: r for r in rows}
+    assert by_name["网摘"]["article_type_key"] == "news"
+    assert by_name["周报"]["strict_mode"] is True
+    assert by_name["要讯"]["outline"]
+    assert all(r["topic"] for r in rows)
+
+
+@pytest.mark.asyncio
+async def test_seed_defaults_is_idempotent(store: ResearchTemplateRepository):
+    assert await store.seed_defaults() == 3
+    # 已有模板 → 再次播种跳过，不重复、不覆盖。
+    assert await store.seed_defaults() == 0
+    assert len(await store.list()) == 3
+
+
+@pytest.mark.asyncio
+async def test_seed_defaults_skips_when_user_has_templates(store: ResearchTemplateRepository):
+    await store.create(name="用户自建")
+    assert await store.seed_defaults() == 0
+    assert [r["name"] for r in await store.list()] == ["用户自建"]
diff --git a/offline-backend-20260512/config.yaml b/offline-backend-20260512/config.yaml
index b753224f..be0d5b21 100644
--- a/offline-backend-20260512/config.yaml
+++ b/offline-backend-20260512/config.yaml
@@ -303,14 +303,27 @@ log_system:
   label: "日志系统"
 
 summarization:
-  enabled: true
+  # 开关已移到聊天框下方(参考文献下面)的「上下文压缩」按钮,默认关闭。
+  # 这里 enabled 仅作为非 Web 路径(渠道/定时任务)的后备开关;下面的 trigger/keep
+  # 等仅是调参,Web 端是否压缩由前端按钮(summarization_enabled)决定。
+  enabled: false
   # model_name: deepseek-chat
   trigger:
     - type: tokens
-      value: 60000
+      value: 8000
     - type: messages
-      value: 80
+      value: 12
+  # Keep a small *token* budget of recent context (same unit as the token
+  # trigger, well below it). This is what the model receives after compaction:
+  # [summary, ~16000 tokens of recent turns]. Small keep = small model context
+  # (fast, no overflow) AND infrequent re-compaction — the preserved window
+  # grows from ~16000 up to the 60000 trigger before folding into a new summary,
+  # which spans many turns. The full conversation still stays in state for the
+  # page; only what is sent to the model is compressed.
   keep:
-    type: messages
-    value: 20
+    type: messages   # 测试值:压缩后只保留最近 6 条原始消息
+    value: 6
+    # 生产值改用 token 预算:
+    # type: tokens
+    # value: 16000
   trim_tokens_to_summarize: 4000