feat: integrate WeKnora llmwiki knowledge flow

.gitignore

@@ -1,6 +1,9 @@
 # Workspace/editor
-.vscode/
-frontend-web/.vscode/
+.vscode/*
+!.vscode/launch.json
+!.vscode/tasks.json
+!.vscode/extensions.json
+frontend-web/.vscode/
 node_modules/
 # Claude Code local-only settings (per-user permissions, not shared)
 .claude/settings.local.json
@@ -52,8 +55,9 @@ offline-backend-20260505/bundle/start_backend.sh
 offline-backend-20260505/bundle/stop_backend.sh
 
 # offline-backend-20260512 runtime artifacts (large files, local config, data)
-offline-backend-20260512/.env.example
-offline-backend-20260512/DEPLOY_BACKEND_OFFLINE_ZH.md
+offline-backend-20260512/.env.example
+offline-backend-20260512/.env
+offline-backend-20260512/DEPLOY_BACKEND_OFFLINE_ZH.md
 offline-backend-20260512/SHA256SUMS
 offline-backend-20260512/config.yaml.example
 offline-backend-20260512/data/
@@ -61,12 +65,14 @@ offline-backend-20260512/deerflow-backend-offline-image-20260505.tar
 offline-backend-20260512/extensions_config.json
 offline-backend-20260512/skills/
 offline-backend-20260512/wheelhouse/
-offline-backend-20260512/backend/.deer-flow/data/*.db
-offline-backend-20260512/backend/.deer-flow/data/*.db-*
-offline-backend-20260512/backend/.deer-flow/*
-offline-backend-20260512/start_backend.sh
-offline-backend-20260512/stop_backend.sh
-offline-backend-20260512/check_backend.sh
+offline-backend-20260512/backend/.deer-flow/data/*.db
+offline-backend-20260512/backend/.deer-flow/data/*.db-*
+offline-backend-20260512/backend/.deer-flow/*
+offline-backend-20260512/backend/.env
+offline-backend-20260512/backend/.env.local
+offline-backend-20260512/start_backend.sh
+offline-backend-20260512/stop_backend.sh
+offline-backend-20260512/check_backend.sh
 
 hermes-agent-main/
 
@@ -80,8 +86,10 @@ offline-backend-20260512/backend/skill_evolution.json
 skills/
 hermes-agent-main/
 .gitignore
-offline-backend-20260512/backend/tests/*
-deploy
+offline-backend-20260512/backend/tests/*
+!offline-backend-20260512/backend/tests/test_llmwiki_weknora.py
+!offline-backend-20260512/backend/tests/test_health_auth_bypass.py
+deploy
 .history
 
 # Cloned third-party repos (local only)

.vscode/extensions.json

@@ -0,0 +1,6 @@
+{
+  "recommendations": [
+    "ms-python.python",
+    "ms-python.debugpy"
+  ]
+}

.vscode/launch.json

@@ -0,0 +1,66 @@
+{
+  "version": "0.2.0",
+  "configurations": [
+    {
+      "name": "DeerFlow Backend FastAPI (8001)",
+      "type": "debugpy",
+      "request": "launch",
+      "module": "uvicorn",
+      "args": [
+        "app.gateway.app:app",
+        "--host",
+        "0.0.0.0",
+        "--port",
+        "8001",
+        "--reload"
+      ],
+      "cwd": "${workspaceFolder}/offline-backend-20260512/backend",
+      "python": "${workspaceFolder}/offline-backend-20260512/backend/.venv/Scripts/python.exe",
+      "envFile": "${workspaceFolder}/offline-backend-20260512/.env",
+      "env": {
+        "PYTHONPATH": "${workspaceFolder}/offline-backend-20260512/backend;${workspaceFolder}/offline-backend-20260512/backend/.deer-flow/python-vendor",
+        "GATEWAY_CORS_ORIGINS": "http://localhost:5174,http://127.0.0.1:5174,http://localhost:8080,http://127.0.0.1:8080"
+      },
+      "console": "integratedTerminal",
+      "justMyCode": false,
+      "subProcess": true
+    },
+    {
+      "name": "DeerFlow Frontend Vite (5174)",
+      "type": "node",
+      "request": "launch",
+      "runtimeExecutable": "pnpm",
+      "runtimeArgs": [
+        "exec",
+        "vite",
+        "--host",
+        "0.0.0.0",
+        "--port",
+        "5174"
+      ],
+      "cwd": "${workspaceFolder}/frontend-web",
+      "env": {
+        "VITE_BACKEND_BASE_URL": "http://127.0.0.1:8001",
+        "VITE_LANGGRAPH_BASE_URL": "http://127.0.0.1:8001/api"
+      },
+      "console": "integratedTerminal",
+      "outputCapture": "std",
+      "serverReadyAction": {
+        "pattern": "Local:\\s+(https?://[^\\s]+)",
+        "uriFormat": "%s",
+        "action": "debugWithChrome",
+        "webRoot": "${workspaceFolder}/frontend-web"
+      }
+    }
+  ],
+  "compounds": [
+    {
+      "name": "DeerFlow Full Stack (Backend + Frontend)",
+      "configurations": [
+        "DeerFlow Backend FastAPI (8001)",
+        "DeerFlow Frontend Vite (5174)"
+      ],
+      "stopAll": true
+    }
+  ]
+}

.vscode/tasks.json

@@ -0,0 +1,52 @@
+{
+  "version": "2.0.0",
+  "tasks": [
+    {
+      "label": "DeerFlow：检查本地开发环境",
+      "type": "shell",
+      "command": "$backendPython = '${workspaceFolder}\\offline-backend-20260512\\backend\\.venv\\Scripts\\python.exe'; if (-not (Test-Path -LiteralPath $backendPython)) { throw '后端虚拟环境不存在，请先在 backend 目录执行 uv sync' }; & $backendPython -c \"import uvicorn; print('Backend Python OK')\"; pnpm --dir '${workspaceFolder}\\frontend-web' exec vite --version",
+      "options": {
+        "shell": {
+          "executable": "powershell.exe",
+          "args": [
+            "-NoProfile",
+            "-ExecutionPolicy",
+            "Bypass",
+            "-Command"
+          ]
+        }
+      },
+      "problemMatcher": []
+    },
+    {
+      "label": "DeerFlow：前端生产构建",
+      "type": "process",
+      "command": "pnpm",
+      "args": [
+        "build"
+      ],
+      "options": {
+        "cwd": "${workspaceFolder}/frontend-web"
+      },
+      "problemMatcher": []
+    },
+    {
+      "label": "DeerFlow：WeKnora 专项测试",
+      "type": "process",
+      "command": "${workspaceFolder}/offline-backend-20260512/backend/.venv/Scripts/python.exe",
+      "args": [
+        "-m",
+        "pytest",
+        "tests/test_llmwiki_weknora.py",
+        "-v"
+      ],
+      "options": {
+        "cwd": "${workspaceFolder}/offline-backend-20260512/backend",
+        "env": {
+          "PYTHONPATH": "."
+        }
+      },
+      "problemMatcher": []
+    }
+  ]
+}

frontend-web/src/components/wiki/WikiHeader.tsx

@@ -8,6 +8,7 @@ import {
   SlidersHorizontal,
   Upload,
 } from "lucide-react";
+import type { ReactNode } from "react";
 
 import { cn } from "@/lib/utils";
 
@@ -18,7 +19,7 @@ function HeaderButton({
   variant = "outline",
   className,
 }: {
-  icon: React.ReactNode;
+  icon: ReactNode;
   label: string;
   onClick: () => void;
   variant?: "outline" | "primary";
@@ -52,6 +53,7 @@ export function WikiHeader({
   onCreate,
   onManageFolders,
   onManageTemplates,
+  extraActions,
 }: {
   onToggleSidebar: () => void;
   onImportFile: () => void;
@@ -61,6 +63,7 @@ export function WikiHeader({
   onCreate: () => void;
   onManageFolders: () => void;
   onManageTemplates: () => void;
+  extraActions?: ReactNode;
 }) {
   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">
@@ -79,6 +82,7 @@ export function WikiHeader({
       </div>
 
       <div className="flex items-center gap-2">
+        {extraActions}
         <HeaderButton
           icon={<Upload className="size-4" />}
           label="导入文件"

frontend-web/src/components/workspace/agents/agent-detail-panel.tsx

@@ -3,6 +3,7 @@
 import {
   BotIcon,
   CheckIcon,
+  DatabaseIcon,
   GlobeIcon,
   InfoIcon,
   Loader2Icon,
@@ -43,6 +44,7 @@ import {
 import { Switch } from "@/components/ui/switch";
 import { Textarea } from "@/components/ui/textarea";
 import { AgentQuestionEditor } from "@/components/workspace/agents/agent-question-editor";
+import { LlmWikiKnowledgeSelector } from "@/components/workspace/llmwiki-knowledge-selector";
 import { AgentRecentChats } from "@/components/workspace/agents/agent-recent-chats";
 import { useArtifactsOptional } from "@/components/workspace/artifacts";
 import { EnabledSkillList } from "@/components/workspace/agents/enabled-skill-list";
@@ -80,6 +82,7 @@ interface AgentDraft {
   templateName: string;
   /** 会商席位「技能识别强化」开关。 */
   seatSkillDirective: boolean;
+  knowledgeBaseIds: string[];
 }
 
 function draftFromAgent(agent: Agent): AgentDraft {
@@ -93,6 +96,7 @@ function draftFromAgent(agent: Agent): AgentDraft {
     featured: agent.featured_order != null,
     templateName: agent.template_name ?? "",
     seatSkillDirective: agent.seat_skill_directive ?? false,
+    knowledgeBaseIds: agent.llmwiki_knowledge_base_ids ?? [],
   };
 }
 
@@ -310,6 +314,11 @@ export function AgentDetailPanel({
       request.template_name = draft.templateName.trim();
     if (draft.seatSkillDirective !== (agent.seat_skill_directive ?? false))
       request.seat_skill_directive = draft.seatSkillDirective;
+    if (
+      JSON.stringify(draft.knowledgeBaseIds) !==
+      JSON.stringify(agent.llmwiki_knowledge_base_ids ?? [])
+    )
+      request.llmwiki_knowledge_base_ids = draft.knowledgeBaseIds;
     if (nextPublished !== agent.published) request.published = nextPublished;
     if (JSON.stringify(draft.skills) !== JSON.stringify(agent.skills ?? []))
       request.skills = draft.skills;
@@ -579,6 +588,13 @@ export function AgentDetailPanel({
           </div>
         )}
 
+        {!editing && (agent.llmwiki_knowledge_base_ids?.length ?? 0) > 0 && (
+          <div className="text-muted-foreground mt-2 flex items-center gap-1 text-xs">
+            <DatabaseIcon className="size-3 shrink-0" />
+            LLMWiki 知识库：{agent.llmwiki_knowledge_base_ids?.length ?? 0} 个
+          </div>
+        )}
+
         {/* Publish / featured toggles — only meaningful in edit mode. */}
         {editing && draft && (
           <div className="mt-3 space-y-2">
@@ -596,6 +612,14 @@ export function AgentDetailPanel({
                 配置后，对话生成的 Markdown 文件可一键发布到知识入库接口。
               </p>
             </div>
+            <LlmWikiKnowledgeSelector
+              source="all"
+              value={draft.knowledgeBaseIds}
+              onChange={(knowledgeBaseIds) =>
+                setDraft({ ...draft, knowledgeBaseIds })
+              }
+              variant="panel"
+            />
             <div className="bg-background/60 flex items-center justify-between rounded-lg border px-3 py-2">
               <div className="pr-3">
                 <p className="text-sm font-medium">会商技能识别强化</p>

frontend-web/src/components/workspace/input-box.tsx

@@ -22,9 +22,10 @@ import {
   useEffect,
   useMemo,
   useRef,
-  useState,
-  type ComponentProps,
-} from "react";
+  useState,
+  type ComponentProps,
+  type ReactNode,
+} from "react";
 
 import {
   PromptInput,
@@ -132,9 +133,10 @@ export function InputBox({
   onFollowupsVisibilityChange,
   onSubmit,
   onStop,
-  promptPrefix,
-  agentSelector,
-  assistantId,
+  promptPrefix,
+  agentSelector,
+  knowledgeSelector,
+  assistantId,
   hideAdminRecommendedQuestions,
   compact,
   ...props
@@ -190,11 +192,13 @@ export function InputBox({
    * 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;
-  };
-}) {
+  agentSelector?: {
+    selectedAgentId: string | null;
+    onAgentChange: (agentId: string | null) => void;
+  };
+  /** WeKnora-style knowledge-space picker rendered inside the composer toolbar. */
+  knowledgeSelector?: ReactNode;
+}) {
   const { t } = useI18n();
   const [localSettings, setLocalSettings] = useLocalSettings();
   const searchParams = useSearchParams();
@@ -1081,9 +1085,10 @@ export function InputBox({
                 </PromptInputActionMenuContent>
               </PromptInputActionMenu>
             )}
-          </PromptInputTools>
-          <PromptInputTools className={cn("gap-1.5", compact && "gap-2")}>
-            {agentSelector && (
+          </PromptInputTools>
+          <PromptInputTools className={cn("gap-1.5", compact && "gap-2")}>
+            {knowledgeSelector}
+            {agentSelector && (
               <AgentSelector
                 selectedAgentId={agentSelector.selectedAgentId}
                 onAgentChange={agentSelector.onAgentChange}

frontend-web/src/components/workspace/llmwiki-knowledge-selector.tsx

@@ -0,0 +1,353 @@
+"use client";
+
+import {
+  Check,
+  Database,
+  Folder,
+  Globe2,
+  Loader2,
+  Search,
+} from "lucide-react";
+import { useEffect, useMemo, useRef, useState } from "react";
+
+import { Badge } from "@/components/ui/badge";
+import { Button } from "@/components/ui/button";
+import {
+  DropdownMenu,
+  DropdownMenuCheckboxItem,
+  DropdownMenuContent,
+  DropdownMenuLabel,
+  DropdownMenuSeparator,
+  DropdownMenuTrigger,
+} from "@/components/ui/dropdown-menu";
+import { Input } from "@/components/ui/input";
+import { cn } from "@/lib/utils";
+import {
+  type LlmWikiKnowledgeBase,
+  useAgentLlmWikiKnowledgeBases,
+  useLlmWikiKnowledgeBases,
+  useLlmWikiRuntime,
+} from "@/strategy-components/api/llmwiki";
+
+type SelectorSource = "all" | "agent";
+type SelectorVariant = "popover" | "panel";
+
+function groupRows(rows: LlmWikiKnowledgeBase[]) {
+  return [
+    {
+      key: "personal",
+      label: "我的知识空间",
+      rows: rows.filter((item) => item.publication_status !== "published" || item.is_owner),
+    },
+    {
+      key: "public",
+      label: "公共知识空间",
+      rows: rows.filter((item) => item.publication_status === "published" && !item.is_owner),
+    },
+  ].filter((group) => group.rows.length > 0);
+}
+
+function SpaceIcon({ item }: { item: LlmWikiKnowledgeBase }) {
+  return (
+    <span
+      className={cn(
+        "flex size-7 shrink-0 items-center justify-center rounded-md",
+        item.publication_status === "published"
+          ? "bg-emerald-500/10 text-emerald-600 dark:text-emerald-400"
+          : "bg-primary/10 text-primary",
+      )}
+    >
+      {item.publication_status === "published" ? (
+        <Globe2 className="size-3.5" />
+      ) : (
+        <Folder className="size-3.5" />
+      )}
+    </span>
+  );
+}
+
+function SelectionContent({
+  rows,
+  value,
+  onChange,
+  search,
+  onSearchChange,
+  loading,
+  error,
+  source,
+  variant,
+}: {
+  rows: LlmWikiKnowledgeBase[];
+  value: string[];
+  onChange: (value: string[]) => void;
+  search: string;
+  onSearchChange: (value: string) => void;
+  loading: boolean;
+  error: unknown;
+  source: SelectorSource;
+  variant: SelectorVariant;
+}) {
+  const selected = useMemo(() => new Set(value), [value]);
+  const filtered = useMemo(() => {
+    const query = search.trim().toLocaleLowerCase();
+    if (!query) return rows;
+    return rows.filter((item) =>
+      `${item.name} ${item.description}`.toLocaleLowerCase().includes(query),
+    );
+  }, [rows, search]);
+  const groups = useMemo(() => groupRows(filtered), [filtered]);
+  const toggle = (id: string) => {
+    onChange(selected.has(id) ? value.filter((item) => item !== id) : [...value, id]);
+  };
+
+  return (
+    <>
+      <div className="border-b p-2.5">
+        <div className="relative">
+          <Search className="pointer-events-none absolute left-2.5 top-1/2 size-3.5 -translate-y-1/2 text-muted-foreground" />
+          <Input
+            value={search}
+            onChange={(event) => onSearchChange(event.target.value)}
+            onKeyDown={(event) => event.stopPropagation()}
+            placeholder="搜索知识空间"
+            className="h-8 bg-muted/40 pl-8 text-xs"
+          />
+        </div>
+      </div>
+
+      <div className={cn("overflow-y-auto p-1.5", variant === "popover" ? "max-h-64" : "max-h-80")}>
+        {loading ? (
+          <div className="flex h-28 items-center justify-center text-xs text-muted-foreground">
+            <Loader2 className="mr-2 size-4 animate-spin" />正在读取知识空间
+          </div>
+        ) : error ? (
+          <div className="m-2 rounded-lg border border-destructive/30 bg-destructive/5 p-3 text-xs text-destructive">
+            {error instanceof Error ? error.message : "知识空间读取失败"}
+          </div>
+        ) : groups.length === 0 ? (
+          <div className="flex h-28 items-center justify-center px-4 text-center text-xs text-muted-foreground">
+            {search
+              ? "没有匹配的知识空间"
+              : source === "agent"
+                ? "该智能体尚未绑定可用知识空间"
+                : "当前账号暂无可用知识空间"}
+          </div>
+        ) : (
+          groups.map((group, groupIndex) => (
+            <div key={group.key}>
+              {groupIndex > 0 && (
+                variant === "popover"
+                  ? <DropdownMenuSeparator className="my-1.5" />
+                  : <div className="my-1.5 border-t" />
+              )}
+              {variant === "popover" ? (
+                <DropdownMenuLabel className="flex items-center gap-1.5 px-2 py-1 text-[11px] font-semibold text-muted-foreground">
+                  {group.key === "personal" ? <Folder className="size-3" /> : <Globe2 className="size-3" />}{group.label}
+                </DropdownMenuLabel>
+              ) : (
+                <div className="flex items-center gap-1.5 px-2 py-1 text-[11px] font-semibold text-muted-foreground">
+                  {group.key === "personal" ? <Folder className="size-3" /> : <Globe2 className="size-3" />}{group.label}
+                </div>
+              )}
+              {group.rows.map((item) => {
+                const rowContent = (
+                  <>
+                    <SpaceIcon item={item} />
+                    <div className="min-w-0 flex-1">
+                      <div className="truncate text-xs font-medium">{item.name}</div>
+                      <div className="mt-0.5 truncate text-[10px] text-muted-foreground">
+                        {item.type === "faq" ? "FAQ" : "文档库"} · {item.knowledge_count ?? 0} 项知识
+                      </div>
+                    </div>
+                    {item.publication_status === "published" && (
+                      <Badge variant="outline" className="h-4 px-1 text-[9px]">公共</Badge>
+                    )}
+                  </>
+                );
+                return variant === "popover" ? (
+                  <DropdownMenuCheckboxItem
+                    key={item.id}
+                    checked={selected.has(item.id)}
+                    onCheckedChange={() => toggle(item.id)}
+                    onSelect={(event) => event.preventDefault()}
+                    className={cn(
+                      "my-0.5 min-h-10 rounded-md py-1.5 pr-2 pl-7",
+                      selected.has(item.id) && "bg-primary/8 focus:bg-primary/12",
+                    )}
+                  >
+                    {rowContent}
+                  </DropdownMenuCheckboxItem>
+                ) : (
+                  <button
+                    key={item.id}
+                    type="button"
+                    role="checkbox"
+                    aria-checked={selected.has(item.id)}
+                    onClick={() => toggle(item.id)}
+                    className={cn(
+                      "my-0.5 flex min-h-11 w-full items-center gap-2 rounded-md px-2 py-1.5 text-left outline-none transition-colors hover:bg-muted/60 focus-visible:ring-2 focus-visible:ring-primary",
+                      selected.has(item.id) && "bg-primary/8",
+                    )}
+                  >
+                    <span className={cn(
+                      "flex size-4 shrink-0 items-center justify-center rounded border",
+                      selected.has(item.id) && "border-primary bg-primary text-primary-foreground",
+                    )}>
+                      {selected.has(item.id) && <Check className="size-3" />}
+                    </span>
+                    {rowContent}
+                  </button>
+                );
+              })}
+            </div>
+          ))
+        )}
+      </div>
+
+      <div className="flex items-center justify-between gap-2 border-t bg-muted/20 px-2.5 py-2">
+        <span className="text-[11px] text-muted-foreground">已选择 {value.length} 个</span>
+        <div className="flex items-center gap-1">
+          <Button type="button" variant="ghost" size="sm" className="h-7 px-2 text-xs" onClick={() => onChange(filtered.map((item) => item.id))}>
+            全选
+          </Button>
+          <Button type="button" variant="ghost" size="sm" className="h-7 px-2 text-xs" onClick={() => onChange([])}>
+            清空
+          </Button>
+        </div>
+      </div>
+    </>
+  );
+}
+
+export function LlmWikiKnowledgeSelector({
+  value,
+  onChange,
+  source = "all",
+  agentId,
+  autoSelectAll = false,
+  disabled = false,
+  className,
+  label = "知识空间",
+  variant = "popover",
+}: {
+  value: string[];
+  onChange: (value: string[]) => void;
+  source?: SelectorSource;
+  agentId?: string | null;
+  autoSelectAll?: boolean;
+  disabled?: boolean;
+  className?: string;
+  label?: string;
+  variant?: SelectorVariant;
+}) {
+  const [open, setOpen] = useState(false);
+  const [search, setSearch] = useState("");
+  const runtime = useLlmWikiRuntime();
+  const all = useLlmWikiKnowledgeBases(
+    "all",
+    runtime.data?.provider === "weknora",
+  );
+  const agent = useAgentLlmWikiKnowledgeBases(
+    agentId,
+    runtime.data?.provider === "weknora" && source === "agent" && !!agentId,
+  );
+  const rows = all.data ?? [];
+  const defaultIds = useMemo(() => {
+    if (source !== "agent") {
+      return rows.map((item) => item.id);
+    }
+    return (agent.data ?? []).map((item) => item.id);
+  }, [agent.data, rows, source]);
+  const loading = all.isLoading || (source === "agent" && !!agentId && agent.isLoading);
+  const error = all.error ?? (source === "agent" ? agent.error : null);
+  const availableIds = useMemo(() => new Set(rows.map((item) => item.id)), [rows]);
+  const initializationKey = `${source}:${agentId ?? "all"}:${defaultIds.join(",")}`;
+  const initialized = useRef<string | null>(null);
+
+  useEffect(() => {
+    if (!all.data || (source === "agent" && agentId && !agent.data) || initialized.current === initializationKey) return;
+    initialized.current = initializationKey;
+    const valid = value.filter((id) => availableIds.has(id));
+    const next = autoSelectAll && valid.length === 0 && defaultIds.length > 0
+      ? defaultIds.filter((id) => availableIds.has(id))
+      : valid;
+    if (next.length !== value.length || next.some((id, index) => id !== value[index])) {
+      onChange(next);
+    }
+  }, [agent.data, agentId, all.data, autoSelectAll, availableIds, defaultIds, initializationKey, onChange, source, value]);
+
+  useEffect(() => {
+    if (!open) setSearch("");
+  }, [open]);
+
+  if (runtime.isError || runtime.data?.provider !== "weknora") return null;
+
+  const content = (
+    <SelectionContent
+      rows={rows}
+      value={value}
+      onChange={onChange}
+      search={search}
+      onSearchChange={setSearch}
+      loading={loading}
+      error={error}
+      source={source}
+      variant={variant}
+    />
+  );
+
+  if (variant === "panel") {
+    return (
+      <div className={cn("overflow-hidden rounded-xl border bg-background shadow-sm", className)}>
+        <div className="flex items-start justify-between gap-4 border-b px-4 py-3">
+          <div>
+            <div className="flex items-center gap-2 text-sm font-semibold">
+              <Database className="size-4 text-primary" />绑定知识空间
+            </div>
+            <p className="mt-1 text-xs text-muted-foreground">
+              新建该智能体对话时，只能从已绑定空间中选择本次检索范围。
+            </p>
+          </div>
+          <Badge variant="secondary" className="shrink-0">{value.length} 个</Badge>
+        </div>
+        {content}
+      </div>
+    );
+  }
+
+  const selectedName = value.length === 1
+    ? rows.find((item) => item.id === value[0])?.name
+    : null;
+
+  return (
+    <DropdownMenu open={open} onOpenChange={setOpen}>
+      <DropdownMenuTrigger asChild>
+        <Button
+          type="button"
+          variant="ghost"
+          size="sm"
+          className={cn(
+            "h-8 max-w-52 gap-1.5 rounded-full border-0 bg-transparent px-2.5 text-muted-foreground shadow-none hover:bg-muted/70 hover:text-foreground",
+            value.length > 0 && "bg-primary/8 text-primary hover:bg-primary/12 hover:text-primary",
+            className,
+          )}
+          disabled={disabled || loading}
+        >
+          {loading ? <Loader2 className="size-3.5 animate-spin" /> : <Database className="size-3.5" />}
+          <span className="truncate text-xs font-normal">{selectedName || label}</span>
+          {value.length > 0 && (
+            <span className="flex size-4 items-center justify-center rounded-full bg-primary text-[9px] font-semibold text-primary-foreground">
+              {value.length}
+            </span>
+          )}
+          {value.length > 0 && value.length === rows.length && <Check className="size-3" />}
+        </Button>
+      </DropdownMenuTrigger>
+      <DropdownMenuContent align="start" sideOffset={8} className="w-80 overflow-hidden rounded-xl p-0 shadow-xl">
+        {content}
+      </DropdownMenuContent>
+    </DropdownMenu>
+  );
+}
+
+export default LlmWikiKnowledgeSelector;

frontend-web/src/components/workspace/skill-display-sources.tsx

@@ -92,7 +92,7 @@ export function SourceCard({
       knowledge_base_id: previewTarget.knowledgeBaseId,
       chunk_id: previewTarget.chunkId,
     });
-    void apiFetch(`${getBackendBaseURL()}/api/weknora/source-context?${params}`)
+    void apiFetch(`${getBackendBaseURL()}/api/llmwiki/sources/context?${params}`)
       .then(async (response) => {
         const payload = (await response.json().catch(() => null)) as unknown;
         if (!response.ok || !isWeKnoraSourceContextResponse(payload)) {

frontend-web/src/core/agents/types.ts

@@ -61,7 +61,9 @@ export interface Agent {
   /** 会商席位「技能识别强化」开关。作为会商席位运行时，是否把本智能体配置的技能
    * (名称 + SKILL.md 路径 + 必须先 read_file 的硬指令) 拼进本轮任务以补偿弱模型。
    * 与业务链条同名开关取 OR。null/false = 关 (默认)。 */
-  seat_skill_directive?: boolean | null;
+  seat_skill_directive?: boolean | null;
+  /** DeerFlow LLMWiki mapping ids this agent allows conversations to select. */
+  llmwiki_knowledge_base_ids?: string[] | null;
   tags?: Tag[];
   featured_order?: number | null;
   /** Whether the current user has favorited (收藏) this agent. */
@@ -96,7 +98,8 @@ export interface CreateAgentRequest {
   common_questions?: AgentQuestion[] | null;
   template_name?: string | null;
   /** 会商席位「技能识别强化」开关；缺省/false = 关。 */
-  seat_skill_directive?: boolean | null;
+  seat_skill_directive?: boolean | null;
+  llmwiki_knowledge_base_ids?: string[] | null;
 }
 
 export interface UpdateAgentRequest {
@@ -113,5 +116,6 @@ export interface UpdateAgentRequest {
   /** 入库模板名；传空字符串表示清除。 */
   template_name?: string | null;
   /** 会商席位「技能识别强化」开关；缺省/false = 关。 */
-  seat_skill_directive?: boolean | null;
-}
+  seat_skill_directive?: boolean | null;
+  llmwiki_knowledge_base_ids?: string[] | null;
+}

frontend-web/src/core/llmwiki/conversation-deposit.ts

@@ -0,0 +1,66 @@
+import type { Message } from "@langchain/langgraph-sdk";
+
+import { textOfMessage } from "@/core/threads/utils";
+import { createLlmWikiConversationDeposit } from "@/strategy-components/api/llmwiki";
+
+function isHidden(message: Message): boolean {
+  const extra = (message as { additional_kwargs?: Record<string, unknown> })
+    .additional_kwargs;
+  return extra?.hide_from_ui === true;
+}
+
+function messageText(message: Message | undefined): string {
+  return (message ? textOfMessage(message) : "")?.trim() ?? "";
+}
+
+function hasText(message: Message | undefined): message is Message {
+  return !!message && messageText(message).length > 0;
+}
+
+export function queueLlmWikiConversationDeposit({
+  threadId,
+  messages,
+  assistantId,
+  knowledgeBaseIds,
+}: {
+  threadId: string | null | undefined;
+  messages: Message[] | undefined;
+  assistantId?: string | null;
+  knowledgeBaseIds?: string[];
+}) {
+  if (!threadId || !messages?.length) {
+    return;
+  }
+  const visible = messages.filter((message) => !isHidden(message));
+  const assistantIndex = [...visible]
+    .map((message, index) => ({ message, index }))
+    .reverse()
+    .find(({ message }) => message.type === "ai" && hasText(message))?.index;
+  if (assistantIndex === undefined) {
+    return;
+  }
+  const assistantMessage = visible[assistantIndex];
+  const humanMessage = [...visible.slice(0, assistantIndex)]
+    .reverse()
+    .find((message) => message.type === "human" && hasText(message));
+  if (!assistantMessage?.id || !humanMessage) {
+    return;
+  }
+  const question = messageText(humanMessage);
+  const answer = messageText(assistantMessage);
+  if (!question || !answer) {
+    return;
+  }
+
+  void createLlmWikiConversationDeposit({
+    thread_id: threadId,
+    question,
+    answer,
+    human_message_id: humanMessage.id ?? null,
+    assistant_message_id: assistantMessage.id,
+    assistant_id: assistantId ?? null,
+    knowledge_base_ids: knowledgeBaseIds ?? [],
+  }).catch((error) => {
+    console.warn("Failed to queue LLMWiki conversation deposit", error);
+  });
+}

frontend-web/src/core/threads/hooks.ts

@@ -1122,20 +1122,26 @@ export function useThreadStream({
           }),
         );
 
-        const ragModeEnabled =
-          localSettings.display.ragModeEnabled || isThreadRagModeLocked(threadId);
-        if (ragModeEnabled) {
-          lockThreadRagMode(threadId);
-        }
+        const mergedContext = {
+          ...context,
+          ...extraContext,
+        };
+        const selectedLlmWikiIds = Array.isArray(
+          mergedContext.llmwiki_knowledge_base_ids,
+        )
+          ? mergedContext.llmwiki_knowledge_base_ids
+          : [];
+        const ragModeEnabled =
+          localSettings.display.ragModeEnabled ||
+          isThreadRagModeLocked(threadId) ||
+          selectedLlmWikiIds.length > 0;
+        if (ragModeEnabled) {
+          lockThreadRagMode(threadId);
+        }
 
         const contextCompressionEnabled =
           localSettings.display.contextCompressionEnabled ?? false;
 
-        const mergedContext = {
-          ...context,
-          ...extraContext,
-        };
-
         const hiddenContextMessages = normalizeHiddenContextMessages(options);
         const submittedMessages: Message[] = [
           ...hiddenContextMessages.map((item) => ({

frontend-web/src/core/threads/types.ts

@@ -24,7 +24,9 @@ export interface AgentThreadContext extends Record<string, unknown> {
   /** 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;
+  knowledge_rag_enabled?: boolean;
+  /** DeerFlow LLMWiki mapping ids selected for this conversation. */
+  llmwiki_knowledge_base_ids?: string[];
   agent_id?: string;
   /** @deprecated kept for reading legacy thread metadata; prefer agent_id */
   agent_name?: string;

frontend-web/src/pages/AgentChatPage.tsx

@@ -16,6 +16,7 @@ import { ArtifactTrigger } from "@/components/workspace/artifacts";
 import { ChatBox, useThreadChat } from "@/components/workspace/chats";
 import { ExportTrigger } from "@/components/workspace/export-trigger";
 import { InputBox } from "@/components/workspace/input-box";
+import { LlmWikiKnowledgeSelector } from "@/components/workspace/llmwiki-knowledge-selector";
 import { MessageList } from "@/components/workspace/messages";
 import { ThreadContext } from "@/components/workspace/messages/context";
 import { QuestionNavigator } from "@/components/workspace/question-navigator";
@@ -24,7 +25,9 @@ import { TodoList } from "@/components/workspace/todo-list";
 import { TokenUsageIndicator } from "@/components/workspace/token-usage-indicator";
 import { Tooltip } from "@/components/workspace/tooltip";
 import { normalizeAgentQuestions, useAgent } from "@/core/agents";
+import { getAPIClient } from "@/core/api";
 import { useI18n } from "@/core/i18n/hooks";
+import { queueLlmWikiConversationDeposit } from "@/core/llmwiki/conversation-deposit";
 import { useModels } from "@/core/models/hooks";
 import { useNotification } from "@/core/notification/hooks";
 import { saveLastAgentId, useThreadSettings } from "@/core/settings";
@@ -69,6 +72,26 @@ const TEMPLATE_BUILDER_AGENT_ID = "template-json-builder";
 const RWFX_SQ_REPORT_CONTEXT_MAX_CHARS = 12000;
 const RWFX_SQ_REPORT_CONTEXT_TAIL_CHARS = 2000;
 
+function readPendingKnowledgeSelection(agentId: string | undefined): string[] | null {
+  if (typeof window === "undefined" || !agentId) return null;
+  try {
+    const raw = sessionStorage.getItem("pendingAgentMessage");
+    if (!raw) return null;
+    const parsed = JSON.parse(raw) as {
+      agentId?: string;
+      knowledgeBaseIds?: unknown;
+    };
+    if (parsed.agentId !== agentId || !Array.isArray(parsed.knowledgeBaseIds)) {
+      return null;
+    }
+    return parsed.knowledgeBaseIds
+      .map((item) => String(item).trim())
+      .filter(Boolean);
+  } catch {
+    return null;
+  }
+}
+
 function stringifyForHiddenContext(value: unknown): string {
   if (typeof value === "string") return value;
   try {
@@ -244,6 +267,79 @@ export default function AgentChatPage({
     useThreadChat();
   const [settings, setSettings] = useThreadSettings(threadId);
   const { tokenUsageEnabled } = useModels();
+  const pendingKnowledgeSelection = useRef(
+    readPendingKnowledgeSelection(agent_id),
+  );
+  const [selectedKnowledgeBaseIds, setSelectedKnowledgeBaseIds] = useState<string[]>(
+    () => pendingKnowledgeSelection.current ?? [],
+  );
+  const [knowledgeSelectionReady, setKnowledgeSelectionReady] = useState(
+    () => pendingKnowledgeSelection.current !== null,
+  );
+  const [knowledgeSelectionLoading, setKnowledgeSelectionLoading] = useState(false);
+  const knowledgeSelectionThread = useRef<string | null>(null);
+  const knowledgeSelectionAgent = useRef(agent_id);
+  const activeThreadIdRef = useRef<string | null | undefined>(threadId);
+
+  const changeKnowledgeSelection = useCallback((ids: string[]) => {
+    setSelectedKnowledgeBaseIds(ids);
+    setKnowledgeSelectionReady(true);
+  }, []);
+
+  useEffect(() => {
+    activeThreadIdRef.current = threadId;
+  }, [threadId]);
+
+  useEffect(() => {
+    if (knowledgeSelectionAgent.current === agent_id) return;
+    knowledgeSelectionAgent.current = agent_id;
+    const pending = readPendingKnowledgeSelection(agent_id);
+    pendingKnowledgeSelection.current = pending;
+    knowledgeSelectionThread.current = null;
+    setSelectedKnowledgeBaseIds(pending ?? []);
+    setKnowledgeSelectionReady(pending !== null);
+  }, [agent_id]);
+
+  useEffect(() => {
+    if (isNewThread) {
+      if (knowledgeSelectionThread.current !== "__new__") {
+        setSelectedKnowledgeBaseIds(pendingKnowledgeSelection.current ?? []);
+        setKnowledgeSelectionReady(pendingKnowledgeSelection.current !== null);
+      }
+      knowledgeSelectionThread.current = "__new__";
+      setKnowledgeSelectionLoading(false);
+      return;
+    }
+    if (!threadId || knowledgeSelectionThread.current === threadId) return;
+    knowledgeSelectionThread.current = threadId;
+    setKnowledgeSelectionLoading(true);
+    void getAPIClient()
+      .threads.get(threadId)
+      .then((record) => {
+        if (knowledgeSelectionThread.current !== threadId) return;
+        const metadata = record.metadata as Record<string, unknown> | undefined;
+        const raw = metadata?.llmwiki_knowledge_base_ids;
+        if (Array.isArray(raw)) {
+          setSelectedKnowledgeBaseIds(
+            raw.map((item) => String(item).trim()).filter(Boolean),
+          );
+          setKnowledgeSelectionReady(true);
+        } else {
+          setSelectedKnowledgeBaseIds([]);
+          setKnowledgeSelectionReady(false);
+        }
+      })
+      .catch(() => {
+        if (knowledgeSelectionThread.current === threadId) {
+          setKnowledgeSelectionReady(false);
+        }
+      })
+      .finally(() => {
+        if (knowledgeSelectionThread.current === threadId) {
+          setKnowledgeSelectionLoading(false);
+        }
+      });
+  }, [isNewThread, threadId]);
 
   useEffect(() => {
     if (isNewThread) {
@@ -253,14 +349,34 @@ export default function AgentChatPage({
   }, [isNewThread, threadId, agent_id]);
 
   const { showNotification } = useNotification();
+  const conversationMetadata = useMemo(
+    () => ({
+      ...(taskId ? taskThreadMetadata(taskContext) : {}),
+      ...(knowledgeSelectionReady
+        ? { llmwiki_knowledge_base_ids: selectedKnowledgeBaseIds }
+        : {}),
+    }),
+    [knowledgeSelectionReady, selectedKnowledgeBaseIds, taskContext, taskId],
+  );
   const [thread, sendMessage] = useThreadStream({
     threadId: isNewThread ? undefined : threadId,
     assistantId: agent_id,
-    context: { ...settings.context, agent_id },
+    context: {
+      ...settings.context,
+      agent_id,
+      ...(knowledgeSelectionReady
+        ? { llmwiki_knowledge_base_ids: selectedKnowledgeBaseIds }
+        : {}),
+    },
     // 任务闸门：新建 thread 打 metadata.taskId 标签，供任务对话列表(1:N)按任务过滤；
     // xdfx 额外打 metadata.actionId 标签，按「行动id」划分聊天记录（见 taskThreadMetadata）。
-    threadMetadata: taskId ? taskThreadMetadata(taskContext) : undefined,
+    threadMetadata:
+      Object.keys(conversationMetadata).length > 0
+        ? conversationMetadata
+        : undefined,
     onStart: (createdThreadId) => {
+      knowledgeSelectionThread.current = createdThreadId;
+      activeThreadIdRef.current = createdThreadId;
       setThreadId(createdThreadId);
       setIsNewThread(false);
       // Keep the current component mounted while updating the hash route.
@@ -286,6 +402,12 @@ export default function AgentChatPage({
       } catch {
         /* 上报绝不影响问答 */
       }
+      queueLlmWikiConversationDeposit({
+        threadId: activeThreadIdRef.current,
+        messages: state.messages,
+        assistantId: agent_id,
+        knowledgeBaseIds: selectedKnowledgeBaseIds,
+      });
       if (document.hidden || !document.hasFocus()) {
         let body = "Conversation finished";
         const lastMessage = state.messages[state.messages.length - 1];
@@ -581,10 +703,23 @@ export default function AgentChatPage({
                         autoFocus
                         status={status}
                         context={settings.context}
-                        disabled={isDemoMode}
+                        disabled={isDemoMode || knowledgeSelectionLoading}
                         onContextChange={(context) =>
                           setSettings("context", context)
                         }
+                        knowledgeSelector={
+                          <LlmWikiKnowledgeSelector
+                            source="agent"
+                            agentId={agent_id}
+                            value={selectedKnowledgeBaseIds}
+                            onChange={changeKnowledgeSelection}
+                            autoSelectAll={
+                              isNewThread && pendingKnowledgeSelection.current === null
+                            }
+                            disabled={isDemoMode || knowledgeSelectionLoading}
+                            label="知识空间"
+                          />
+                        }
                         onFollowupsVisibilityChange={setShowFollowups}
                         onSubmit={handleSubmit}
                         onStop={handleStop}
@@ -636,7 +771,7 @@ export default function AgentChatPage({
                         assistantId={agent_id}
                         status={status}
                         context={settings.context}
-                        disabled={isDemoMode}
+                        disabled={isDemoMode || knowledgeSelectionLoading}
                         onContextChange={(context) =>
                           setSettings("context", context)
                         }

frontend-web/src/pages/ChatPage.tsx

@@ -16,6 +16,7 @@ import {
 } from "@/components/workspace/chats";
 import { ExportTrigger } from "@/components/workspace/export-trigger";
 import { KnowledgeSaveTrigger } from "@/components/workspace/knowledge-save-trigger";
+import { LlmWikiKnowledgeSelector } from "@/components/workspace/llmwiki-knowledge-selector";
 import { InputBox } from "@/components/workspace/input-box";
 import {
   MessageList,
@@ -52,6 +53,7 @@ import {
   useThreadSettings,
 } from "@/core/settings";
 import { useThreadStream } from "@/core/threads/hooks";
+import { queueLlmWikiConversationDeposit } from "@/core/llmwiki/conversation-deposit";
 import { useLayoutMode } from "@/core/system-settings/hooks";
 import { useRouteBase } from "@/core/route-base";
 import { textOfMessage } from "@/core/threads/utils";
@@ -121,12 +123,29 @@ export default function ChatPage() {
   // In-memory only — explicitly overrides any value persisted by the old
   // per-thread header toggle so stale localStorage can't re-enable it.
   const [knowledgeRagEnabled] = useKnowledgeRagEnabled();
+  const [selectedKnowledgeBaseIds, setSelectedKnowledgeBaseIds] = useState<string[]>([]);
+  const [knowledgeSelectionReady, setKnowledgeSelectionReady] = useState(false);
+  const [knowledgeSelectionLoading, setKnowledgeSelectionLoading] = useState(false);
+  const knowledgeSelectionThread = useRef<string | null>(null);
+  const activeThreadIdRef = useRef<string | null | undefined>(threadId);
+  const changeKnowledgeSelection = useCallback((ids: string[]) => {
+    setSelectedKnowledgeBaseIds(ids);
+    setKnowledgeSelectionReady(true);
+  }, []);
   const streamContext = useMemo(
     () => ({
       ...settings.context,
       knowledge_rag_enabled: knowledgeRagEnabled,
+      ...(knowledgeSelectionReady
+        ? { llmwiki_knowledge_base_ids: selectedKnowledgeBaseIds }
+        : {}),
     }),
-    [settings.context, knowledgeRagEnabled],
+    [
+      settings.context,
+      knowledgeRagEnabled,
+      knowledgeSelectionReady,
+      selectedKnowledgeBaseIds,
+    ],
   );
   const [mounted, setMounted] = useState(false);
   const { tokenUsageEnabled } = useModels();
@@ -150,6 +169,10 @@ export default function ChatPage() {
   const { selectedArtifact } = useArtifacts();
   useSpecificChatMode();
 
+  useEffect(() => {
+    activeThreadIdRef.current = threadId;
+  }, [threadId]);
+
   useEffect(() => {
     setMounted(true);
   }, []);
@@ -158,12 +181,54 @@ export default function ChatPage() {
     if (isNewThread) {
       setHasStartedConversation(false);
       setPromptPrefixOn(false);
+      setSelectedKnowledgeBaseIds([]);
+      setKnowledgeSelectionReady(false);
+      setKnowledgeSelectionLoading(false);
+      knowledgeSelectionThread.current = "__new__";
       // Pre-select the agent the user last started a conversation with so a
       // fresh chat defaults to it. Validated against the known agent list below.
       setSelectedAgentId(getLastAgentId());
     }
   }, [isNewThread, threadId]);
 
+  // The knowledge-base choice belongs to the conversation, not to a browser
+  // component instance. Restore it from LangGraph thread metadata whenever an
+  // existing conversation is opened. An explicit empty array means the user
+  // deliberately disabled LLMWiki retrieval for that conversation.
+  useEffect(() => {
+    if (isNewThread || !threadId || knowledgeSelectionThread.current === threadId) {
+      return;
+    }
+    knowledgeSelectionThread.current = threadId;
+    setKnowledgeSelectionLoading(true);
+    void getAPIClient(isMock)
+      .threads.get(threadId)
+      .then((record) => {
+        if (knowledgeSelectionThread.current !== threadId) return;
+        const metadata = record.metadata as Record<string, unknown> | undefined;
+        const raw = metadata?.llmwiki_knowledge_base_ids;
+        if (Array.isArray(raw)) {
+          setSelectedKnowledgeBaseIds(
+            raw.map((item) => String(item).trim()).filter(Boolean),
+          );
+          setKnowledgeSelectionReady(true);
+        } else {
+          setSelectedKnowledgeBaseIds([]);
+          setKnowledgeSelectionReady(false);
+        }
+      })
+      .catch(() => {
+        if (knowledgeSelectionThread.current === threadId) {
+          setKnowledgeSelectionReady(false);
+        }
+      })
+      .finally(() => {
+        if (knowledgeSelectionThread.current === threadId) {
+          setKnowledgeSelectionLoading(false);
+        }
+      });
+  }, [isMock, isNewThread, threadId]);
+
   // Drop a restored last-used agent that is no longer in the selector list
   // (deleted / unpublished / no longer curated), so we never hand off to a
   // dead agent. Wait for the list to load before deciding.
@@ -175,6 +240,8 @@ export default function ChatPage() {
     if (known.length === 0) return;
     if (!known.some((a) => a.id === selectedAgentId)) {
       setSelectedAgentId(null);
+      setSelectedKnowledgeBaseIds([]);
+      setKnowledgeSelectionReady(false);
     }
   }, [selectedAgentId, selectorAgents, mineAgents, squareAgents, selectorLoading]);
 
@@ -183,8 +250,13 @@ export default function ChatPage() {
   const [thread, sendMessage, isUploading] = useThreadStream({
     threadId: isNewThread ? undefined : threadId,
     context: streamContext,
+    threadMetadata: knowledgeSelectionReady
+      ? { llmwiki_knowledge_base_ids: selectedKnowledgeBaseIds }
+      : undefined,
     isMock,
     onStart: (createdThreadId) => {
+      knowledgeSelectionThread.current = createdThreadId;
+      activeThreadIdRef.current = createdThreadId;
       setThreadId(createdThreadId);
       setIsNewThread(false);
       // Keep the current component mounted while updating the hash route.
@@ -210,6 +282,12 @@ export default function ChatPage() {
       } catch {
         /* 上报绝不影响问答 */
       }
+      queueLlmWikiConversationDeposit({
+        threadId: activeThreadIdRef.current,
+        messages: state.messages,
+        assistantId: selectedAgentId,
+        knowledgeBaseIds: selectedKnowledgeBaseIds,
+      });
       if (document.hidden || !document.hasFocus()) {
         let body = "Conversation finished";
         const lastMessage = state.messages.at(-1);
@@ -262,7 +340,12 @@ export default function ChatPage() {
         void (async () => {
           try {
             const created = await getAPIClient(isMock).threads.create({
-              metadata: { agent_id: selectedAgentId },
+              metadata: {
+                agent_id: selectedAgentId,
+                ...(knowledgeSelectionReady
+                  ? { llmwiki_knowledge_base_ids: selectedKnowledgeBaseIds }
+                  : {}),
+              },
             });
             sessionStorage.setItem(
               "pendingAgentMessage",
@@ -270,6 +353,9 @@ export default function ChatPage() {
                 agentId: selectedAgentId,
                 threadId: created.thread_id,
                 text,
+                ...(knowledgeSelectionReady
+                  ? { knowledgeBaseIds: selectedKnowledgeBaseIds }
+                  : {}),
               }),
             );
             saveLastAgentId(selectedAgentId);
@@ -328,6 +414,8 @@ export default function ChatPage() {
       hideAdminPrompt,
       preferencesLoading,
       selectedAgentId,
+      selectedKnowledgeBaseIds,
+      knowledgeSelectionReady,
       agentHandoffPending,
       isMock,
       router,
@@ -418,6 +506,8 @@ export default function ChatPage() {
   // Passing null (the ✕ clear) wipes the stored preference.
   const handleAgentChange = useCallback((agentId: string | null) => {
     setSelectedAgentId(agentId);
+    setSelectedKnowledgeBaseIds([]);
+    setKnowledgeSelectionReady(false);
     saveLastAgentId(agentId);
   }, []);
   const handleStop = useCallback(async () => {
@@ -620,7 +710,8 @@ export default function ChatPage() {
                       disabled={
                         env.NEXT_PUBLIC_STATIC_WEBSITE_ONLY === "true" ||
                         isUploading ||
-                        agentHandoffPending
+                        agentHandoffPending ||
+                        knowledgeSelectionLoading
                       }
                       onContextChange={(context) =>
                         setSettings("context", context)
@@ -634,6 +725,22 @@ export default function ChatPage() {
                             }
                           : undefined
                       }
+                      knowledgeSelector={
+                        isLandingState ? (
+                          <LlmWikiKnowledgeSelector
+                            source={selectedAgentId ? "agent" : "all"}
+                            agentId={selectedAgentId}
+                            value={selectedKnowledgeBaseIds}
+                            onChange={changeKnowledgeSelection}
+                            autoSelectAll={!!selectedAgentId}
+                            disabled={
+                              env.NEXT_PUBLIC_STATIC_WEBSITE_ONLY === "true" ||
+                              knowledgeSelectionLoading
+                            }
+                            label="知识空间"
+                          />
+                        ) : undefined
+                      }
                       onFollowupsVisibilityChange={setShowFollowups}
                       onSubmit={handleSubmit}
                       onStop={handleStop}

frontend-web/src/pages/KnowledgeBasePage.tsx

@@ -1,5 +1,5 @@
 import { Loader2 } from "lucide-react";
-import { useEffect, useMemo, useRef, useState } from "react";
+import { type ReactNode, useEffect, useMemo, useRef, useState } from "react";
 import { useNavigate, useParams } from "react-router-dom";
 import { toast } from "sonner";
 
@@ -530,7 +530,11 @@ function VaultPageDialog({
   );
 }
 
-export default function KnowledgeBasePage() {
+export default function KnowledgeBasePage({
+  headerExtraActions,
+}: {
+  headerExtraActions?: ReactNode;
+} = {}) {
   const navigate = useNavigate();
   const { noteId } = useParams<{ noteId: string }>();
 
@@ -653,6 +657,7 @@ export default function KnowledgeBasePage() {
         onCreate={() => setCreateOpen(true)}
         onManageFolders={() => setFoldersOpen(true)}
         onManageTemplates={() => setTemplatesOpen(true)}
+        extraActions={headerExtraActions}
       />
 
       <div

frontend-web/src/pages/LlmWikiKnowledgePage.tsx

@@ -0,0 +1,91 @@
+import { Loader2, Settings2 } from "lucide-react";
+import { useState } from "react";
+import { Navigate } from "react-router-dom";
+
+import { Button } from "@/components/ui/button";
+import {
+  Dialog,
+  DialogContent,
+  DialogHeader,
+  DialogTitle,
+} from "@/components/ui/dialog";
+import { getStoredAuth } from "@/core/auth";
+import { useLlmWikiRuntime } from "@/strategy-components/api/llmwiki";
+import WeKnoraKnowledgeWorkspace, {
+  ConnectionSettings,
+} from "@/strategy-components/components/resource-management/llmwiki/WeKnoraKnowledgeWorkspace";
+import WeKnoraKnowledgeDetailPage from "@/strategy-components/components/resource-management/llmwiki/WeKnoraKnowledgeDetailPage";
+
+import KnowledgeBasePage from "./KnowledgeBasePage";
+import KnowledgeGraphPage from "./KnowledgeGraphPage";
+
+type LlmWikiKnowledgePageMode = "list" | "detail" | "graph";
+
+function RuntimeLoading() {
+  return (
+    <div className="flex h-full min-h-64 items-center justify-center text-sm text-muted-foreground">
+      <Loader2 className="mr-2 size-5 animate-spin" />
+      正在读取 LLMWiki 运行模式
+    </div>
+  );
+}
+
+function AdminWeKnoraSettingsButton() {
+  const [open, setOpen] = useState(false);
+
+  return (
+    <>
+      <Button size="sm" variant="outline" onClick={() => setOpen(true)}>
+        <Settings2 className="mr-2 size-4" />
+        配置 WeKnora
+      </Button>
+      <Dialog open={open} onOpenChange={setOpen}>
+        <DialogContent className="max-w-4xl">
+          <DialogHeader>
+            <DialogTitle>LLMWiki 运行模式</DialogTitle>
+          </DialogHeader>
+          <ConnectionSettings />
+        </DialogContent>
+      </Dialog>
+    </>
+  );
+}
+
+function LegacyKnowledgePage() {
+  const isAdmin = getStoredAuth()?.system_role === "admin";
+
+  return (
+    <KnowledgeBasePage
+      headerExtraActions={isAdmin ? <AdminWeKnoraSettingsButton /> : undefined}
+    />
+  );
+}
+
+export default function LlmWikiKnowledgePage({
+  mode = "list",
+}: {
+  mode?: LlmWikiKnowledgePageMode;
+}) {
+  const runtime = useLlmWikiRuntime();
+
+  if (runtime.isLoading) {
+    return <RuntimeLoading />;
+  }
+
+  if (runtime.isError || runtime.data?.provider !== "weknora") {
+    if (mode === "graph") {
+      return <KnowledgeGraphPage />;
+    }
+    return <LegacyKnowledgePage />;
+  }
+
+  if (mode === "graph") {
+    return <Navigate to="/page/workspace/knowledge" replace />;
+  }
+
+  if (mode === "detail") {
+    return <WeKnoraKnowledgeDetailPage />;
+  }
+
+  return <WeKnoraKnowledgeWorkspace />;
+}

frontend-web/src/pages/NewAgentPage.tsx

@@ -31,6 +31,7 @@ import {
 } from "@/components/ui/select";
 import { Textarea } from "@/components/ui/textarea";
 import { ArtifactsProvider } from "@/components/workspace/artifacts";
+import { LlmWikiKnowledgeSelector } from "@/components/workspace/llmwiki-knowledge-selector";
 import {
   MessageList,
   type SetupAgentApprovalPayload,
@@ -95,6 +96,7 @@ export default function NewAgentPage() {
   const [isCreatingAgent, setIsCreatingAgent] = useState(false);
   const [isSavingDraft, setIsSavingDraft] = useState(false);
   const [selectedSkills, setSelectedSkills] = useState<string[]>([]);
+  const [selectedKnowledgeBaseIds, setSelectedKnowledgeBaseIds] = useState<string[]>([]);
   const [isSkillSelectionTouched, setIsSkillSelectionTouched] = useState(false);
   const [detailSkillName, setDetailSkillName] = useState<string | null>(null);
   // 恢复已有 bootstrap 会话时该 thread 已存在，无需再 create（置为 true，跳过创建）。
@@ -230,6 +232,7 @@ export default function NewAgentPage() {
           model: modelOverride,
           template_name: templateNameInput.trim(),
           seat_skill_directive: seatSkillDirectiveInput,
+          llmwiki_knowledge_base_ids: selectedKnowledgeBaseIds,
         });
         setAgent(updated);
         return updated;
@@ -252,6 +255,7 @@ export default function NewAgentPage() {
         model: modelOverride,
         template_name: templateNameInput.trim(),
         seat_skill_directive: seatSkillDirectiveInput,
+        llmwiki_knowledge_base_ids: selectedKnowledgeBaseIds,
       });
       setAgentId(created.id);
       setAgent(created);
@@ -280,6 +284,7 @@ export default function NewAgentPage() {
     descriptionInput,
     modelInput,
     selectedSkills,
+    selectedKnowledgeBaseIds,
     soulInput,
     templateNameInput,
     seatSkillDirectiveInput,
@@ -590,6 +595,13 @@ export default function NewAgentPage() {
                   </p>
                 </div>
 
+                <LlmWikiKnowledgeSelector
+                  source="all"
+                  value={selectedKnowledgeBaseIds}
+                  onChange={setSelectedKnowledgeBaseIds}
+                  variant="panel"
+                />
+
                 <div className="space-y-2">
                   <label className="text-sm font-medium">SOUL.md</label>
                   <Textarea

frontend-web/src/pages/WorkspaceRoutes.tsx

@@ -35,8 +35,7 @@ import AdminScheduledTasksPage from './AdminScheduledTasksPage';
 import AdminSkillsPage from './AdminSkillsPage';
 import ScheduledTaskRunDetailPage from './ScheduledTaskRunDetailPage';
 import ScheduledTasksPage from './ScheduledTasksPage';
-import KnowledgeBasePage from './KnowledgeBasePage';
-import KnowledgeGraphPage from './KnowledgeGraphPage';
+import LlmWikiKnowledgePage from './LlmWikiKnowledgePage';
 import LlmMetricsPage from './LlmMetricsPage';
 import CheckpointMigrationPage from './CheckpointMigrationPage';
 import MemoryPage from './MemoryPage';
@@ -604,7 +603,7 @@ export default function WorkspaceRoutes({
           path="knowledge"
           element={
             <WorkspaceLayout {...layoutProps}>
-              <KnowledgeBasePage />
+              <LlmWikiKnowledgePage />
             </WorkspaceLayout>
           }
         />
@@ -612,7 +611,7 @@ export default function WorkspaceRoutes({
           path="knowledge/graph"
           element={
             <WorkspaceLayout {...layoutProps}>
-              <KnowledgeGraphPage />
+              <LlmWikiKnowledgePage mode="graph" />
             </WorkspaceLayout>
           }
         />
@@ -620,7 +619,7 @@ export default function WorkspaceRoutes({
           path="knowledge/:noteId"
           element={
             <WorkspaceLayout {...layoutProps}>
-              <KnowledgeBasePage />
+              <LlmWikiKnowledgePage mode="detail" />
             </WorkspaceLayout>
           }
         />

frontend-web/src/strategy-components/api/llmwiki.ts

@@ -0,0 +1,689 @@
+import { useMutation, useQuery, useQueryClient } from "@tanstack/react-query";
+
+import { apiFetch } from "@/core/api/fetch-client";
+import { getBackendBaseURL } from "@/core/config";
+
+const API = () => `${getBackendBaseURL()}/api/llmwiki`;
+const SETTINGS_API = () => `${getBackendBaseURL()}/api/system-settings/llmwiki`;
+
+export type LlmWikiProvider = "legacy" | "weknora";
+export type LlmWikiScope = "personal" | "public" | "all";
+
+export interface LlmWikiRuntime {
+  provider: LlmWikiProvider;
+  api_base_url: string;
+  web_base_url: string;
+  override_enabled: boolean;
+}
+
+export interface LlmWikiSettings extends LlmWikiRuntime {
+  file_api_base_url: string;
+  file_web_base_url: string;
+  effective_provider: LlmWikiProvider;
+}
+
+export interface LlmWikiKnowledgeBase {
+  id: string;
+  remote_status: "available" | "missing";
+  name: string;
+  description: string;
+  type: "document" | "faq";
+  publication_status: "private" | "pending" | "published" | "rejected";
+  owner_user_id?: string | null;
+  is_owner: boolean;
+  can_write: boolean;
+  knowledge_count: number;
+  chunk_count: number;
+  processing_count: number;
+  embedding_model_id?: string | null;
+  vector_store_id?: string | null;
+  vector_store_name?: string | null;
+  vector_store_source?: string | null;
+  vector_store_engine_type?: string | null;
+  vector_store_status?: string | null;
+  chunking_config?: Record<string, unknown>;
+  indexing_strategy?: Record<string, boolean>;
+  capabilities?: Record<string, boolean>;
+  extract_config?: Record<string, unknown>;
+  wiki_config?: Record<string, unknown>;
+  question_generation_config?: Record<string, unknown>;
+  summary_model_id?: string | null;
+  storage_provider?: string | null;
+  created_at?: string | null;
+  updated_at?: string | null;
+  published_at?: string | null;
+}
+
+export interface LlmWikiDocument {
+  id: string;
+  type?: string;
+  title?: string;
+  description?: string;
+  source?: string;
+  channel?: string;
+  file_name?: string;
+  file_type?: string;
+  file_size?: number;
+  storage_size?: number;
+  parse_status?: string;
+  pending_subtasks_count?: number;
+  summary_status?: string;
+  enable_status?: string;
+  embedding_model_id?: string;
+  created_at?: string;
+  updated_at?: string;
+  processed_at?: string;
+  error_message?: string;
+}
+
+export interface LlmWikiChunk {
+  id: string;
+  content: string;
+  chunk_index: number;
+  is_enabled: boolean;
+  status?: number | string | null;
+  start_at: number;
+  end_at: number;
+  chunk_type: string;
+  parent_chunk_id?: string;
+  relation_chunk_ids: string[];
+  indirect_relation_chunk_ids: string[];
+  metadata: Record<string, unknown>;
+  created_at?: string;
+  updated_at?: string;
+}
+
+export interface LlmWikiChunkPage {
+  items: LlmWikiChunk[];
+  total: number;
+  page: number;
+  page_size: number;
+}
+
+export interface LlmWikiGraph {
+  nodes: Array<{ id: string; title: string; type: string; link_count: number }>;
+  edges: Array<{ source: string; target: string }>;
+  meta: {
+    mode?: string;
+    total?: number;
+    returned?: number;
+    truncated?: boolean;
+  };
+}
+
+export interface LlmWikiWikiPage {
+  id: string;
+  slug: string;
+  title: string;
+  page_type: string;
+  status: string;
+  content: string;
+  summary: string;
+  aliases: string[];
+  parent_slug?: string;
+  category_path?: string[];
+  depth?: number;
+  source_refs: string[];
+  in_links: string[];
+  out_links: string[];
+  version: number;
+  created_at?: string;
+  updated_at?: string;
+}
+
+export interface LlmWikiWikiPageList {
+  pages: LlmWikiWikiPage[];
+  total: number;
+  page: number;
+  page_size: number;
+  total_pages: number;
+}
+
+export interface LlmWikiSearchResult {
+  chunk_id: string;
+  content: string;
+  knowledge_id: string;
+  knowledge_base_id: string;
+  knowledge_base_name: string;
+  title: string;
+  filename: string;
+  score: number;
+  chunk_index: number;
+  url: string;
+}
+
+async function readJson<T>(response: Response, context: string): Promise<T> {
+  if (!response.ok) {
+    const payload = await response.json().catch(() => null as unknown);
+    const detail =
+      payload && typeof payload === "object" && "detail" in payload
+        ? String((payload as { detail: unknown }).detail)
+        : `${context}失败（${response.status}）`;
+    throw new Error(detail);
+  }
+  if (response.status === 204) return undefined as T;
+  return (await response.json()) as T;
+}
+
+export async function getLlmWikiRuntime(): Promise<LlmWikiRuntime> {
+  return readJson(
+    await apiFetch(`${API()}/runtime`),
+    "读取 LLMWiki 运行模式",
+  );
+}
+
+export async function listLlmWikiKnowledgeBases(
+  scope: LlmWikiScope,
+): Promise<LlmWikiKnowledgeBase[]> {
+  const response = await apiFetch(
+    `${API()}/knowledge-bases?scope=${encodeURIComponent(scope)}`,
+  );
+  const payload = await readJson<{ knowledge_bases: LlmWikiKnowledgeBase[] }>(
+    response,
+    "读取知识库",
+  );
+  return payload.knowledge_bases ?? [];
+}
+
+export async function listAgentLlmWikiKnowledgeBases(
+  agentId: string,
+): Promise<LlmWikiKnowledgeBase[]> {
+  const payload = await readJson<{ knowledge_bases: LlmWikiKnowledgeBase[] }>(
+    await apiFetch(
+      `${API()}/agents/${encodeURIComponent(agentId)}/knowledge-bases`,
+    ),
+    "读取智能体知识库",
+  );
+  return payload.knowledge_bases ?? [];
+}
+
+export async function getLlmWikiKnowledgeBase(
+  id: string,
+): Promise<LlmWikiKnowledgeBase> {
+  return readJson(
+    await apiFetch(`${API()}/knowledge-bases/${encodeURIComponent(id)}`),
+    "读取知识库详情",
+  );
+}
+
+export async function createLlmWikiKnowledgeBase(input: {
+  name: string;
+  description: string;
+  type: "document" | "faq";
+  wiki_enabled?: boolean;
+}): Promise<LlmWikiKnowledgeBase> {
+  return readJson(
+    await apiFetch(`${API()}/knowledge-bases`, {
+      method: "POST",
+      headers: { "Content-Type": "application/json" },
+      body: JSON.stringify(input),
+    }),
+    "创建知识库",
+  );
+}
+
+export async function updateLlmWikiKnowledgeBase(
+  id: string,
+  input: { name?: string; description?: string },
+): Promise<LlmWikiKnowledgeBase> {
+  return readJson(
+    await apiFetch(`${API()}/knowledge-bases/${encodeURIComponent(id)}`, {
+      method: "PATCH",
+      headers: { "Content-Type": "application/json" },
+      body: JSON.stringify(input),
+    }),
+    "更新知识库",
+  );
+}
+
+export async function deleteLlmWikiKnowledgeBase(id: string): Promise<void> {
+  await readJson<void>(
+    await apiFetch(`${API()}/knowledge-bases/${encodeURIComponent(id)}`, {
+      method: "DELETE",
+    }),
+    "删除知识库",
+  );
+}
+
+export async function listLlmWikiDocuments(
+  knowledgeBaseId: string,
+): Promise<LlmWikiDocument[]> {
+  const payload = await readJson<{ items: LlmWikiDocument[] }>(
+    await apiFetch(
+      `${API()}/knowledge-bases/${encodeURIComponent(knowledgeBaseId)}/documents`,
+    ),
+    "读取文档",
+  );
+  return payload.items ?? [];
+}
+
+export async function getLlmWikiDocument(
+  knowledgeBaseId: string,
+  documentId: string,
+): Promise<LlmWikiDocument> {
+  return readJson(
+    await apiFetch(
+      `${API()}/knowledge-bases/${encodeURIComponent(knowledgeBaseId)}/documents/${encodeURIComponent(documentId)}`,
+    ),
+    "读取文档详情",
+  );
+}
+
+export async function listLlmWikiDocumentChunks(
+  knowledgeBaseId: string,
+  documentId: string,
+  page: number,
+  pageSize: number,
+): Promise<LlmWikiChunkPage> {
+  return readJson(
+    await apiFetch(
+      `${API()}/knowledge-bases/${encodeURIComponent(knowledgeBaseId)}/documents/${encodeURIComponent(documentId)}/chunks?page=${page}&page_size=${pageSize}`,
+    ),
+    "读取文档索引",
+  );
+}
+
+export async function getLlmWikiDocumentPreview(
+  knowledgeBaseId: string,
+  documentId: string,
+): Promise<Blob> {
+  const response = await apiFetch(
+    `${API()}/knowledge-bases/${encodeURIComponent(knowledgeBaseId)}/documents/${encodeURIComponent(documentId)}/preview`,
+  );
+  if (!response.ok) {
+    await readJson<void>(response, "读取文档原件");
+    throw new Error("读取文档原件失败");
+  }
+  return response.blob();
+}
+
+export async function getLlmWikiGraph(knowledgeBaseId: string): Promise<LlmWikiGraph> {
+  return readJson(
+    await apiFetch(`${API()}/knowledge-bases/${encodeURIComponent(knowledgeBaseId)}/graph?limit=200`),
+    "读取知识图谱",
+  );
+}
+
+export async function listLlmWikiWikiPages(
+  knowledgeBaseId: string,
+  page: number,
+  pageSize: number,
+  query = "",
+): Promise<LlmWikiWikiPageList> {
+  const params = new URLSearchParams({ page: String(page), page_size: String(pageSize) });
+  if (query.trim()) params.set("query", query.trim());
+  return readJson(
+    await apiFetch(
+      `${API()}/knowledge-bases/${encodeURIComponent(knowledgeBaseId)}/wiki/pages?${params.toString()}`,
+    ),
+    "读取 Wiki 页面",
+  );
+}
+
+export async function getLlmWikiWikiPage(
+  knowledgeBaseId: string,
+  slug: string,
+): Promise<LlmWikiWikiPage> {
+  const safeSlug = slug.split("/").map(encodeURIComponent).join("/");
+  return readJson(
+    await apiFetch(
+      `${API()}/knowledge-bases/${encodeURIComponent(knowledgeBaseId)}/wiki/pages/${safeSlug}`,
+    ),
+    "读取 Wiki 页面详情",
+  );
+}
+
+export async function uploadLlmWikiDocument(
+  knowledgeBaseId: string,
+  file: File,
+): Promise<LlmWikiDocument> {
+  const form = new FormData();
+  form.append("file", file);
+  return readJson(
+    await apiFetch(
+      `${API()}/knowledge-bases/${encodeURIComponent(knowledgeBaseId)}/documents`,
+      { method: "POST", body: form },
+    ),
+    "上传文档",
+  );
+}
+
+export async function importLlmWikiDocumentUrl(
+  knowledgeBaseId: string,
+  url: string,
+): Promise<LlmWikiDocument> {
+  return readJson(
+    await apiFetch(
+      `${API()}/knowledge-bases/${encodeURIComponent(knowledgeBaseId)}/documents/url`,
+      {
+        method: "POST",
+        headers: { "Content-Type": "application/json" },
+        body: JSON.stringify({ url }),
+      },
+    ),
+    "导入网页",
+  );
+}
+
+export async function createManualLlmWikiDocument(
+  knowledgeBaseId: string,
+  input: { title: string; content: string },
+): Promise<LlmWikiDocument> {
+  return readJson(
+    await apiFetch(
+      `${API()}/knowledge-bases/${encodeURIComponent(knowledgeBaseId)}/documents/manual`,
+      {
+        method: "POST",
+        headers: { "Content-Type": "application/json" },
+        body: JSON.stringify(input),
+      },
+    ),
+    "创建 Wiki 内容",
+  );
+}
+
+export async function deleteLlmWikiDocument(
+  knowledgeBaseId: string,
+  documentId: string,
+): Promise<void> {
+  await readJson<void>(
+    await apiFetch(
+      `${API()}/knowledge-bases/${encodeURIComponent(knowledgeBaseId)}/documents/${encodeURIComponent(documentId)}`,
+      { method: "DELETE" },
+    ),
+    "删除文档",
+  );
+}
+
+export async function reprocessLlmWikiDocument(
+  knowledgeBaseId: string,
+  documentId: string,
+): Promise<void> {
+  await readJson(
+    await apiFetch(
+      `${API()}/knowledge-bases/${encodeURIComponent(knowledgeBaseId)}/documents/${encodeURIComponent(documentId)}/reprocess`,
+      { method: "POST" },
+    ),
+    "重新解析文档",
+  );
+}
+
+export async function setLlmWikiPublication(
+  knowledgeBaseId: string,
+  publish: boolean,
+): Promise<void> {
+  await readJson(
+    await apiFetch(
+      `${API()}/knowledge-bases/${encodeURIComponent(knowledgeBaseId)}/${publish ? "publish" : "unpublish"}`,
+      { method: "POST" },
+    ),
+    publish ? "发布知识库" : "取消发布",
+  );
+}
+
+export async function searchLlmWiki(input: {
+  query: string;
+  scope: LlmWikiScope;
+  knowledge_base_ids?: string[];
+}): Promise<LlmWikiSearchResult[]> {
+  const payload = await readJson<{ results: LlmWikiSearchResult[] }>(
+    await apiFetch(`${API()}/search`, {
+      method: "POST",
+      headers: { "Content-Type": "application/json" },
+      body: JSON.stringify(input),
+    }),
+    "检索知识库",
+  );
+  return payload.results ?? [];
+}
+
+export async function createLlmWikiConversationDeposit(input: {
+  thread_id: string;
+  question: string;
+  answer: string;
+  human_message_id?: string | null;
+  assistant_message_id: string;
+  assistant_id?: string | null;
+  knowledge_base_ids?: string[];
+}): Promise<{ queued: boolean; provider: LlmWikiProvider }> {
+  return readJson(
+    await apiFetch(`${API()}/conversation-deposits/turn`, {
+      method: "POST",
+      headers: { "Content-Type": "application/json" },
+      body: JSON.stringify(input),
+    }),
+    "写入对话沉淀",
+  );
+}
+
+export async function getLlmWikiSettings(): Promise<LlmWikiSettings> {
+  return readJson(await apiFetch(SETTINGS_API()), "读取 WeKnora 连接配置");
+}
+
+export async function saveLlmWikiSettings(input: {
+  override_enabled: boolean;
+  api_base_url: string;
+  web_base_url: string;
+}): Promise<LlmWikiRuntime> {
+  return readJson(
+    await apiFetch(SETTINGS_API(), {
+      method: "PUT",
+      headers: { "Content-Type": "application/json" },
+      body: JSON.stringify(input),
+    }),
+    "保存 WeKnora 连接配置",
+  );
+}
+
+export async function resetLlmWikiSettings(): Promise<LlmWikiRuntime> {
+  return readJson(
+    await apiFetch(`${SETTINGS_API()}/override`, { method: "DELETE" }),
+    "恢复配置文件设置",
+  );
+}
+
+export async function testLlmWikiConnection(
+  apiBaseUrl: string,
+): Promise<void> {
+  await readJson(
+    await apiFetch(`${SETTINGS_API()}/test`, {
+      method: "POST",
+      headers: { "Content-Type": "application/json" },
+      body: JSON.stringify({ api_base_url: apiBaseUrl }),
+    }),
+    "测试 WeKnora 连接",
+  );
+}
+
+const runtimeKey = ["llmwiki", "runtime"] as const;
+const settingsKey = ["llmwiki", "settings"] as const;
+const knowledgeBasesKey = (scope: LlmWikiScope) =>
+  ["llmwiki", "knowledge-bases", scope] as const;
+
+export function useLlmWikiRuntime() {
+  return useQuery({
+    queryKey: runtimeKey,
+    queryFn: getLlmWikiRuntime,
+    staleTime: 30_000,
+    retry: 1,
+  });
+}
+
+export function useLlmWikiKnowledgeBases(
+  scope: LlmWikiScope,
+  enabled = true,
+) {
+  return useQuery({
+    queryKey: knowledgeBasesKey(scope),
+    queryFn: () => listLlmWikiKnowledgeBases(scope),
+    enabled,
+    refetchInterval: 15_000,
+  });
+}
+
+export function useAgentLlmWikiKnowledgeBases(
+  agentId: string | null | undefined,
+  enabled = true,
+) {
+  return useQuery({
+    queryKey: ["llmwiki", "agents", agentId, "knowledge-bases"],
+    queryFn: () => listAgentLlmWikiKnowledgeBases(agentId!),
+    enabled: enabled && !!agentId,
+  });
+}
+
+export function useLlmWikiKnowledgeBase(
+  knowledgeBaseId: string | null,
+  enabled = true,
+) {
+  return useQuery({
+    queryKey: ["llmwiki", "knowledge-base", knowledgeBaseId],
+    queryFn: () => getLlmWikiKnowledgeBase(knowledgeBaseId!),
+    enabled: enabled && !!knowledgeBaseId,
+    refetchInterval: 8_000,
+  });
+}
+
+export function useLlmWikiDocuments(
+  knowledgeBaseId: string | null,
+  enabled = true,
+) {
+  return useQuery({
+    queryKey: ["llmwiki", "documents", knowledgeBaseId],
+    queryFn: () => listLlmWikiDocuments(knowledgeBaseId!),
+    enabled: enabled && !!knowledgeBaseId,
+    refetchInterval: 8_000,
+  });
+}
+
+export function useLlmWikiDocument(
+  knowledgeBaseId: string | null,
+  documentId: string | null,
+  enabled = true,
+) {
+  return useQuery({
+    queryKey: ["llmwiki", "document", knowledgeBaseId, documentId],
+    queryFn: () => getLlmWikiDocument(knowledgeBaseId!, documentId!),
+    enabled: enabled && !!knowledgeBaseId && !!documentId,
+  });
+}
+
+export function useLlmWikiDocumentChunks(
+  knowledgeBaseId: string | null,
+  documentId: string | null,
+  page: number,
+  pageSize = 20,
+  enabled = true,
+) {
+  return useQuery({
+    queryKey: ["llmwiki", "document-chunks", knowledgeBaseId, documentId, page, pageSize],
+    queryFn: () => listLlmWikiDocumentChunks(knowledgeBaseId!, documentId!, page, pageSize),
+    enabled: enabled && !!knowledgeBaseId && !!documentId,
+  });
+}
+
+export function useLlmWikiGraph(knowledgeBaseId: string | null, enabled = true) {
+  return useQuery({
+    queryKey: ["llmwiki", "graph", knowledgeBaseId],
+    queryFn: () => getLlmWikiGraph(knowledgeBaseId!),
+    enabled: enabled && !!knowledgeBaseId,
+  });
+}
+
+export function useLlmWikiWikiPages(
+  knowledgeBaseId: string | null,
+  page: number,
+  pageSize = 50,
+  query = "",
+  enabled = true,
+) {
+  return useQuery({
+    queryKey: ["llmwiki", "wiki-pages", knowledgeBaseId, page, pageSize, query],
+    queryFn: () => listLlmWikiWikiPages(knowledgeBaseId!, page, pageSize, query),
+    enabled: enabled && !!knowledgeBaseId,
+    refetchInterval: 10_000,
+  });
+}
+
+export function useLlmWikiWikiPage(
+  knowledgeBaseId: string | null,
+  slug: string | null,
+  enabled = true,
+) {
+  return useQuery({
+    queryKey: ["llmwiki", "wiki-page", knowledgeBaseId, slug],
+    queryFn: () => getLlmWikiWikiPage(knowledgeBaseId!, slug!),
+    enabled: enabled && !!knowledgeBaseId && !!slug,
+  });
+}
+
+export function useLlmWikiSettings(enabled = true) {
+  return useQuery({
+    queryKey: settingsKey,
+    queryFn: getLlmWikiSettings,
+    enabled,
+  });
+}
+
+export function useLlmWikiActions() {
+  const queryClient = useQueryClient();
+  const invalidateKnowledge = () =>
+    queryClient.invalidateQueries({ queryKey: ["llmwiki"] });
+  return {
+    create: useMutation({
+      mutationFn: createLlmWikiKnowledgeBase,
+      onSuccess: invalidateKnowledge,
+    }),
+    update: useMutation({
+      mutationFn: ({ id, ...input }: { id: string; name?: string; description?: string }) =>
+        updateLlmWikiKnowledgeBase(id, input),
+      onSuccess: invalidateKnowledge,
+    }),
+    remove: useMutation({
+      mutationFn: deleteLlmWikiKnowledgeBase,
+      onSuccess: invalidateKnowledge,
+    }),
+    upload: useMutation({
+      mutationFn: ({ id, file }: { id: string; file: File }) =>
+        uploadLlmWikiDocument(id, file),
+      onSuccess: invalidateKnowledge,
+    }),
+    importUrl: useMutation({
+      mutationFn: ({ id, url }: { id: string; url: string }) =>
+        importLlmWikiDocumentUrl(id, url),
+      onSuccess: invalidateKnowledge,
+    }),
+    createManual: useMutation({
+      mutationFn: ({ id, title, content }: { id: string; title: string; content: string }) =>
+        createManualLlmWikiDocument(id, { title, content }),
+      onSuccess: invalidateKnowledge,
+    }),
+    removeDocument: useMutation({
+      mutationFn: ({ id, documentId }: { id: string; documentId: string }) =>
+        deleteLlmWikiDocument(id, documentId),
+      onSuccess: invalidateKnowledge,
+    }),
+    reprocessDocument: useMutation({
+      mutationFn: ({ id, documentId }: { id: string; documentId: string }) =>
+        reprocessLlmWikiDocument(id, documentId),
+      onSuccess: invalidateKnowledge,
+    }),
+    publication: useMutation({
+      mutationFn: ({ id, publish }: { id: string; publish: boolean }) =>
+        setLlmWikiPublication(id, publish),
+      onSuccess: invalidateKnowledge,
+    }),
+  };
+}
+
+export function useLlmWikiSettingsActions() {
+  const queryClient = useQueryClient();
+  const invalidate = async () => {
+    await queryClient.invalidateQueries({ queryKey: ["llmwiki"] });
+  };
+  return {
+    save: useMutation({ mutationFn: saveLlmWikiSettings, onSuccess: invalidate }),
+    reset: useMutation({ mutationFn: resetLlmWikiSettings, onSuccess: invalidate }),
+    test: useMutation({ mutationFn: testLlmWikiConnection }),
+  };
+}

frontend-web/src/strategy-components/components/resource-management/llmwiki/WeKnoraKnowledgeDetailPage.tsx

@@ -0,0 +1,152 @@
+import { AlertCircle, Loader2, RefreshCw } from "lucide-react";
+import { useEffect, useMemo, useRef } from "react";
+import { Navigate, useNavigate, useParams } from "react-router-dom";
+
+import { Button } from "@/components/ui/button";
+import { getBackendBaseURL } from "@/core/config";
+import {
+  useLlmWikiKnowledgeBase,
+  useLlmWikiRuntime,
+} from "@/strategy-components/api/llmwiki";
+
+const LIST_PATH = "/page/workspace/knowledge";
+
+type DeerFlowThemePayload = {
+  type: "deerflow-weknora-theme";
+  theme: "dark" | "light";
+  vars: Record<string, string>;
+};
+
+function cssVar(name: string, fallback = ""): string {
+  if (typeof window === "undefined") return fallback;
+  return getComputedStyle(document.documentElement).getPropertyValue(name).trim() || fallback;
+}
+
+function collectThemePayload(): DeerFlowThemePayload {
+  const isDark =
+    document.documentElement.classList.contains("dark") ||
+    document.documentElement.dataset.theme === "dark";
+  return {
+    type: "deerflow-weknora-theme",
+    theme: isDark ? "dark" : "light",
+    vars: {
+      primary: cssVar("--primary", isDark ? "#2b81ff" : "#0062d9"),
+      primaryForeground: cssVar("--button-foreground", cssVar("--primary-foreground", "#ffffff")),
+      background: cssVar("--background", isDark ? "#0e172b" : "#ffffff"),
+      foreground: cssVar("--foreground", isDark ? "#eff2f5" : "#202a3a"),
+      card: cssVar("--card", isDark ? "#1a2539" : "#ffffff"),
+      cardForeground: cssVar("--card-foreground", cssVar("--foreground", isDark ? "#eff2f5" : "#202a3a")),
+      muted: cssVar("--muted", isDark ? "#232e40" : "#f3f5f8"),
+      mutedForeground: cssVar("--muted-foreground", isDark ? "#8e9aa4" : "#7f7f7f"),
+      border: cssVar("--border", isDark ? "rgba(255,255,255,0.12)" : "#dbdbdb"),
+      popover: cssVar("--popover", cssVar("--card", isDark ? "#1a2539" : "#ffffff")),
+      destructive: cssVar("--destructive", "#d54941"),
+    },
+  };
+}
+
+function postThemeToFrame(frame: HTMLIFrameElement | null) {
+  if (!frame?.contentWindow) return;
+  const src = frame.getAttribute("src") || "";
+  const targetOrigin = new URL(src, window.location.href).origin;
+  frame.contentWindow.postMessage(collectThemePayload(), targetOrigin);
+}
+
+export function WeKnoraKnowledgeDetailPage() {
+  const {
+    knowledge_base_id: strategyKnowledgeBaseId,
+    noteId: workspaceKnowledgeBaseId,
+  } = useParams();
+  const knowledgeBaseId = strategyKnowledgeBaseId ?? workspaceKnowledgeBaseId ?? "";
+  const navigate = useNavigate();
+  const frameRef = useRef<HTMLIFrameElement>(null);
+  const runtime = useLlmWikiRuntime();
+  const details = useLlmWikiKnowledgeBase(knowledgeBaseId || null, !!knowledgeBaseId);
+
+  const preferredTab = details.data?.indexing_strategy?.wiki_enabled ? "wiki" : "documents";
+  const frameSrc = useMemo(() => {
+    if (!knowledgeBaseId || runtime.data?.provider !== "weknora") return "";
+    const base = getBackendBaseURL().replace(/\/+$/, "");
+    const params = new URLSearchParams({ tab: preferredTab });
+    return `${base}/api/llmwiki/knowledge-bases/${encodeURIComponent(
+      knowledgeBaseId,
+    )}/weknora/frame?${params.toString()}`;
+  }, [knowledgeBaseId, preferredTab, runtime.data?.provider]);
+
+  useEffect(() => {
+    if (!frameSrc) return;
+    const frame = frameRef.current;
+    const send = () => postThemeToFrame(frame);
+    const timer = window.setTimeout(send, 300);
+    const observer = new MutationObserver(send);
+    observer.observe(document.documentElement, {
+      attributes: true,
+      attributeFilter: ["class", "style", "data-theme"],
+    });
+    window.addEventListener("storage", send);
+    return () => {
+      window.clearTimeout(timer);
+      observer.disconnect();
+      window.removeEventListener("storage", send);
+    };
+  }, [frameSrc]);
+
+  if (runtime.isLoading || details.isLoading) {
+    return (
+      <div className="flex h-full items-center justify-center text-sm text-muted-foreground">
+        <Loader2 className="mr-2 size-5 animate-spin" />
+        正在打开 WeKnora 详情页
+      </div>
+    );
+  }
+
+  if (runtime.data?.provider !== "weknora") return <Navigate to={LIST_PATH} replace />;
+
+  if (details.error || !details.data) {
+    return (
+      <div className="flex h-full flex-col items-center justify-center gap-4 bg-background px-6 text-center">
+        <div className="flex size-12 items-center justify-center rounded-full bg-destructive/10 text-destructive">
+          <AlertCircle className="size-6" />
+        </div>
+        <div>
+          <div className="text-base font-semibold text-foreground">无法打开知识库详情</div>
+          <p className="mt-2 text-sm text-muted-foreground">
+            当前账号没有访问权限，或 WeKnora 资源已不存在。
+          </p>
+        </div>
+        <Button variant="outline" onClick={() => navigate(LIST_PATH)}>
+          返回知识库
+        </Button>
+      </div>
+    );
+  }
+
+  return (
+    <div className="flex h-full min-h-0 flex-col bg-background">
+      <iframe
+        key={frameSrc}
+        ref={frameRef}
+        title={`${details.data.name || "WeKnora"} 详情`}
+        src={frameSrc}
+        className="h-full min-h-0 w-full flex-1 border-0 bg-background"
+        sandbox="allow-same-origin allow-scripts allow-forms allow-downloads allow-popups"
+        referrerPolicy="no-referrer"
+        onLoad={() => postThemeToFrame(frameRef.current)}
+      />
+      {!frameSrc && (
+        <div className="absolute inset-0 flex items-center justify-center bg-background text-sm text-muted-foreground">
+          <Loader2 className="mr-2 size-5 animate-spin" />
+          正在准备 iframe
+        </div>
+      )}
+      {details.isFetching && (
+        <div className="pointer-events-none fixed bottom-4 right-4 rounded-full border bg-card px-3 py-2 text-xs text-muted-foreground shadow-sm">
+          <RefreshCw className="mr-2 inline size-3 animate-spin" />
+          同步知识库状态
+        </div>
+      )}
+    </div>
+  );
+}
+
+export default WeKnoraKnowledgeDetailPage;

frontend-web/src/strategy-components/components/resource-management/llmwiki/WeKnoraKnowledgeWorkspace.tsx

@@ -0,0 +1,318 @@
+import {
+  BookOpen,
+  Database,
+  FileText,
+  Globe2,
+  Loader2,
+  Plus,
+  RefreshCw,
+  Search,
+  Settings2,
+  Trash2,
+} from "lucide-react";
+import { useEffect, useMemo, useState } from "react";
+import { useNavigate } from "react-router-dom";
+import { toast } from "sonner";
+
+import { Badge } from "@/components/ui/badge";
+import { Button } from "@/components/ui/button";
+import { Card, CardContent, CardDescription, CardHeader, CardTitle } from "@/components/ui/card";
+import {
+  Dialog,
+  DialogContent,
+  DialogDescription,
+  DialogFooter,
+  DialogHeader,
+  DialogTitle,
+} from "@/components/ui/dialog";
+import { Input } from "@/components/ui/input";
+import { Tabs, TabsContent, TabsList, TabsTrigger } from "@/components/ui/tabs";
+import { Textarea } from "@/components/ui/textarea";
+import { getStoredAuth } from "@/core/auth";
+import {
+  type LlmWikiKnowledgeBase,
+  type LlmWikiScope,
+  searchLlmWiki,
+  useLlmWikiActions,
+  useLlmWikiKnowledgeBases,
+  useLlmWikiSettings,
+  useLlmWikiSettingsActions,
+} from "@/strategy-components/api/llmwiki";
+
+type WorkspaceTab = "personal" | "public" | "settings";
+const LIST_PATH = "/page/workspace/knowledge";
+
+const publicationLabels: Record<string, string> = {
+  private: "仅自己",
+  pending: "仅自己",
+  published: "已公开",
+  rejected: "仅自己",
+};
+
+function messageOf(error: unknown): string {
+  return error instanceof Error ? error.message : "操作失败，请稍后重试";
+}
+
+export function WeKnoraKnowledgeWorkspace() {
+  const navigate = useNavigate();
+  const isAdmin = getStoredAuth()?.system_role === "admin";
+  const [tab, setTab] = useState<WorkspaceTab>("personal");
+  const scope: LlmWikiScope = tab === "public" ? "public" : "personal";
+  const knowledge = useLlmWikiKnowledgeBases(scope, tab !== "settings");
+  const actions = useLlmWikiActions();
+  const [createOpen, setCreateOpen] = useState(false);
+  const [searchText, setSearchText] = useState("");
+  const [searching, setSearching] = useState(false);
+  const [searchResults, setSearchResults] = useState<Array<{
+    chunk_id: string;
+    title: string;
+    content: string;
+    score: number;
+    knowledge_base_name: string;
+  }>>([]);
+
+  useEffect(() => setSearchResults([]), [tab]);
+
+  const busy = actions.create.isPending || actions.remove.isPending || actions.publication.isPending;
+  const openDetails = (item: LlmWikiKnowledgeBase) => {
+    if (item.remote_status === "missing") {
+      toast.error("该知识库已从 WeKnora 删除，只能清理当前记录");
+      return;
+    }
+    navigate(`${LIST_PATH}/${encodeURIComponent(item.id)}`);
+  };
+
+  const runSearch = async () => {
+    const query = searchText.trim();
+    if (!query) return;
+    setSearching(true);
+    try {
+      setSearchResults(await searchLlmWiki({ query, scope }));
+    } catch (error) {
+      toast.error(messageOf(error));
+    } finally {
+      setSearching(false);
+    }
+  };
+
+  const removeKnowledgeBase = async (item: LlmWikiKnowledgeBase) => {
+    const warning = item.remote_status === "missing"
+      ? `知识库“${item.name}”已不在 WeKnora 中。确定清理 DeerFlow 中的这条失效记录吗？`
+      : `确定删除知识库“${item.name}”吗？WeKnora 中的文档和索引也会删除。`;
+    if (!window.confirm(warning)) return;
+    try {
+      await actions.remove.mutateAsync(item.id);
+      toast.success("知识库已删除");
+    } catch (error) {
+      toast.error(messageOf(error));
+    }
+  };
+
+  const togglePublication = async (item: LlmWikiKnowledgeBase) => {
+    const publish = item.publication_status !== "published";
+    try {
+      await actions.publication.mutateAsync({ id: item.id, publish });
+      toast.success(publish ? "知识库已发布到公共 LLMWiki" : "知识库已取消公开");
+    } catch (error) {
+      toast.error(messageOf(error));
+    }
+  };
+
+  return (
+    <div className="flex h-full min-h-0 flex-col bg-background">
+      <div className="border-b bg-gradient-to-r from-primary/8 via-card to-card px-6 py-5">
+        <div className="mx-auto flex max-w-[1500px] flex-wrap items-center justify-between gap-4">
+          <div>
+            <div className="flex items-center gap-2">
+              <Database className="size-5 text-primary" />
+              <h1 className="text-xl font-semibold">问答梳理</h1>
+              <Badge variant="secondary">WeKnora</Badge>
+            </div>
+            <p className="mt-1 text-sm text-muted-foreground">由 WeKnora 负责解析与检索，用户、权限和发布范围由 DeerFlow 管理。</p>
+          </div>
+          <div className="flex items-center gap-2 rounded-full border bg-background/80 p-1 shadow-sm backdrop-blur">
+            {tab !== "settings" && (
+              <Button variant="ghost" size="sm" className="rounded-full" onClick={() => knowledge.refetch()} disabled={knowledge.isFetching}>
+                <RefreshCw className={`mr-2 size-4 ${knowledge.isFetching ? "animate-spin" : ""}`} />刷新
+              </Button>
+            )}
+            {tab === "personal" && <Button size="sm" className="rounded-full px-4 shadow-sm" onClick={() => setCreateOpen(true)}><Plus className="mr-2 size-4" />新建知识库</Button>}
+          </div>
+        </div>
+      </div>
+
+      <div className="min-h-0 flex-1 overflow-auto px-6 py-5">
+        <div className="mx-auto max-w-[1500px]">
+          <Tabs value={tab} onValueChange={(value) => setTab(value as WorkspaceTab)}>
+            <div className="flex flex-wrap items-center justify-between gap-3 rounded-2xl border bg-card/80 p-3 shadow-sm">
+              <TabsList className="rounded-xl bg-muted/70">
+                <TabsTrigger value="personal"><BookOpen className="mr-2 size-4" />我的知识库</TabsTrigger>
+                <TabsTrigger value="public"><Globe2 className="mr-2 size-4" />公共 LLMWiki</TabsTrigger>
+                {isAdmin && <TabsTrigger value="settings"><Settings2 className="mr-2 size-4" />连接配置</TabsTrigger>}
+              </TabsList>
+              {tab !== "settings" && (
+                <div className="flex w-full max-w-xl items-center gap-2 rounded-xl border bg-background/80 p-1 shadow-sm sm:w-auto">
+                  <div className="relative min-w-0 flex-1 sm:w-80">
+                    <Search className="absolute left-3 top-1/2 size-4 -translate-y-1/2 text-muted-foreground" />
+                    <Input value={searchText} onChange={(event) => setSearchText(event.target.value)} onKeyDown={(event) => event.key === "Enter" && void runSearch()} placeholder={tab === "public" ? "检索公共知识" : "检索我的知识"} className="border-0 bg-transparent pl-9 shadow-none focus-visible:ring-0" />
+                  </div>
+                  <Button variant="secondary" className="rounded-lg px-4" onClick={() => void runSearch()} disabled={searching || !searchText.trim()}>{searching ? <Loader2 className="size-4 animate-spin" /> : "检索"}</Button>
+                </div>
+              )}
+            </div>
+
+            <TabsContent value="personal" className="mt-5"><KnowledgeGrid rows={knowledge.data ?? []} loading={knowledge.isLoading} error={knowledge.error} emptyText="还没有个人知识库，点击右上角创建。" busy={busy} onOpen={openDetails} onPublish={togglePublication} onDelete={removeKnowledgeBase} /></TabsContent>
+            <TabsContent value="public" className="mt-5"><KnowledgeGrid rows={knowledge.data ?? []} loading={knowledge.isLoading} error={knowledge.error} emptyText="公共 LLMWiki 暂无已发布知识库。" busy={busy} onOpen={openDetails} onPublish={togglePublication} onDelete={removeKnowledgeBase} /></TabsContent>
+            {isAdmin && <TabsContent value="settings" className="mt-5"><ConnectionSettings /></TabsContent>}
+          </Tabs>
+
+          {searchResults.length > 0 && tab !== "settings" && (
+            <Card className="mt-5">
+              <CardHeader className="pb-3"><CardTitle className="text-base">检索结果</CardTitle><CardDescription>结果由 WeKnora 检索返回，问答中使用同一授权链路。</CardDescription></CardHeader>
+              <CardContent className="space-y-3">{searchResults.map((result) => <div key={result.chunk_id} className="rounded-lg border p-3"><div className="flex items-center justify-between gap-3 text-sm font-medium"><span>{result.title || "未命名来源"}</span><Badge variant="outline">{result.knowledge_base_name}</Badge></div><p className="mt-2 line-clamp-3 whitespace-pre-wrap text-sm text-muted-foreground">{result.content}</p></div>)}</CardContent>
+            </Card>
+          )}
+        </div>
+      </div>
+
+      <CreateKnowledgeDialog open={createOpen} pending={actions.create.isPending} onOpenChange={setCreateOpen} onCreate={async (input) => {
+        try {
+          const created = await actions.create.mutateAsync(input);
+          setCreateOpen(false);
+          navigate(`${LIST_PATH}/${encodeURIComponent(created.id)}`);
+          toast.success("知识库已创建");
+        } catch (error) {
+          toast.error(messageOf(error));
+        }
+      }} />
+    </div>
+  );
+}
+
+function KnowledgeGrid({
+  rows,
+  loading,
+  error,
+  emptyText,
+  busy,
+  onOpen,
+  onPublish,
+  onDelete,
+}: {
+  rows: LlmWikiKnowledgeBase[];
+  loading: boolean;
+  error: unknown;
+  emptyText: string;
+  busy: boolean;
+  onOpen: (item: LlmWikiKnowledgeBase) => void;
+  onPublish: (item: LlmWikiKnowledgeBase) => void;
+  onDelete: (item: LlmWikiKnowledgeBase) => void;
+}) {
+  if (loading) return <div className="flex h-64 items-center justify-center text-muted-foreground"><Loader2 className="mr-2 size-5 animate-spin" />正在读取知识库</div>;
+  if (error) return <div className="rounded-lg border border-destructive/30 bg-destructive/5 p-6 text-sm text-destructive">{messageOf(error)}</div>;
+  if (!rows.length) return <div className="flex h-64 items-center justify-center rounded-xl border border-dashed text-sm text-muted-foreground">{emptyText}</div>;
+  return (
+    <div className="grid gap-4 md:grid-cols-2 xl:grid-cols-3">
+      {rows.map((item) => (
+        <Card
+          key={item.id}
+          className={`group overflow-hidden rounded-2xl transition-all hover:-translate-y-0.5 hover:shadow-lg ${
+            item.remote_status === "missing"
+              ? "border-amber-500/40 bg-amber-500/[0.03]"
+              : "border-border/70 bg-card/95 hover:border-primary/30"
+          }`}
+        >
+          <div className={`h-1 ${item.publication_status === "published" ? "bg-primary" : "bg-muted"}`} />
+          <button type="button" disabled={item.remote_status === "missing"} className="w-full text-left outline-none transition-colors enabled:hover:bg-muted/25 focus-visible:ring-2 focus-visible:ring-primary focus-visible:ring-offset-2 disabled:cursor-default" onClick={() => onOpen(item)}>
+            <CardHeader className="pb-3 pt-5">
+              <div className="flex items-start justify-between gap-3">
+                <div className="min-w-0"><CardTitle className="truncate text-base">{item.name}</CardTitle><CardDescription className="mt-1 line-clamp-2 min-h-10">{item.description || "暂无描述"}</CardDescription></div>
+                <div className="flex shrink-0 flex-col items-end gap-1.5">{item.remote_status === "missing" ? <Badge variant="outline" className="border-amber-500/50 text-amber-700 dark:text-amber-300">远端已删除</Badge> : <Badge variant={item.publication_status === "published" ? "default" : "secondary"}>{publicationLabels[item.publication_status] ?? "仅自己"}</Badge>}{item.remote_status !== "missing" && item.indexing_strategy?.wiki_enabled && <Badge className="bg-violet-100 text-violet-700 hover:bg-violet-100 dark:bg-violet-950 dark:text-violet-300">Wiki</Badge>}</div>
+              </div>
+            </CardHeader>
+            <div className="px-6 pb-4"><div className="grid grid-cols-3 gap-2 rounded-xl border bg-muted/35 p-3 text-center text-xs"><div><div className="text-base font-semibold text-foreground">{item.knowledge_count}</div><span className="text-muted-foreground">文档</span></div><div><div className="text-base font-semibold text-foreground">{item.chunk_count}</div><span className="text-muted-foreground">分块</span></div><div><div className="text-base font-semibold text-foreground">{item.processing_count}</div><span className="text-muted-foreground">处理中</span></div></div></div>
+          </button>
+          <CardContent className="border-t bg-muted/20 px-6 py-4">
+            <div className="flex items-center justify-between gap-3">
+              <div className="flex min-w-0 flex-wrap items-center gap-2">
+                {item.remote_status === "missing" ? <span className="text-xs text-muted-foreground">WeKnora 资源不存在，可清理本地记录</span> : <Button size="sm" variant="outline" className="rounded-full" onClick={() => onOpen(item)}><FileText className="mr-2 size-4" />查看详情</Button>}
+                {item.remote_status !== "missing" && item.is_owner && <Button size="sm" variant="secondary" className="rounded-full" disabled={busy} onClick={() => void onPublish(item)}>{item.publication_status === "published" ? "取消公开" : "发布公共库"}</Button>}
+              </div>
+              {item.can_write && <Button size="icon" variant="ghost" className="size-8 shrink-0 rounded-full text-destructive hover:bg-destructive/10 hover:text-destructive" aria-label={item.remote_status === "missing" ? `清理失效记录 ${item.name}` : `删除知识库 ${item.name}`} title={item.remote_status === "missing" ? "清理失效记录" : "删除知识库"} disabled={busy} onClick={() => void onDelete(item)}><Trash2 className="size-4" /></Button>}
+            </div>
+          </CardContent>
+        </Card>
+      ))}
+    </div>
+  );
+}
+
+function CreateKnowledgeDialog({
+  open,
+  pending,
+  onOpenChange,
+  onCreate,
+}: {
+  open: boolean;
+  pending: boolean;
+  onOpenChange: (value: boolean) => void;
+  onCreate: (input: { name: string; description: string; type: "document" | "faq"; wiki_enabled: boolean }) => Promise<void>;
+}) {
+  const [name, setName] = useState("");
+  const [description, setDescription] = useState("");
+  const [wikiEnabled, setWikiEnabled] = useState(true);
+  useEffect(() => {
+    if (!open) { setName(""); setDescription(""); setWikiEnabled(true); }
+  }, [open]);
+  return (
+    <Dialog open={open} onOpenChange={onOpenChange}>
+      <DialogContent>
+        <DialogHeader><DialogTitle>新建个人知识库</DialogTitle><DialogDescription>选择知识库模式；模型、切块和存储策略仍由 WeKnora 管理。</DialogDescription></DialogHeader>
+        <div className="space-y-4 py-2">
+          <div className="space-y-2"><label className="text-sm font-medium" htmlFor="llmwiki-name">名称</label><Input id="llmwiki-name" value={name} onChange={(event) => setName(event.target.value)} maxLength={255} /></div>
+          <div className="space-y-2"><label className="text-sm font-medium" htmlFor="llmwiki-description">描述</label><Textarea id="llmwiki-description" value={description} onChange={(event) => setDescription(event.target.value)} rows={4} maxLength={2048} /></div>
+          <div className="space-y-2"><div className="text-sm font-medium">知识库模式</div><div className="grid gap-3 sm:grid-cols-2">
+            <button type="button" className={`rounded-xl border p-4 text-left transition-colors ${wikiEnabled ? "border-primary bg-primary/5 ring-1 ring-primary" : "hover:bg-muted/40"}`} onClick={() => setWikiEnabled(true)}><div className="flex items-center gap-2 text-sm font-semibold"><BookOpen className="size-4 text-violet-500" />Wiki 知识库</div><p className="mt-2 text-xs leading-5 text-muted-foreground">自动综合资料，生成 Wiki 页面和页面关系图谱。</p></button>
+            <button type="button" className={`rounded-xl border p-4 text-left transition-colors ${!wikiEnabled ? "border-primary bg-primary/5 ring-1 ring-primary" : "hover:bg-muted/40"}`} onClick={() => setWikiEnabled(false)}><div className="flex items-center gap-2 text-sm font-semibold"><FileText className="size-4 text-blue-500" />标准知识库</div><p className="mt-2 text-xs leading-5 text-muted-foreground">仅进行文档解析、切块与向量/关键词索引。</p></button>
+          </div></div>
+        </div>
+        <DialogFooter><Button variant="outline" onClick={() => onOpenChange(false)}>取消</Button><Button disabled={pending || !name.trim()} onClick={() => void onCreate({ name: name.trim(), description: description.trim(), type: "document", wiki_enabled: wikiEnabled })}>{pending && <Loader2 className="mr-2 size-4 animate-spin" />}创建</Button></DialogFooter>
+      </DialogContent>
+    </Dialog>
+  );
+}
+
+export function ConnectionSettings() {
+  const settings = useLlmWikiSettings(true);
+  const actions = useLlmWikiSettingsActions();
+  const [apiUrl, setApiUrl] = useState("");
+  const [webUrl, setWebUrl] = useState("");
+
+  useEffect(() => {
+    if (!settings.data) return;
+    setApiUrl(settings.data.api_base_url || settings.data.file_api_base_url || "");
+    setWebUrl(settings.data.web_base_url || settings.data.file_web_base_url || "");
+  }, [settings.data]);
+
+  const status = useMemo(() => settings.data?.effective_provider === "weknora" ? "WeKnora 模式" : "原 LLMWiki 模式", [settings.data]);
+  if (settings.isLoading) return <div className="flex h-52 items-center justify-center"><Loader2 className="size-5 animate-spin" /></div>;
+  if (settings.error) return <div className="rounded-lg border border-destructive/30 p-5 text-sm text-destructive">{messageOf(settings.error)}</div>;
+  return (
+    <Card className="max-w-3xl">
+      <CardHeader><CardTitle className="flex items-center gap-2 text-base"><Settings2 className="size-5" />WeKnora 连接</CardTitle><CardDescription>这里只管理 DeerFlow 到 WeKnora 的连接地址。数据库、Redis、模型、存储和解析器仍在 WeKnora 自己的配置中维护。</CardDescription></CardHeader>
+      <CardContent className="space-y-5">
+        <div className="flex flex-wrap gap-2 text-sm"><Badge>{status}</Badge><Badge variant="secondary">服务端代理鉴权</Badge>{settings.data?.override_enabled && <Badge variant="outline">页面配置覆盖中</Badge>}</div>
+        <div className="space-y-2"><label className="text-sm font-medium" htmlFor="weknora-api-url">WeKnora API 地址</label><Input id="weknora-api-url" value={apiUrl} onChange={(event) => setApiUrl(event.target.value)} placeholder="http://weknora-app:8080" /><p className="text-xs text-muted-foreground">非空启用 WeKnora；保存为空会切回原 LLMWiki。</p></div>
+        <div className="space-y-2"><label className="text-sm font-medium" htmlFor="weknora-web-url">WeKnora 管理地址（可选）</label><Input id="weknora-web-url" value={webUrl} onChange={(event) => setWebUrl(event.target.value)} placeholder="https://weknora-admin.example.com" /></div>
+        <div className="rounded-lg border border-amber-500/30 bg-amber-500/5 p-3 text-xs text-muted-foreground">配置文件地址：{settings.data?.file_api_base_url || "未配置"}。页面只配置地址；WeKnora iframe 免登录所需的管理员账号/密码只从后端 config.yaml 或环境变量读取，不在前端保存。</div>
+        <div className="flex flex-wrap gap-2">
+          <Button variant="outline" disabled={!apiUrl.trim() || actions.test.isPending} onClick={async () => { try { await actions.test.mutateAsync(apiUrl.trim()); toast.success("WeKnora 连接正常"); } catch (error) { toast.error(messageOf(error)); } }}>{actions.test.isPending && <Loader2 className="mr-2 size-4 animate-spin" />}测试连接</Button>
+          <Button disabled={actions.save.isPending} onClick={async () => { try { await actions.save.mutateAsync({ override_enabled: true, api_base_url: apiUrl.trim(), web_base_url: webUrl.trim() }); toast.success(apiUrl.trim() ? "已启用 WeKnora 模式" : "已切换到原 LLMWiki 模式"); } catch (error) { toast.error(messageOf(error)); } }}>{actions.save.isPending && <Loader2 className="mr-2 size-4 animate-spin" />}保存页面配置</Button>
+          <Button variant="secondary" disabled={actions.reset.isPending} onClick={async () => { try { await actions.reset.mutateAsync(); toast.success("已恢复 config.yaml 配置"); } catch (error) { toast.error(messageOf(error)); } }}>恢复配置文件</Button>
+        </div>
+      </CardContent>
+    </Card>
+  );
+}
+
+export default WeKnoraKnowledgeWorkspace;

offline-backend-20260512/README.md

@@ -1,13 +1,13 @@
 # DeerFlow 离线后端部署补充
 
-本目录提供与原始 `start_backend.sh` 配套的 wheelhouse。启动脚本保持原有结构，只新增一行 Redis Python 客户端的离线安装命令。
+本目录提供与 `start_backend.sh` 配套的 wheelhouse。启动脚本会在容器启动时检查可选运行时依赖，并在缺包时仅从本地 wheelhouse 安装。
 
 ## 文件
 
 - `offline-runtime-requirements.txt`：固定版本的完整依赖闭包。
 - `prepare_wheelhouse.sh`：联网机器下载 Linux/Python 3.12 wheel，并验证传递依赖完整性。
-- `start_backend.sh`：原始 Docker 启动脚本，加一行 Redis 缺包安装命令。
-- `wheelhouse/`：22 个 wheel 和 `SHA256SUMS`，需要随发布目录一起复制。
+- `start_backend.sh`：Docker 启动脚本，包含 `croniter`、`hindsight-client`、`aiomysql` 和 Redis 客户端的缺包补装逻辑。
+- `wheelhouse/`：24 个 wheel 和 `SHA256SUMS`，需要随发布目录一起复制。
 
 ## 联网机器准备
 
@@ -23,8 +23,9 @@ bash prepare_wheelhouse.sh
 - `croniter==6.2.2`
 - `hindsight-client==0.6.2`
 - `redis==5.2.1`
+- `aiomysql==0.3.2`
 
-上述依赖在 Python 3.12/Linux x86_64 下的 19 个传递依赖也已锁定在 `offline-runtime-requirements.txt`。
+上述依赖在 Python 3.12/Linux x86_64 下的 20 个传递依赖（包括 `PyMySQL==1.2.0`）也已锁定在 `offline-runtime-requirements.txt`。
 
 国内镜像示例：
 
@@ -57,7 +58,9 @@ bash start_backend.sh
 docker logs -f deerflow-backend
 ```
 
-原脚本已有 `croniter` 和 `hindsight-client` 的离线补装逻辑；新增的一行命令只负责在缺少 `redis` 模块时，从 `/app/wheelhouse` 安装 `redis==5.2.1`。
+脚本会先尝试导入依赖，仅在缺少模块时从 `/app/wheelhouse` 安装。`aiomysql` 补装默认开启，可通过 `AIOMYSQL_INSTALL_DEPS=0` 禁用，或通过 `AIOMYSQL_PIP_PACKAGES` 覆盖默认的 `aiomysql==0.3.2`。
+
+MySQL 连接 URL 使用 aiomysql 驱动时应写为 `mysql+aiomysql://user:password@host:3306/database?charset=utf8mb4`。
 
 ## Redis 说明
 

offline-backend-20260512/backend/CLAUDE.md

@@ -257,7 +257,7 @@ Configuration priority:
 
 ### 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).
+FastAPI application on port 8001 with health checks at `GET|HEAD /health`, `/api/health`, and `/api/langgraph/health`. These exact paths are always public in `auth_middleware.PUBLIC_HEALTH_PATHS`, including when authentication is enabled; reverse-proxy prefixes and trailing slashes are normalized before the whitelist check. Set `GATEWAY_ENABLE_DOCS=false` to disable `/docs`, `/redoc`, and `/openapi.json` in production (default: enabled). Regression coverage: `tests/test_health_auth_bypass.py`.
 
 **Routers**:
 
@@ -600,7 +600,7 @@ Operators upgrading an existing DB run the whole file (`USE <db>;` then execute)
 ### 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()`
+1. **Config-defined tools** - Enabled entries are resolved from `config.yaml` via `resolve_variable()`; entries with `enabled: false` are omitted before import and are not registered with the agent
 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`)
@@ -613,6 +613,7 @@ Operators upgrading an existing DB run the whole file (`USE <db>;` then execute)
 - `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
+- `configurable_search/` - Config-driven `POST` JSON `web_search` replacement. It deep-copies the configured payload and overwrites only `query`, supports HTTP/HTTPS plus configurable `verify_ssl`, timeout and result limit, normalizes nullable `recUuid`/`content`/`title`/`url` fields, and rewrites a result URL through `result_url_template` only when `recUuid` is non-empty. The required response envelope is `{"results": [...]}`. Tests use `httpx.MockTransport` and never require the deployment endpoint.
 
 **ACP agent tools**:
 - `invoke_acp_agent` - Invokes external ACP-compatible agents from `config.yaml`
@@ -921,9 +922,9 @@ Both endpoints are whitelisted in `auth_middleware._PUBLIC_EXACT_PATHS`. The sha
 
 Both can be modified at runtime via Gateway API endpoints or `DeerFlowClient` methods.
 
-### WeKnora Retrieval Skill
+### Conditional WeKnora LLMWiki
 
-`skills/custom/weknora-knowledge-search/` is a read-only integration for a WeKnora tenant. The Skill itself only directs the model to two MCP tools, `weknora_search` and `weknora_list_knowledge_bases`; the credential remains inside the stdio MCP subprocess (`scripts/weknora_retrieval_mcp.py`). Its retrieval core groups document/FAQ knowledge bases by `embedding_model_id` and calls WeKnora's multi-KB `hybrid-search` API once per group, preventing invalid cross-embedding score comparisons. It does not discard a KB merely because the legacy list API reports `chunk_count=0`; hybrid-search decides its actual readiness. Wiki pages are opt-in and use their dedicated Wiki search route. In explicit DeerFlow sharing mode, `WEKNORA_DEERFLOW_SHARE_TOKEN_FILE` selects WeKnora's narrow `/api/v1/deerflow/*` capability API: only owner-published document/FAQ knowledge bases are discoverable and searchable, while Wiki is skipped so the skill cannot fall back to a user bearer-token endpoint. Shared document hits include a credential-free source-preview URL. The frontend source card recognizes it and opens a context dialog; its call to authenticated Gateway route `/api/weknora/source-context` is proxied to WeKnora's token-gated `/api/v1/deerflow/knowledge-bases/:id/chunks/:chunk_id/context`, returning only the matching chunk and two neighbours on either side. Never expose the share token to the browser or add it to a URL. Configure through the bundled `templates/extensions_config.fragment.json`; otherwise prefer `WEKNORA_API_KEY_FILE` for scoped-key deployments and `WEKNORA_BEARER_TOKEN_FILE` for legacy deployments. `WEKNORA_ALLOWED_KB_IDS` remains an optional allow-list. Tests: `tests/test_weknora_knowledge_search_skill.py`, `tests/test_weknora_source_preview.py`.
+`llmwiki.weknora.api_base_url` is the provider switch: non-empty enables WeKnora, empty preserves the legacy LLMWiki UI. WeKnora is independently deployed and owns PostgreSQL/Redis/object storage/vector storage/parsers/models; DeerFlow config contains only `api_base_url` and optional `web_base_url`. DeerFlow deliberately sends no WeKnora credential, so the WeKnora endpoint must be restricted to a trusted network and allow DeerFlow traffic itself or through its reverse proxy. There is intentionally no `default_embedding_model_id`: `WeKnoraClient` queries `/api/v1/models` and uses WeKnora's active Embedding model when creating a KB; Wiki creation additionally resolves an active KnowledgeQA model inside WeKnora for synthesis. The native BFF is `app/gateway/routers/llmwiki.py`; ownership/publication mappings live in `deerflow.persistence.llmwiki`; the browser only sees DeerFlow mapping IDs. Knowledge-base owners publish immediately (`private` -> `published`) without a review queue, and can unpublish. Knowledge cards navigate to a dedicated WeKnora-style detail route. Wiki-enabled bases render and default to `Wiki -> documents -> graph`; standard bases omit Wiki and render `documents -> graph`. The detail page supports file, URL and published manual-Markdown imports, browses real Wiki pages/content, proxies authenticated original-file previews, pages through `/chunks/{knowledge_id}`, and renders the Wiki page graph from `/knowledgebase/{kb_id}/wiki/graph`; document details use a right sheet rather than the removed KB modal. The credential-free proxy uses one WeKnora service identity, so per-user separation is enforced by DeerFlow mappings while all remote KBs physically share that identity's WeKnora workspace. `llmwiki_search` is conditionally registered in `tools.py`, is read-only, and rechecks the current DeerFlow user, agent `llmwiki_knowledge_base_ids`, and thread-context selection on every call. Agent bindings use an inline WeKnora-style searchable/grouped multi-select in the standard agent create/detail UI; new chats use the same interaction pattern as a compact composer dropdown, and selections are written to thread metadata/context and restored on reopen. The DeerFlow frontend is native React and the WeKnora frontend is not modified or embedded. Full deployment and behavior notes: `docs/LLMWIKI_WEKNORA_INTEGRATION_ZH.md`. Tests: `tests/test_llmwiki_weknora.py`.
 
 ### Embedded Client (`packages/harness/deerflow/client.py`)
 
@@ -1133,7 +1134,7 @@ not delete a built-in id because historical chain nodes persist those ids;
 custom role ids may be added and removed.
 
 See `docs/` directory for detailed documentation:
-- [../README.md](../README.md) - Offline Docker wheelhouse preparation and startup workflow. The pinned versions in `../offline-runtime-requirements.txt` and `../start_backend.sh` must stay synchronized. `start_backend.sh` stays aligned with the deployment script and adds only the Redis wheel fallback line. The download target defaults to CPython 3.12 on Linux x86_64 and verifies dependency closure using wheelhouse only.
+- [../README.md](../README.md) - Offline Docker wheelhouse preparation and startup workflow. The pinned versions in `../offline-runtime-requirements.txt` and `../start_backend.sh` must stay synchronized. Startup checks `croniter`, `hindsight-client`, `aiomysql`, and Redis, installing missing packages from wheelhouse only; `aiomysql==0.3.2` includes the pinned `PyMySQL==1.2.0` transitive dependency. The download target defaults to CPython 3.12 on Linux x86_64 and verifies dependency closure using wheelhouse only.
 - [CONFIGURATION.md](docs/CONFIGURATION.md) - Configuration options
 - [ARCHITECTURE.md](docs/ARCHITECTURE.md) - Architecture details
 - [API.md](docs/API.md) - API reference

offline-backend-20260512/backend/README.md

@@ -39,6 +39,8 @@ DeerFlow is a LangGraph-based AI super agent with sandbox execution, persistent
 - `/api/*` (other) → Gateway API - models, MCP, skills, memory, artifacts, uploads, thread-local cleanup
 - `/` (non-API) → Frontend - Next.js web interface
 
+Health probes are always unauthenticated, even when API authentication is enabled: `GET|HEAD /health`, `/api/health`, and `/api/langgraph/health` all return the Gateway health status without requiring a cookie or bearer token.
+
 ---
 
 ## Core Components
@@ -78,7 +80,7 @@ Per-thread isolated execution with virtual path translation:
 - **Virtual paths**: `/mnt/user-data/{workspace,uploads,outputs}` → thread-specific physical directories
 - **Skills path**: `/mnt/skills` → `deer-flow/skills/` directory
 - **Skills loading**: Recursively discovers nested `SKILL.md` files under `skills/{public,custom}` and preserves nested container paths
-- **WeKnora retrieval skill**: `skills/custom/weknora-knowledge-search` pairs a concise Skill with a stdio MCP server. It retrieves tenant-scoped WeKnora document/FAQ passages through the public API, groups multi-KB calls by embedding model, optionally searches Wiki pages, and can use the local reranker. It does not reject a KB solely because the legacy list API reports `chunk_count=0`; hybrid-search remains the source of truth. For WeKnora's explicit DeerFlow sharing mode, configure `WEKNORA_DEERFLOW_SHARE_TOKEN_FILE`: the MCP uses the restricted `/api/v1/deerflow/*` capability API and can only list/search knowledge bases whose owner enabled sharing; it never receives a user bearer token. Shared document results now contain a credential-free `url` pointer; the source card's “查看原文” dialog calls DeerFlow's authenticated `/api/weknora/source-context` proxy, which retrieves the hit plus two adjacent chunks while keeping the share token server-side. Shared mode currently excludes Wiki pages. Otherwise use `WEKNORA_API_KEY_FILE` for scoped-key deployments or `WEKNORA_BEARER_TOKEN_FILE` for legacy deployments; never put any credential in `SKILL.md` or a chat prompt.
+- **Conditional WeKnora LLMWiki**: a non-empty `llmwiki.weknora.api_base_url` enables the DeerFlow-native WeKnora BFF and UI; an empty address preserves the existing LLMWiki page. WeKnora remains independently deployed and owns its database, Redis, storage, parsers and models. DeerFlow keeps only the API/admin addresses, user/publication mappings, agent knowledge-base bindings and per-conversation selections; it stores and sends no WeKnora credential, so the endpoint must be restricted to a trusted network and allow DeerFlow traffic. Owners publish directly to the public LLMWiki without an approval queue. Knowledge cards open a dedicated WeKnora-style detail route: Wiki-enabled bases default to Wiki, then documents and graph; standard bases hide Wiki and default to documents. It supports file/URL/manual-Markdown imports, Wiki page browsing, document preview, real parsed chunks and Wiki graph nodes/edges. New chats use a compact WeKnora-style searchable/grouped multi-select inside the composer, while agent create/edit pages bind spaces through an inline panel. A single proxy service identity means DeerFlow users are isolated logically while their WeKnora entities share that identity's workspace. The read-only `llmwiki_search` tool reauthorizes the current DeerFlow user on every call and renders credential-free source previews. See [docs/LLMWIKI_WEKNORA_INTEGRATION_ZH.md](docs/LLMWIKI_WEKNORA_INTEGRATION_ZH.md).
 - **File-write safety**: `str_replace` serializes read-modify-write per `(sandbox.id, path)` so isolated sandboxes keep concurrency even when virtual paths match
 - **Tools**: `bash`, `ls`, `read_file`, `write_file`, `str_replace` (`bash` is disabled by default when using `LocalSandboxProvider`; use `AioSandboxProvider` for isolated shell access)
 
@@ -111,6 +113,19 @@ LLM-powered persistent context retention across conversations:
 | **MCP** | Any Model Context Protocol server (stdio, SSE, HTTP transports) |
 | **Skills** | Domain-specific workflows injected via system prompt |
 
+#### Configurable HTTP web search
+
+The config-defined `web_search` tool can call a deployment-specific JSON API
+instead of a public search provider. Set its `endpoint`, static `payload`, SSL
+verification, timeout, result limit, and `result_url_template` in `config.yaml`.
+At runtime only `payload.query` is overwritten with the model's search text.
+The API response must use `{"results": [{"recUuid": "", "content": "",
+"title": "", "url": ""}]}`; missing or null result fields are normalized to empty strings. When
+`recUuid` is present, `{recUuid}` in `result_url_template` replaces the original
+result URL; otherwise the original URL is retained. Set `enabled: false` to keep
+the tool completely out of the agent's registered tool list. `verify_ssl: false`
+supports trusted internal HTTPS services with self-signed certificates.
+
 ### Gateway API
 
 FastAPI application providing REST endpoints for frontend integration:

offline-backend-20260512/backend/app/gateway/app.py

@@ -1,26 +1,26 @@
 import asyncio
 import logging
-import os
-import queue
-import sys
-import threading
-from collections.abc import AsyncGenerator
-from contextlib import asynccontextmanager
-from datetime import datetime
-from time import perf_counter
-
-from dotenv import load_dotenv
-from fastapi import FastAPI
+import os
+import queue
+import sys
+import threading
+from collections.abc import AsyncGenerator
+from contextlib import asynccontextmanager
+from datetime import datetime
+from time import perf_counter
+
+from dotenv import load_dotenv
+from fastapi import FastAPI
 from fastapi.middleware.cors import CORSMiddleware
 
 # Load .env BEFORE any module that reads os.environ at import time.
 # Without this, GATEWAY_CORS_ORIGINS / AUTH_JWT_SECRET / DEER_FLOW_AUTH_DISABLED
 # placed in /workspace/.env are silently ignored when the container is started
 # without explicit `-e` flags, leading to ephemeral JWT secrets, missing CORS
-# origins, and "Failed to fetch" in the browser.
-load_dotenv()
-
-from app.gateway.auth.disabled_mode import is_auth_disabled  # noqa: E402
+# origins, and "Failed to fetch" in the browser.
+load_dotenv()
+
+from app.gateway.auth.disabled_mode import is_auth_disabled  # noqa: E402
 from app.gateway.auth_middleware import AuthMiddleware
 from app.gateway.config import get_gateway_config
 from app.gateway.csrf_middleware import CSRFMiddleware
@@ -43,16 +43,17 @@ from app.gateway.routers import (
     intent,
     knowledge,
     knowledge_ingest,
-    light_apps,
-    llm_metrics,
-    mcp,
+    light_apps,
+    llm_metrics,
+    llmwiki,
+    mcp,
     memory,
     menu_overrides,
     models,
     multi_agent,
     notifications,
     open_3qfx,
-    open_chat,
+    open_chat,
     parallel_agent_tasks,
     parallel_agents,
     position_roundtable,
@@ -89,8 +90,7 @@ from app.gateway.routers import (
     uploads,
     user_preferences,
     user_prompts,
-    weknora_sources,
-    writing,
+    writing,
 )
 from deerflow.config import app_config as deerflow_app_config
 from deerflow.config.app_config import apply_logging_level
@@ -1430,8 +1430,8 @@ This gateway provides custom endpoints for models, MCP configuration, skills, an
     # MCP API is mounted at /api/mcp
     app.include_router(mcp.router)
 
-    # Authenticated source-context proxy for the WeKnora retrieval skill.
-    app.include_router(weknora_sources.router)
+    # DeerFlow-native LLMWiki management/search BFF (conditionally backed by WeKnora).
+    app.include_router(llmwiki.router)
 
     # Memory API is mounted at /api/memory
     app.include_router(memory.router)
@@ -1622,11 +1622,12 @@ This gateway provides custom endpoints for models, MCP configuration, skills, an
     app.include_router(recommend.router)     # Step 1?2?/api/recommend/stream?? init?
     # Step 3 ????/?????? artifacts.router?/api/threads/{id}/artifacts?
 
-    # ?????????(? auth_middleware._PUBLIC_PATH_PREFIXES)?
-    # ????? ``/health`` ?? API ??? ``/api/health``,?????
-    # ?? base URL ????,?????????? 401?
-    @app.get("/health", tags=["health"])
-    @app.get("/api/health", tags=["health"])
+    # Health probes are always unauthenticated (see
+    # auth_middleware.PUBLIC_HEALTH_PATHS). Support the bare Gateway path,
+    # API base URL, and LangGraph proxy base URL for deployment monitors.
+    @app.api_route("/health", methods=["GET", "HEAD"], tags=["health"])
+    @app.api_route("/api/health", methods=["GET", "HEAD"], tags=["health"])
+    @app.api_route("/api/langgraph/health", methods=["GET", "HEAD"], tags=["health"])
     async def health_check() -> dict:
         """Health check endpoint.
 

offline-backend-20260512/backend/app/gateway/auth_middleware.py

@@ -24,15 +24,25 @@ from app.gateway.internal_auth import INTERNAL_AUTH_HEADER_NAME, get_internal_us
 from app.gateway.proxy_path import app_relative_path
 from deerflow.runtime.user_context import reset_current_user, set_current_user
 
-# Paths that never require authentication.
+# Health probes are deliberately public, including the LangGraph-compatible
+# gateway alias used by some deployments. Keep these as exact paths: health
+# checks must bypass authentication, but similarly named application routes
+# must not inherit that exemption.
+PUBLIC_HEALTH_PATHS: frozenset[str] = frozenset(
+    {
+        "/health",
+        "/api/health",
+        "/api/langgraph/health",
+    }
+)
+
+# Path prefixes that never require authentication.
 #
 # ``/api/public/`` is a namespace reserved for intentionally world-readable,
 # read-only endpoints (e.g. public conversation share snapshots). Only mount
 # safe read-only routes under it — everything there bypasses auth entirely.
-_PUBLIC_PATH_PREFIXES: tuple[str, ...] = (
-    "/health",
-    "/api/health",
-    "/docs",
+_PUBLIC_PATH_PREFIXES: tuple[str, ...] = (
+    "/docs",
     "/redoc",
     "/openapi.json",
     "/api/public/",
@@ -43,9 +53,10 @@ _PUBLIC_PATH_PREFIXES: tuple[str, ...] = (
 
 # Exact auth paths that are public (login/register/status check).
 # /api/v1/auth/me, /api/v1/auth/change-password etc. are NOT public.
-_PUBLIC_EXACT_PATHS: frozenset[str] = frozenset(
-    {
-        "/api/v1/auth/login/local",
+_PUBLIC_EXACT_PATHS: frozenset[str] = frozenset(
+    {
+        *PUBLIC_HEALTH_PATHS,
+        "/api/v1/auth/login/local",
         "/api/v1/auth/login/username",
         "/api/v1/auth/login/token",
         "/api/v1/auth/register",

offline-backend-20260512/backend/app/gateway/deps.py

@@ -293,10 +293,16 @@ async def langgraph_runtime(app: FastAPI) -> AsyncGenerator[None, None]:
 
         app.state.dashboard_session_store = make_dashboard_session_store(sf)
 
-        from deerflow.persistence.task_buttons import make_task_button_store
-
+        from deerflow.persistence.task_buttons import make_task_button_store
+
         app.state.task_button_store = make_task_button_store(sf)
 
+        from deerflow.integrations.weknora.runtime import register_llmwiki_store
+        from deerflow.persistence.llmwiki import make_llmwiki_store
+
+        app.state.llmwiki_store = make_llmwiki_store(sf)
+        register_llmwiki_store(app.state.llmwiki_store)
+
         from deerflow.persistence.taskcop_tasks import make_taskcop_task_store
 
         app.state.taskcop_task_store = make_taskcop_task_store(sf)

offline-backend-20260512/backend/app/gateway/llmwiki_deposit.py

@@ -0,0 +1,241 @@
+"""Asynchronous conversation sedimentation into WeKnora-backed LLMWiki."""
+
+from __future__ import annotations
+
+import asyncio
+import logging
+from dataclasses import dataclass
+from typing import Any
+
+from deerflow.integrations.weknora.client import WeKnoraClient, WeKnoraError
+from deerflow.integrations.weknora.runtime import build_weknora_client, get_resolved_llmwiki_runtime
+from deerflow.persistence.llmwiki import LlmWikiStore
+
+logger = logging.getLogger(__name__)
+
+CONVERSATION_DEPOSIT_KB_NAME = "对话沉淀"
+_DEPOSIT_KB_LOCK = asyncio.Lock()
+
+
+@dataclass(frozen=True)
+class ConversationTurnDeposit:
+    thread_id: str
+    question: str
+    answer: str
+    human_message_id: str | None
+    assistant_message_id: str
+    assistant_id: str | None
+    knowledge_base_ids: list[str]
+    created_by_user_id: str | None
+
+
+def _store(app: Any) -> LlmWikiStore | None:
+    return getattr(app.state, "llmwiki_store", None)
+
+
+def _client(app: Any) -> WeKnoraClient | None:
+    runtime = get_resolved_llmwiki_runtime(getattr(app.state, "config", None))
+    if not runtime.weknora_enabled:
+        return None
+    try:
+        return build_weknora_client(runtime)
+    except RuntimeError:
+        logger.debug("LLMWiki conversation deposit skipped: WeKnora client is unavailable", exc_info=True)
+        return None
+
+
+def _shorten(text: str, max_chars: int) -> str:
+    value = " ".join(str(text or "").split())
+    if len(value) <= max_chars:
+        return value
+    return f"{value[: max_chars - 1].rstrip()}…"
+
+
+def _title_for_turn(turn: ConversationTurnDeposit) -> str:
+    question = _shorten(turn.question, 48)
+    return f"对话沉淀 - {question or turn.assistant_message_id[:12]}"
+
+
+def _format_markdown(turn: ConversationTurnDeposit, title: str) -> str:
+    kb_line = "、".join(turn.knowledge_base_ids) if turn.knowledge_base_ids else "未显式选择知识库"
+    return f"""---
+source: DeerFlow
+type: conversation-deposit
+thread_id: {turn.thread_id}
+human_message_id: {turn.human_message_id or ""}
+assistant_message_id: {turn.assistant_message_id}
+assistant_id: {turn.assistant_id or ""}
+knowledge_base_ids: {kb_line}
+created_by_user_id: {turn.created_by_user_id or ""}
+---
+
+# {title}
+
+> 本页由 DeerFlow 在一轮问答结束后自动沉淀，用于后续 LLMWiki 检索、Wiki 索引与图谱构建。
+
+## 摘要
+
+{_shorten(turn.answer, 360) or "本轮回答为空。"}
+
+## 用户问题
+
+{turn.question.strip() or "（空）"}
+
+## 助手回答
+
+{turn.answer.strip() or "（空）"}
+
+## 结构化线索
+
+- 来源线程：`{turn.thread_id}`
+- 用户消息：`{turn.human_message_id or ""}`
+- 回答消息：`{turn.assistant_message_id}`
+- 智能体：`{turn.assistant_id or "通用问答"}`
+- 本轮显式知识库：{kb_line}
+
+## LLMWiki 标签
+
+#对话沉淀 #DeerFlow #问答知识
+"""
+
+
+async def ensure_conversation_deposit_knowledge_base(
+    app: Any,
+    *,
+    owner_user_id: str | None,
+) -> dict[str, Any] | None:
+    store = _store(app)
+    client = _client(app)
+    if store is None or client is None:
+        return None
+
+    existing = await store.get_conversation_deposit_mapping()
+    if existing is not None:
+        return existing
+
+    async with _DEPOSIT_KB_LOCK:
+        existing = await store.get_conversation_deposit_mapping()
+        if existing is not None:
+            return existing
+
+        remote_id = ""
+        try:
+            for remote in await client.list_knowledge_bases():
+                if str(remote.get("name") or "").strip() == CONVERSATION_DEPOSIT_KB_NAME and remote.get("id"):
+                    remote_id = str(remote["id"])
+                    break
+        except WeKnoraError:
+            logger.debug("Could not scan WeKnora knowledge bases before creating deposit base", exc_info=True)
+
+        if remote_id:
+            mapped = await store.get_by_weknora_id(remote_id)
+            if mapped is not None:
+                if mapped.get("publication_status") != "published":
+                    return await store.request_publish(
+                        str(mapped["id"]),
+                        owner_user_id or "system",
+                        is_admin=True,
+                    )
+                return mapped
+            remote = {"id": remote_id, "name": CONVERSATION_DEPOSIT_KB_NAME, "description": "DeerFlow 全局问答沉淀知识库", "type": "document"}
+        else:
+            try:
+                remote = await client.create_knowledge_base(
+                    name=CONVERSATION_DEPOSIT_KB_NAME,
+                    description="DeerFlow 全局问答沉淀知识库。所有用户可见，不默认参与问答检索。",
+                    kb_type="document",
+                    wiki_enabled=True,
+                )
+            except WeKnoraError:
+                logger.warning("Failed to create WeKnora conversation deposit knowledge base", exc_info=True)
+                return None
+
+        try:
+            row = await store.create_mapping(
+                weknora_id=str(remote["id"]),
+                owner_user_id=owner_user_id or "system",
+                name=CONVERSATION_DEPOSIT_KB_NAME,
+                description=str(remote.get("description") or "DeerFlow 全局问答沉淀知识库"),
+                kb_type="document",
+            )
+            published = await store.request_publish(str(row["id"]), owner_user_id or "system", is_admin=True)
+            return published or row
+        except Exception:
+            logger.exception("Failed to persist conversation deposit knowledge-base mapping")
+            return None
+
+
+async def deposit_conversation_turn(app: Any, turn: ConversationTurnDeposit) -> None:
+    if not turn.thread_id or not turn.assistant_message_id:
+        return
+    if not turn.question.strip() and not turn.answer.strip():
+        return
+
+    store = _store(app)
+    client = _client(app)
+    if store is None or client is None:
+        return
+
+    kb = await ensure_conversation_deposit_knowledge_base(app, owner_user_id=turn.created_by_user_id)
+    if kb is None:
+        return
+
+    previous = await store.get_conversation_deposit(
+        thread_id=turn.thread_id,
+        assistant_message_id=turn.assistant_message_id,
+    )
+    previous_doc_id = str((previous or {}).get("weknora_knowledge_id") or "")
+    if previous_doc_id:
+        try:
+            await client.delete_document(previous_doc_id)
+        except WeKnoraError as exc:
+            if exc.status_code != 404:
+                logger.warning("Failed to delete old conversation deposit document %s", previous_doc_id, exc_info=True)
+
+    title = _title_for_turn(turn)
+    content = _format_markdown(turn, title)
+    try:
+        remote_doc = await client.create_manual_document(
+            str(kb["weknora_id"]),
+            title=title,
+            content=content,
+        )
+    except WeKnoraError:
+        logger.warning("Failed to create conversation deposit document", exc_info=True)
+        return
+
+    await store.upsert_conversation_deposit(
+        thread_id=turn.thread_id,
+        human_message_id=turn.human_message_id,
+        assistant_message_id=turn.assistant_message_id,
+        knowledge_base_mapping_id=str(kb["id"]),
+        weknora_knowledge_id=str(remote_doc.get("id") or remote_doc.get("knowledge_id") or ""),
+        title=title,
+        created_by_user_id=turn.created_by_user_id,
+    )
+
+
+async def delete_conversation_deposit_documents(
+    app: Any,
+    *,
+    thread_id: str,
+    message_ids: list[str] | None = None,
+) -> None:
+    store = _store(app)
+    client = _client(app)
+    if store is None or client is None:
+        return
+    try:
+        rows = await store.delete_conversation_deposits(thread_id=thread_id, message_ids=message_ids)
+    except Exception:
+        logger.debug("Failed to delete local conversation deposit rows", exc_info=True)
+        return
+    for row in rows:
+        doc_id = str(row.get("weknora_knowledge_id") or "")
+        if not doc_id:
+            continue
+        try:
+            await client.delete_document(doc_id)
+        except WeKnoraError as exc:
+            if exc.status_code != 404:
+                logger.warning("Failed to delete WeKnora conversation deposit document %s", doc_id, exc_info=True)

offline-backend-20260512/backend/app/gateway/llmwiki_rag.py

@@ -0,0 +1,290 @@
+"""LLMWiki retrieval helpers for normal DeerFlow chat runs.
+
+The public chat API accepts DeerFlow-local LLMWiki knowledge-base mapping ids.
+When WeKnora mode is enabled, this module resolves those local ids under the
+current DeerFlow user's permissions, searches WeKnora once before the model
+call, and attaches both:
+
+* a hidden prompt with numbered context snippets, and
+* structured citation sources for the existing reference panel.
+
+The helper lives in ``app.gateway`` because it needs Gateway auth state and the
+DeerFlow-owned mapping store.  Harness code remains app-import free.
+"""
+
+from __future__ import annotations
+
+import logging
+from collections.abc import Mapping
+from dataclasses import dataclass
+from typing import Any
+from urllib.parse import urlencode
+
+from fastapi import HTTPException, Request
+from langchain_core.messages import HumanMessage
+
+from app.gateway.deps import get_optional_user_from_request
+from app.gateway.utils import sanitize_log_param
+from deerflow.config.agents_config import load_agent_config, validate_agent_id
+from deerflow.integrations.weknora.client import WeKnoraError
+from deerflow.integrations.weknora.runtime import build_weknora_client, get_resolved_llmwiki_runtime
+from deerflow.persistence.llmwiki import LlmWikiStore
+from deerflow.runtime.user_context import get_effective_user_id
+
+logger = logging.getLogger(__name__)
+
+_MAX_SOURCES = 8
+_MAX_SOURCE_CHARS = 1400
+
+
+@dataclass(frozen=True)
+class LlmWikiRagContext:
+    knowledge_base_ids: list[str]
+    source_count: int
+    prompt: str
+    sources: list[dict[str, Any]]
+    query: str
+
+
+def _store(request: Request) -> LlmWikiStore | None:
+    return getattr(request.app.state, "llmwiki_store", None)
+
+
+async def _actor(request: Request) -> tuple[str, bool]:
+    user = await get_optional_user_from_request(request)
+    if user is None:
+        return get_effective_user_id(), True
+    return str(user.id), getattr(user, "system_role", None) == "admin"
+
+
+def _as_id_list(value: Any) -> list[str]:
+    if value is None:
+        return []
+    if isinstance(value, str):
+        parts = [value]
+    elif isinstance(value, (list, tuple, set)):
+        parts = list(value)
+    else:
+        return []
+    return list(dict.fromkeys(str(item).strip() for item in parts if str(item).strip()))
+
+
+def _body_selected_ids(body: Any) -> list[str]:
+    for source in (getattr(body, "context", None), getattr(body, "metadata", None)):
+        if isinstance(source, Mapping) and "llmwiki_knowledge_base_ids" in source:
+            return _as_id_list(source.get("llmwiki_knowledge_base_ids"))
+    return []
+
+
+def _assistant_agent_id(body: Any) -> str | None:
+    assistant_id = validate_agent_id(getattr(body, "assistant_id", None))
+    if assistant_id:
+        return assistant_id
+    context = getattr(body, "context", None)
+    if isinstance(context, Mapping):
+        return validate_agent_id(str(context.get("agent_id") or "")) if context.get("agent_id") else None
+    metadata = getattr(body, "metadata", None)
+    if isinstance(metadata, Mapping):
+        return validate_agent_id(str(metadata.get("agent_id") or "")) if metadata.get("agent_id") else None
+    return None
+
+
+def _agent_bound_ids(body: Any) -> list[str]:
+    agent_id = _assistant_agent_id(body)
+    if not agent_id:
+        return []
+    try:
+        config = load_agent_config(agent_id)
+    except (FileNotFoundError, ValueError):
+        return []
+    return _as_id_list(config.llmwiki_knowledge_base_ids if config else None)
+
+
+def resolve_requested_llmwiki_knowledge_base_ids(body: Any) -> list[str]:
+    """Return selected ids plus custom-agent defaults, de-duplicated in order."""
+
+    return list(dict.fromkeys([*_agent_bound_ids(body), *_body_selected_ids(body)]))
+
+
+def llmwiki_selection_requested(body: Any) -> bool:
+    return bool(resolve_requested_llmwiki_knowledge_base_ids(body))
+
+
+def _latest_human_query(graph_input: dict[str, Any]) -> str:
+    messages = graph_input.get("messages")
+    if not isinstance(messages, list):
+        return ""
+    for message in reversed(messages):
+        if isinstance(message, HumanMessage):
+            return _message_text(getattr(message, "content", ""))
+        if isinstance(message, dict) and str(message.get("type") or message.get("role") or "") in {"human", "user"}:
+            return _message_text(message.get("content"))
+    return ""
+
+
+def _message_text(content: Any) -> str:
+    if isinstance(content, str):
+        return content.strip()
+    if isinstance(content, list):
+        parts: list[str] = []
+        for item in content:
+            if isinstance(item, str):
+                parts.append(item)
+            elif isinstance(item, Mapping):
+                text = item.get("text") or item.get("content")
+                if isinstance(text, str):
+                    parts.append(text)
+        return "\n".join(part.strip() for part in parts if part and part.strip()).strip()
+    return str(content or "").strip()
+
+
+def _truncate(text: str, max_chars: int) -> str:
+    value = text.strip()
+    if len(value) <= max_chars:
+        return value
+    return f"{value[: max_chars - 1].rstrip()}…"
+
+
+def _source_from_result(result: dict[str, Any], mapping: dict[str, Any], index: int) -> dict[str, Any]:
+    chunk_id = str(result.get("chunk_id") or "")
+    title = str(result.get("title") or result.get("filename") or mapping.get("name") or f"LLMWiki 片段 {index}")
+    content = str(result.get("content") or "")
+    params = urlencode({"knowledge_base_id": mapping["id"], "chunk_id": chunk_id})
+    raw = {
+        **result,
+        "knowledge_base_id": mapping["id"],
+        "knowledge_base_name": mapping.get("name") or "",
+        "source": mapping.get("name") or "LLMWiki",
+        "url": f"/weknora-source-preview?{params}",
+    }
+    return {
+        "index": index,
+        "id": chunk_id or str(result.get("knowledge_id") or f"{mapping['id']}:{index}"),
+        "title": title,
+        "type": "llmwiki",
+        "snippet": _truncate(content, 360),
+        "content": content,
+        "source": mapping.get("name") or "LLMWiki",
+        "score": float(result.get("score") or 0),
+        "url": f"/weknora-source-preview?{params}",
+        "skillName": "llmwiki_search",
+        "mode": "citation",
+        "raw": raw,
+    }
+
+
+def _build_prompt(query: str, sources: list[dict[str, Any]]) -> str:
+    if not sources:
+        return (
+            "当前用户选择了 LLMWiki 知识库，但 DeerFlow 未检索到可用片段。"
+            "请直接说明没有检索到可用背景资料，不要编造知识库内容或引用编号。"
+        )
+
+    lines = [
+        "当前用户选择了 LLMWiki 知识库。DeerFlow 已在模型回答前自动检索到以下背景资料。",
+        "请优先依据这些资料回答；如果资料不足，请明确说明不足之处，不要编造。",
+        "引用规则：正文关键结论必须使用 [1]、[2] 这样的编号引用下方资料；只能引用存在的编号；不要在正文末尾重复输出完整参考文献列表，前端会展示引用卡片。",
+        "",
+        f"用户问题：{query}",
+        "",
+        f"<llmwiki_reference_batch source_count=\"{len(sources)}\">",
+    ]
+    for source in sources:
+        meta = " / ".join(
+            str(item)
+            for item in (
+                source.get("source"),
+                f"score={source.get('score'):.4f}" if isinstance(source.get("score"), float) else None,
+            )
+            if item
+        )
+        lines.append(f"[{source['index']}] {source.get('title') or 'LLMWiki 片段'}")
+        if meta:
+            lines.append(f"来源：{meta}")
+        lines.append(_truncate(str(source.get("content") or source.get("snippet") or ""), _MAX_SOURCE_CHARS))
+        lines.append("")
+    lines.append("</llmwiki_reference_batch>")
+    return "\n".join(lines).strip()
+
+
+async def build_llmwiki_rag_context(
+    *,
+    request: Request,
+    body: Any,
+    graph_input: dict[str, Any],
+) -> LlmWikiRagContext:
+    selected_ids = resolve_requested_llmwiki_knowledge_base_ids(body)
+    if not selected_ids:
+        return LlmWikiRagContext([], 0, "", [], "")
+
+    runtime = get_resolved_llmwiki_runtime(getattr(request.app.state, "config", None))
+    if not runtime.weknora_enabled:
+        return LlmWikiRagContext(selected_ids, 0, "", [], "")
+    store = _store(request)
+    if store is None:
+        return LlmWikiRagContext(selected_ids, 0, "", [], "")
+
+    query = _latest_human_query(graph_input)
+    if not query:
+        return LlmWikiRagContext(selected_ids, 0, "", [], "")
+
+    user_id, is_admin = await _actor(request)
+    rows: list[dict[str, Any]] = []
+    for mapping_id in selected_ids:
+        row = await store.get_authorized(mapping_id, user_id, write=False, is_admin=is_admin)
+        if row is None:
+            raise HTTPException(status_code=404, detail=f"Knowledge base {mapping_id} not found")
+        rows.append(row)
+    remote_to_local = {str(row["weknora_id"]): row for row in rows}
+
+    try:
+        results = await build_weknora_client(runtime).search(query, [str(row["weknora_id"]) for row in rows])
+    except WeKnoraError as exc:
+        logger.warning(
+            "LLMWiki pre-retrieval failed for user=%s kb_count=%s: %s",
+            sanitize_log_param(user_id),
+            len(rows),
+            exc,
+        )
+        return LlmWikiRagContext(
+            selected_ids,
+            0,
+            f"当前用户选择了 LLMWiki 知识库，但自动检索失败：{exc}。请说明暂时无法检索知识库背景资料，不要编造引用。",
+            [],
+            query,
+        )
+
+    sources: list[dict[str, Any]] = []
+    for result in results:
+        mapping = remote_to_local.get(str(result.get("knowledge_base_id") or ""))
+        if mapping is None or not str(result.get("content") or "").strip():
+            continue
+        sources.append(_source_from_result(result, mapping, len(sources) + 1))
+        if len(sources) >= _MAX_SOURCES:
+            break
+    return LlmWikiRagContext(selected_ids, len(sources), _build_prompt(query, sources), sources, query)
+
+
+def attach_llmwiki_rag_to_graph_input(
+    graph_input: dict[str, Any],
+    rag: LlmWikiRagContext,
+) -> dict[str, Any]:
+    if not rag.prompt and not rag.sources:
+        return graph_input
+    messages = graph_input.get("messages")
+    if not isinstance(messages, list) or not messages:
+        return graph_input
+    new_messages = list(messages)
+    for index, message in enumerate(new_messages):
+        if not isinstance(message, HumanMessage):
+            continue
+        extra = dict(getattr(message, "additional_kwargs", None) or {})
+        existing = str(extra.get("prompt_prefix") or "").strip()
+        extra["prompt_prefix"] = f"{existing}\n\n{rag.prompt}".strip() if existing else rag.prompt
+        extra["llmwiki_knowledge_base_ids"] = rag.knowledge_base_ids
+        extra["llmwiki_reference_query"] = rag.query
+        if rag.sources:
+            extra["llmwiki_reference_sources"] = rag.sources
+        new_messages[index] = message.model_copy(update={"additional_kwargs": extra})
+        break
+    return {**graph_input, "messages": new_messages}

offline-backend-20260512/backend/app/gateway/routers/agents.py

@@ -64,10 +64,15 @@ class AgentResponse(BaseModel):
         description="入库模板名 (external knowledge-ingest templateName). When set, markdown artifacts produced in this agent's chats can be published to the external knowledge-ingest API.",
     )
 
-    seat_skill_directive: bool | None = Field(
+    seat_skill_directive: bool | None = Field(
         default=None,
         description="会商席位「技能识别强化」开关 (config.yaml). True = 本智能体作为会商席位时强化技能识别 (与业务链条同名开关取 OR). None/False = 关 (默认).",
-    )
+    )
+
+    llmwiki_knowledge_base_ids: list[str] | None = Field(
+        default=None,
+        description="DeerFlow LLMWiki mapping ids this agent may offer for conversation retrieval.",
+    )
 
     tags: list[dict] = Field(default_factory=list, description="Tags assigned to this agent")
 
@@ -120,7 +125,12 @@ class AgentCreateRequest(BaseModel):
     tool_groups: list[str] | None = Field(default=None, description="Optional tool group whitelist persisted to config.yaml")
     common_questions: list[CommonQuestion] | None = Field(default=None, description="Recommended questions (常用问题) persisted to config.yaml. Each item is {title, prompt}; legacy strings are coerced.")
     template_name: str | None = Field(default=None, description="入库模板名 persisted to config.yaml. Enables publishing markdown artifacts to the external knowledge-ingest API.")
-    seat_skill_directive: bool | None = Field(default=None, description="会商席位「技能识别强化」开关 persisted to config.yaml. None/False = off (默认).")
+    seat_skill_directive: bool | None = Field(default=None, description="会商席位「技能识别强化」开关 persisted to config.yaml. None/False = off (默认).")
+    llmwiki_knowledge_base_ids: list[str] | None = Field(
+        default=None,
+        max_length=100,
+        description="Allowed DeerFlow LLMWiki mapping ids persisted to config.yaml.",
+    )
 
 
 class AgentUpdateRequest(BaseModel):
@@ -137,7 +147,12 @@ class AgentUpdateRequest(BaseModel):
     tool_groups: list[str] | None = Field(default=None, description="Updated tool group whitelist (config.yaml)")
     common_questions: list[CommonQuestion] | None = Field(default=None, description="Updated recommended questions (常用问题, config.yaml). Each item is {title, prompt}; legacy strings are coerced.")
     template_name: str | None = Field(default=None, description="Updated 入库模板名 (config.yaml). Empty string clears it.")
-    seat_skill_directive: bool | None = Field(default=None, description="Updated 会商席位「技能识别强化」开关 (config.yaml).")
+    seat_skill_directive: bool | None = Field(default=None, description="Updated 会商席位「技能识别强化」开关 (config.yaml).")
+    llmwiki_knowledge_base_ids: list[str] | None = Field(
+        default=None,
+        max_length=100,
+        description="Updated allowed DeerFlow LLMWiki mapping ids (config.yaml).",
+    )
 
 
 class AgentSelectorResponse(BaseModel):
@@ -178,10 +193,38 @@ def _current_user_id(request: Request) -> str:
     return get_effective_user_id()
 
 
-def _is_admin_user(request: Request) -> bool:
+def _is_admin_user(request: Request) -> bool:
     """Return True when the request is authenticated as a system admin."""
     user = getattr(request.state, "user", None)
-    return getattr(user, "system_role", None) == "admin"
+    return getattr(user, "system_role", None) == "admin"
+
+
+async def _validate_llmwiki_bindings(
+    request: Request,
+    mapping_ids: list[str] | None,
+) -> list[str] | None:
+    """Normalize bindings and require read access for the editing actor."""
+
+    if mapping_ids is None:
+        return None
+    normalized = list(dict.fromkeys(item.strip() for item in mapping_ids if item.strip()))
+    if not normalized:
+        return []
+    store = getattr(request.app.state, "llmwiki_store", None)
+    if store is None:
+        raise HTTPException(status_code=503, detail="LLMWiki metadata store is unavailable")
+    user_id = _current_user_id(request)
+    is_admin = _is_admin_user(request)
+    for mapping_id in normalized:
+        row = await store.get_authorized(
+            mapping_id,
+            user_id,
+            write=False,
+            is_admin=is_admin,
+        )
+        if row is None:
+            raise HTTPException(status_code=422, detail=f"LLMWiki knowledge base '{mapping_id}' is unavailable")
+    return normalized
 
 
 def _email_local_part(email: str | None) -> str | None:
@@ -233,7 +276,7 @@ def _validate_storage_id(agent_id: str) -> str:
         raise HTTPException(status_code=422, detail=str(exc)) from exc
 
 
-def _write_agent_config(
+def _write_agent_config(
     agent_id: str,
     *,
     name: str,
@@ -243,8 +286,9 @@ def _write_agent_config(
     tool_groups: list[str] | None = None,
     common_questions: list[CommonQuestion] | None = None,
     template_name: str | None = None,
-    seat_skill_directive: bool | None = None,
-) -> None:
+    seat_skill_directive: bool | None = None,
+    llmwiki_knowledge_base_ids: list[str] | None = None,
+) -> None:
     agent_dir = get_paths().agent_dir(agent_id)
     agent_dir.mkdir(parents=True, exist_ok=True)
     config_file = agent_dir / "config.yaml"
@@ -273,15 +317,19 @@ def _write_agent_config(
     if template_name is not None and template_name.strip():
         config_data["template_name"] = template_name.strip()
     # 会商席位技能识别强化开关：仅当显式设置时写入（None = 不写, 由 AgentConfig 默认为关）。
-    if seat_skill_directive is not None:
-        config_data["seat_skill_directive"] = bool(seat_skill_directive)
+    if seat_skill_directive is not None:
+        config_data["seat_skill_directive"] = bool(seat_skill_directive)
+    if llmwiki_knowledge_base_ids is not None:
+        config_data["llmwiki_knowledge_base_ids"] = list(
+            dict.fromkeys(item.strip() for item in llmwiki_knowledge_base_ids if item.strip())
+        )
     with open(config_file, "w", encoding="utf-8") as f:
         yaml.dump(config_data, f, default_flow_style=False, allow_unicode=True)
 
 
 def _load_config_extras(agent_id: str) -> dict[str, Any]:
     """Read skills/model/tool_groups/common_questions from the agent's config.yaml, tolerant of missing files."""
-    empty = {"skills": None, "model": None, "tool_groups": None, "common_questions": None, "template_name": None, "seat_skill_directive": None}
+    empty = {"skills": None, "model": None, "tool_groups": None, "common_questions": None, "template_name": None, "seat_skill_directive": None, "llmwiki_knowledge_base_ids": None}
     try:
         cfg = load_agent_config(agent_id)
     except (FileNotFoundError, ValueError):
@@ -294,7 +342,8 @@ def _load_config_extras(agent_id: str) -> dict[str, Any]:
         "tool_groups": cfg.tool_groups,
         "common_questions": cfg.common_questions,
         "template_name": cfg.template_name,
-        "seat_skill_directive": cfg.seat_skill_directive,
+        "seat_skill_directive": cfg.seat_skill_directive,
+        "llmwiki_knowledge_base_ids": cfg.llmwiki_knowledge_base_ids,
     }
 
 
@@ -455,7 +504,7 @@ def _record_to_response(
             can happen lazily when the agent is actually opened.
     """
     soul: str | None = None
-    extras: dict[str, Any] = {"skills": None, "model": None, "tool_groups": None, "common_questions": None, "template_name": None, "seat_skill_directive": None}
+    extras: dict[str, Any] = {"skills": None, "model": None, "tool_groups": None, "common_questions": None, "template_name": None, "seat_skill_directive": None, "llmwiki_knowledge_base_ids": None}
     if include_extras:
         _ensure_agent_dir_from_record(record)
         extras = _load_config_extras(record["id"])
@@ -479,7 +528,8 @@ def _record_to_response(
         skills=extras["skills"],
         common_questions=extras.get("common_questions"),
         template_name=extras.get("template_name"),
-        seat_skill_directive=extras.get("seat_skill_directive"),
+        seat_skill_directive=extras.get("seat_skill_directive"),
+        llmwiki_knowledge_base_ids=extras.get("llmwiki_knowledge_base_ids"),
         featured_order=record.get("featured_order"),
     )
 
@@ -780,7 +830,7 @@ async def get_agent(agent_id: str, request: Request) -> AgentResponse:
     summary="Create Agent",
     description="Create a new user-owned agent. The display name can be Chinese and may duplicate existing names.",
 )
-async def create_agent_endpoint(request: AgentCreateRequest, http_request: Request) -> AgentResponse:
+async def create_agent_endpoint(request: AgentCreateRequest, http_request: Request) -> AgentResponse:
     _require_agents_api_enabled()
     name = _validate_display_name(request.name)
     store = get_agent_store(http_request)
@@ -789,9 +839,13 @@ async def create_agent_endpoint(request: AgentCreateRequest, http_request: Reque
         agent_id = _validate_storage_id(request.id)
         if await store.get_any(agent_id) is not None:
             raise HTTPException(status_code=409, detail=f"Agent id '{agent_id}' already exists")
-    else:
-        agent_id = uuid4().hex
-    agent_dir = get_paths().agent_dir(agent_id)
+    else:
+        agent_id = uuid4().hex
+    agent_dir = get_paths().agent_dir(agent_id)
+    llmwiki_ids = await _validate_llmwiki_bindings(
+        http_request,
+        request.llmwiki_knowledge_base_ids,
+    )
 
     try:
         if agent_dir.exists():
@@ -806,7 +860,8 @@ async def create_agent_endpoint(request: AgentCreateRequest, http_request: Reque
             tool_groups=request.tool_groups,
             common_questions=request.common_questions,
             template_name=request.template_name,
-            seat_skill_directive=request.seat_skill_directive,
+            seat_skill_directive=request.seat_skill_directive,
+            llmwiki_knowledge_base_ids=llmwiki_ids,
         )
         _write_agent_soul(agent_id, request.soul)
 
@@ -887,9 +942,14 @@ async def update_agent(agent_id: str, request: AgentUpdateRequest, http_request:
         next_tool_groups = request.tool_groups if "tool_groups" in fields_set else existing_extras["tool_groups"]
         next_common_questions = request.common_questions if "common_questions" in fields_set else existing_extras["common_questions"]
         next_template_name = request.template_name if "template_name" in fields_set else existing_extras["template_name"]
-        next_seat_skill_directive = request.seat_skill_directive if "seat_skill_directive" in fields_set else existing_extras["seat_skill_directive"]
-
-        config_dirty = bool(update_data) or any(key in fields_set for key in ("skills", "model", "tool_groups", "common_questions", "template_name", "seat_skill_directive"))
+        next_seat_skill_directive = request.seat_skill_directive if "seat_skill_directive" in fields_set else existing_extras["seat_skill_directive"]
+        next_llmwiki_ids = (
+            await _validate_llmwiki_bindings(http_request, request.llmwiki_knowledge_base_ids)
+            if "llmwiki_knowledge_base_ids" in fields_set
+            else existing_extras["llmwiki_knowledge_base_ids"]
+        )
+
+        config_dirty = bool(update_data) or any(key in fields_set for key in ("skills", "model", "tool_groups", "common_questions", "template_name", "seat_skill_directive", "llmwiki_knowledge_base_ids"))
         if config_dirty:
             _write_agent_config(
                 agent_id,
@@ -900,7 +960,8 @@ async def update_agent(agent_id: str, request: AgentUpdateRequest, http_request:
                 tool_groups=next_tool_groups,
                 common_questions=next_common_questions,
                 template_name=next_template_name,
-                seat_skill_directive=next_seat_skill_directive,
+                seat_skill_directive=next_seat_skill_directive,
+                llmwiki_knowledge_base_ids=next_llmwiki_ids,
             )
             # Keep the listing cache in lock-step with config.yaml, but only
             # when the cached fields were actually touched — a partial update

offline-backend-20260512/backend/app/gateway/routers/threads.py

@@ -12,11 +12,12 @@ matching the LangGraph Platform wire format expected by the
 
 from __future__ import annotations
 
-import logging
+import asyncio
 import json
+import logging
 import os
-from pathlib import Path
 import uuid
+from pathlib import Path
 from typing import Any
 
 from fastapi import APIRouter, HTTPException, Request
@@ -26,6 +27,7 @@ from sqlalchemy import select
 
 from app.gateway.authz import require_permission
 from app.gateway.deps import get_checkpointer
+from app.gateway.llmwiki_deposit import delete_conversation_deposit_documents
 from app.gateway.utils import sanitize_log_param
 from deerflow.config.extensions_config import SkillDisplayConfig
 from deerflow.config.paths import Paths, get_paths
@@ -34,10 +36,12 @@ from deerflow.persistence.skills.model import SkillDisplayAdapterRow
 from deerflow.runtime import serialize_channel_values
 from deerflow.runtime.references import (
     adapt_reference_batches_with_displays,
-    build_reference_batches as build_runtime_reference_batches,
     build_reference_debug_payload,
     build_reference_debug_summary,
 )
+from deerflow.runtime.references import (
+    build_reference_batches as build_runtime_reference_batches,
+)
 from deerflow.runtime.user_context import get_effective_user_id
 from deerflow.utils.time import coerce_iso, now_iso
 
@@ -1250,6 +1254,13 @@ async def clear_thread_messages(thread_id: str, request: Request) -> dict[str, A
         len(messages),
         sanitize_log_param(thread_id),
     )
+    asyncio.create_task(
+        delete_conversation_deposit_documents(
+            request.app,
+            thread_id=thread_id,
+            message_ids=None,
+        )
+    )
     return {
         "ok": True,
         "deleted_ids": deleted_ids,
@@ -1345,6 +1356,13 @@ async def delete_thread_message(thread_id: str, message_id: str, request: Reques
         len(messages) - len(filtered),
         sanitize_log_param(thread_id),
     )
+    asyncio.create_task(
+        delete_conversation_deposit_documents(
+            request.app,
+            thread_id=thread_id,
+            message_ids=deleted_ids,
+        )
+    )
     return {"ok": True, "deleted_ids": deleted_ids, "remaining": len(filtered)}
 
 

offline-backend-20260512/backend/app/gateway/services.py

@@ -21,6 +21,11 @@ from fastapi import HTTPException, Request
 from langchain_core.messages import HumanMessage
 
 from app.gateway.deps import get_concurrency_gate, get_run_context, get_run_manager, get_stream_bridge
+from app.gateway.llmwiki_rag import (
+    attach_llmwiki_rag_to_graph_input,
+    build_llmwiki_rag_context,
+    llmwiki_selection_requested,
+)
 from app.gateway.utils import sanitize_log_param
 from deerflow.config import get_app_config
 from deerflow.config.agents_config import load_agent_config, validate_agent_id
@@ -275,6 +280,8 @@ def _prompt_prefix_opted_in(body: Any) -> bool:
 
 
 def _rag_mode_requested(body: Any) -> bool:
+    if llmwiki_selection_requested(body):
+        return True
     context = getattr(body, "context", None)
     if isinstance(context, Mapping) and context.get("rag_mode_enabled"):
         return True
@@ -325,6 +332,7 @@ _CONTEXT_CONFIGURABLE_KEYS: frozenset[str] = frozenset(
         "thinking_force_disabled",
         "reasoning_effort",
         "rag_mode_enabled",
+        "llmwiki_knowledge_base_ids",
         "writing_mode",
         "writing_artifact_path",
         "canvas_artifact_title",
@@ -1003,6 +1011,27 @@ async def start_run(
         prefix_settings = load_system_settings().prompt_prefix
         if prefix_settings.enabled and prefix_settings.prompt_prefix.strip():
             graph_input = _prepend_prompt_prefix(graph_input, prefix_settings.prompt_prefix.strip())
+    llmwiki_rag = await build_llmwiki_rag_context(
+        request=request,
+        body=body,
+        graph_input=graph_input,
+    )
+    if llmwiki_rag.prompt or llmwiki_rag.sources:
+        graph_input = attach_llmwiki_rag_to_graph_input(graph_input, llmwiki_rag)
+        effective_rag_mode = True
+        run_metadata["llmwiki_knowledge_base_ids"] = llmwiki_rag.knowledge_base_ids
+        run_metadata["llmwiki_rag_source_count"] = llmwiki_rag.source_count
+        try:
+            await run_ctx.thread_store.update_metadata(
+                thread_id,
+                {
+                    "llmwiki_knowledge_base_ids": llmwiki_rag.knowledge_base_ids,
+                    "rag_mode_enabled": True,
+                    "rag_mode": True,
+                },
+            )
+        except Exception:
+            logger.debug("Failed to persist LLMWiki RAG metadata for %s", sanitize_log_param(thread_id), exc_info=True)
     if effective_rag_mode:
         graph_input = _append_invisible_prompt(graph_input, _RAG_CITATION_PROMPT)
     config = build_run_config(thread_id, body.config, body.metadata, assistant_id=body.assistant_id)
@@ -1014,6 +1043,8 @@ async def start_run(
     body_context = dict(getattr(body, "context", None) or {})
     if effective_rag_mode:
         body_context["rag_mode_enabled"] = True
+    if "llmwiki_knowledge_base_ids" not in body_context and llmwiki_rag.knowledge_base_ids:
+        body_context["llmwiki_knowledge_base_ids"] = llmwiki_rag.knowledge_base_ids
     # Promote whitelisted top-level body fields into the context map so
     # callers can pass them either nested under ``context`` or as flat
     # top-level fields (``excluded_tools``, ``skill_stop_names``, …).

offline-backend-20260512/backend/config.yaml

@@ -258,8 +258,20 @@ tool_groups:
 tools:
   - name: web_search
     group: web
-    use: deerflow.community.ddg_search.tools:web_search_tool
+    use: deerflow.community.configurable_search.tools:web_search_tool
+    # Keep disabled until the deployment-specific endpoint is filled in.
+    # When false, web_search is not registered with the agent at all.
+    enabled: false
+    endpoint: "https://xxx.xxx.xxx/searchName"
+    verify_ssl: false
+    timeout: 30
     max_results: 5
+    result_url_template: "https://xxx.xxx.xxx/mindId/xxx?recUuid={recUuid}"
+    # The runtime query always overwrites payload.query; all other fields are
+    # sent unchanged so deployment-specific request parameters stay in config.
+    payload:
+      query: ""
+      xxx: "xxxx"
 
   - name: web_fetch
     group: web
@@ -478,6 +490,21 @@ ai_writing:
     writer: false # 作家（逐章正文写作，逐章流式与严格模式求助逻辑不变）
     editor: false # 编辑（事实/逻辑/语言三维度审查）
 
+# Optional independently deployed WeKnora integration. DeerFlow owns only
+# these connection values; database/Redis/storage/parser/model settings stay
+# in WeKnora's own docker-compose/.env/config files. A non-empty api_base_url
+# enables WeKnora mode; leaving it empty preserves the existing LLMWiki iframe.
+# admin_email/admin_password are used only by DeerFlow's server-side WeKnora
+# detail-page iframe proxy. Prefer environment variables in production, e.g.
+# admin_email: "$WEKNORA_ADMIN_EMAIL"
+# admin_password: "$WEKNORA_ADMIN_PASSWORD"
+llmwiki:
+  weknora:
+    api_base_url: "http://119.254.86.251:8088/deerflow-api"
+    web_base_url: "http://119.254.86.251:8088"
+    admin_email: "lqq@qq.com"
+    admin_password: "$WEKNORA_ADMIN_PASSWORD"
+
 log_system:
   enabled: true
   url: 'https://your-log-system.example.com'

offline-backend-20260512/backend/docs/LLMWIKI_WEKNORA_INTEGRATION_ZH.md

@@ -0,0 +1,105 @@
+# LLMWiki × WeKnora 集成与部署
+
+## 目标与边界
+
+DeerFlow 只把 WeKnora 当作独立部署的知识抽取、向量化和检索服务，不复制或修改 WeKnora 前端，也不接管它的运行配置。
+
+本版本按 WeKnora `v0.7.2` 的官方 REST API 实现；升级 WeKnora 大版本前应先回归下文列出的主要接口。
+
+- WeKnora 自己负责 PostgreSQL、Redis、对象存储、解析器、向量库、Embedding/Rerank/LLM 等依赖。
+- DeerFlow 只保存 WeKnora 连接地址，以及 DeerFlow 用户、发布状态、智能体绑定与 WeKnora 知识库之间的映射。
+- DeerFlow 不保存也不发送 WeKnora API Key、Bearer Token 或其他 WeKnora 凭证；浏览器拿不到原始 WeKnora 知识库 ID。
+- 用户权限、智能体所有权和页面风格全部沿用 DeerFlow；库所有者发布后立即进入公共 LLMWiki，不经过审核队列。
+
+## 模式开关
+
+`config.yaml` 中 `llmwiki.weknora.api_base_url` 非空时启用 WeKnora 模式；为空时继续使用原有 LLMWiki 页面和流程。
+
+```yaml
+llmwiki:
+  weknora:
+    api_base_url: "http://weknora-app:8080"
+    web_base_url: "http://weknora-app:8080"
+```
+
+管理员也可在 DeerFlow 原 LLMWiki 页面中的“集成配置”修改地址。页面配置是运行时覆盖项：启用覆盖并填写地址会切换到 WeKnora；启用覆盖但留空会强制回到原模式；“恢复配置文件设置”会重新采用 `config.yaml`。
+
+这两个地址是 DeerFlow 中全部的 WeKnora 配置，不需要创建密钥文件，也不需要在 DeerFlow 中配置 WeKnora 的数据库、模型或存储参数。
+
+## WeKnora 侧准备
+
+先按 WeKnora 官方部署方式完整启动 WeKnora。数据库、Redis、MinIO/S3、解析服务和所有模型均在 WeKnora 自己的 compose、`.env` 或管理页面配置，DeerFlow 无对应字段。
+
+DeerFlow 对 WeKnora 使用无凭证的内部 HTTP 请求，因此 `api_base_url` 必须指向可信内网服务，不能把匿名接口直接暴露到公网。需要在 WeKnora 自身或前置反向代理中，仅按容器网络、源 IP、安全组或 mTLS 等基础设施边界放行 DeerFlow。若 WeKnora 保持默认接口鉴权且未放行，连接测试会收到 401/403，并提示配置 WeKnora 或反向代理；DeerFlow 不提供凭证回退。
+
+WeKnora 中至少需要一个可用且标为默认的 Embedding 模型。创建知识库时，DeerFlow 调用 WeKnora `/api/v1/models` 查找 `type=Embedding` 的默认模型，并把它的 ID 传给 WeKnora；因此 DeerFlow 不提供 `default_embedding_model_id` 配置。这个 ID 的含义是“新知识库使用哪个编码模型生成向量”，其生命周期属于 WeKnora。
+
+WeKnora 侧放行范围需要覆盖本版本调用的模型读取、知识库管理、文档写入和知识检索接口。如果基础设施只放行检索接口，可以检索已有知识库，但无法通过 DeerFlow 创建知识库或上传文档。
+
+## 数据与权限模型
+
+DeerFlow 的 `llmwiki_knowledge_bases` 表保存以下控制面数据：DeerFlow 映射 ID、WeKnora ID、所有者用户 ID、名称快照和公开状态。文档、切片、向量和模型参数仍只存在 WeKnora。
+
+- 个人库：仅所有者和 DeerFlow 管理员可管理。
+- 公共库：所有者点击发布后立即公开，全部 DeerFlow 用户只读可检索；所有者可随时取消公开。
+- 删除个人库时先删除 WeKnora 实体，再删除 DeerFlow 映射。
+- 所有 BFF 接口和智能体工具都会按当前 DeerFlow 登录用户重新鉴权，不能靠前端提交任意知识库 ID 越权。
+
+当前无凭证/反向代理方案使用一个固定的 WeKnora 服务身份。因此 DeerFlow 的用户隔离是映射层的逻辑隔离：不同 DeerFlow 用户只能看到自己拥有的库和已公开库；但这些库在 WeKnora 物理上都创建于该服务身份所属的同一个空间。上传文件不会落到 DeerFlow 用户目录，而是经 DeerFlow BFF 转发给 WeKnora，由 WeKnora 把文件对象、知识元数据、切片和向量分别写入自己的对象存储、数据库和向量存储。需要在 WeKnora 后台核对时，应登录这个服务身份所属空间，在“知识库管理”中按同名知识库进入查看。
+
+如果要求 WeKnora 本身也按 DeerFlow 用户划分空间，则需要另一种部署：为每个 DeerFlow 用户建立 WeKnora 用户/租户并安全保存、轮换各自凭证。它与当前“不在 DeerFlow 配置 WeKnora 凭证”的单服务身份方案不同，不能只靠前端切换实现。
+
+## 智能体和对话选择
+
+智能体 `config.yaml` 可保存：
+
+```yaml
+llmwiki_knowledge_base_ids:
+  - "deerflow-mapping-id-1"
+  - "deerflow-mapping-id-2"
+```
+
+这里必须是 DeerFlow 映射 ID，不是 WeKnora 原始 ID。智能体创建页和详情编辑页均提供内嵌的“绑定知识空间”面板，可直接搜索、分组、多选、全选或清空当前用户有权读取的个人库和公共库，不再通过额外弹窗配置。
+
+新建智能体对话时，页面只展示“智能体已绑定”且“当前登录用户仍有权读取”的交集，默认全选，可在发送第一条消息前调整或清空。普通新建对话也可选择知识库；选中了智能体时会自动切换成该智能体的绑定范围。问答选择器仿照 WeKnora 的交互结构，作为输入框工具栏的锚定下拉呈现，包含搜索、个人/公共分组、文档数量、多选、全选和清空，不再在输入框外单独占据一行。选择结果写入 LangGraph thread metadata 和运行 context，重新打开对话会恢复，开始对话后不可中途扩大权限范围。
+
+`llmwiki_search` 是只读工具，只在 WeKnora 模式注册。每次调用会同时校验：
+
+1. 当前对话选择；
+2. 智能体允许的绑定范围；
+3. 当前 DeerFlow 用户对映射的实时读取权限。
+
+明确选择空列表表示本次对话不使用 LLMWiki；旧对话没有选择字段时，智能体对话默认使用其绑定范围，通用 Lead Agent 默认使用当前用户全部可见知识库。
+
+## 前端实现
+
+WeKnora 前端不做任何修改。DeerFlow 原“知识库资源管理”入口根据运行模式渲染：
+
+- 原模式：保留现有 LLMWiki iframe 页面；
+- WeKnora 模式：渲染 DeerFlow 原生 React 页面，提供个人库、公共库、直接发布和检索。点击卡片进入独立知识库路由，内容区复刻 WeKnora 的详情页结构，同时保留 DeerFlow 的导航、登录和权限壳。Wiki 库按 **Wiki → 文档 → 图谱** 排列并默认打开 Wiki；标准库不显示 Wiki 标签，按 **文档 → 图谱** 排列。Wiki 视图包含页面目录、搜索、页面正文和引用摘要；文档以卡片网格展示，点击后从右侧抽屉查看元数据、原文件与真实解析分块；图谱视图渲染 WeKnora Wiki 页面节点和边。新建知识库可选择“Wiki 知识库”或“标准知识库”，详情页支持文件、网页 URL、手工 Markdown 三种导入方式。问答选择器沿用 WeKnora 的紧凑搜索/分组/多选交互，颜色、间距和组件沿用 DeerFlow 设计系统；智能体创建和编辑页使用同一数据源的内嵌绑定面板。
+
+问答检索命中以 DeerFlow 来源卡片展示。“查看原文”调用 DeerFlow 的 `/api/llmwiki/sources/context`，由服务端在重新校验 DeerFlow 用户权限后向 WeKnora 读取命中切片及相邻上下文，不向浏览器暴露 WeKnora 原始资源 ID。
+
+## 主要接口
+
+- `GET /api/llmwiki/runtime`：当前原模式/WeKnora 模式。
+- `/api/system-settings/llmwiki`：管理员地址配置、重置和连通性测试。
+- `/api/llmwiki/knowledge-bases`：知识库 CRUD、标准/Wiki 模式创建、详情、文档上传、解析、直接发布和取消公开。
+- `POST /api/llmwiki/knowledge-bases/{id}/documents/url`：导入网页 URL。
+- `POST /api/llmwiki/knowledge-bases/{id}/documents/manual`：发布手工 Markdown 内容。
+- `GET /api/llmwiki/knowledge-bases/{id}/documents/{document_id}`：具体文档元数据。
+- `GET /api/llmwiki/knowledge-bases/{id}/documents/{document_id}/preview`：鉴权后代理原文件预览。
+- `GET /api/llmwiki/knowledge-bases/{id}/documents/{document_id}/chunks`：分页读取解析后的真实分块索引。
+- `GET /api/llmwiki/knowledge-bases/{id}/graph`：读取 Wiki 页面图谱的节点和边；未开启 Wiki 索引时返回明确的空状态。
+- `GET /api/llmwiki/knowledge-bases/{id}/wiki/pages` 与 `.../wiki/pages/{slug}`：读取 Wiki 页面目录和正文。
+- `GET /api/llmwiki/agents/{agent_id}/knowledge-bases`：当前用户可用于该智能体的绑定交集。
+- `POST /api/llmwiki/search`：页面检索。
+- `GET /api/llmwiki/sources/context`：引用原文上下文。
+
+## 验收建议
+
+1. 地址留空启动，确认原 LLMWiki 页面和原有流程不变。
+2. 在可信网络中放行 DeerFlow，配置地址，测试连接并创建个人库、上传文档、等待解析完成；确认 DeerFlow 请求没有认证头。
+3. 普通用户直接发布，换另一用户确认能立即读取和检索、但不能修改；再由所有者取消公开并确认访问随即失效。
+4. 给智能体绑定两个库，新建对话取消其中一个，确认引用只来自保留的库；重新打开对话确认选择恢复。
+5. 撤销公共发布或删除用户权限后继续提问，确认工具实时拒绝已失效的知识库。

offline-backend-20260512/backend/extensions_config.json

@@ -1,24 +1,5 @@
 {
-  "mcpServers": {
-    "weknora-retrieval": {
-      "enabled": true,
-      "type": "stdio",
-      "command": "uv",
-      "args": [
-        "run",
-        "--no-sync",
-        "python",
-        "skills/custom/weknora-knowledge-search/scripts/weknora_retrieval_mcp.py"
-      ],
-      "env": {
-        "WEKNORA_BASE_URL": "http://[::1]:8081",
-        "WEKNORA_DEERFLOW_SHARE_TOKEN_FILE": "F:\\react01\\deeflow-code\\deerflow-server\\offline-backend-20260512\\backend\\.runtime\\secrets\\weknora_deerflow_share_token",
-        "WEKNORA_ALLOWED_KB_IDS": "",
-        "WEKNORA_REQUEST_TIMEOUT_SECONDS": "30",
-        "WEKNORA_RERANK_URL": "http://47.88.25.99:7022"
-      }
-    }
-  },
+  "mcpServers": {},
   "skills": {
     "web-research": {
       "enabled": true
@@ -49,9 +30,6 @@
     },
     "browser_act": {
       "enabled": true
-    },
-    "weknora-knowledge-search": {
-      "enabled": true
     }
   }
 }

offline-backend-20260512/backend/packages/harness/deerflow/community/configurable_search/__init__.py

@@ -0,0 +1,3 @@
+from .tools import web_search_tool
+
+__all__ = ["web_search_tool"]

offline-backend-20260512/backend/packages/harness/deerflow/community/configurable_search/tools.py

@@ -0,0 +1,202 @@
+"""Configurable JSON-over-HTTP implementation of the ``web_search`` tool."""
+
+from __future__ import annotations
+
+import json
+import logging
+from copy import deepcopy
+from dataclasses import dataclass
+from typing import Any
+from urllib.parse import quote
+
+import httpx
+from langchain.tools import tool
+
+from deerflow.config import get_app_config
+from deerflow.config.tool_config import ToolConfig
+
+logger = logging.getLogger(__name__)
+
+
+@dataclass(frozen=True)
+class ConfigurableSearchSettings:
+    """Validated runtime settings read from the ``web_search`` tool entry."""
+
+    enabled: bool
+    endpoint: str
+    payload: dict[str, Any]
+    verify_ssl: bool
+    timeout: float
+    max_results: int
+    result_url_template: str
+
+    @classmethod
+    def from_tool_config(cls, config: ToolConfig | None) -> ConfigurableSearchSettings:
+        if config is None:
+            raise ValueError("web_search is not configured")
+
+        extra = config.model_extra or {}
+        endpoint = str(extra.get("endpoint") or "").strip()
+        if not endpoint.startswith(("http://", "https://")):
+            raise ValueError("web_search.endpoint must use http:// or https://")
+
+        raw_payload = extra.get("payload", {})
+        if not isinstance(raw_payload, dict):
+            raise ValueError("web_search.payload must be an object")
+
+        try:
+            timeout = float(extra.get("timeout", 30))
+        except (TypeError, ValueError) as exc:
+            raise ValueError("web_search.timeout must be a number") from exc
+        if timeout <= 0:
+            raise ValueError("web_search.timeout must be greater than zero")
+
+        try:
+            max_results = int(extra.get("max_results", 5))
+        except (TypeError, ValueError) as exc:
+            raise ValueError("web_search.max_results must be an integer") from exc
+        if max_results <= 0:
+            raise ValueError("web_search.max_results must be greater than zero")
+
+        result_url_template = str(extra.get("result_url_template") or "").strip()
+        if result_url_template and "{recUuid}" not in result_url_template:
+            raise ValueError("web_search.result_url_template must contain {recUuid}")
+
+        verify_ssl = extra.get("verify_ssl", True)
+        if not isinstance(verify_ssl, bool):
+            raise ValueError("web_search.verify_ssl must be true or false")
+
+        return cls(
+            enabled=config.enabled,
+            endpoint=endpoint,
+            payload=deepcopy(raw_payload),
+            verify_ssl=verify_ssl,
+            timeout=timeout,
+            max_results=max_results,
+            result_url_template=result_url_template,
+        )
+
+
+def _text(value: Any) -> str:
+    """Normalize optional scalar result fields without leaking ``None``."""
+
+    if value is None:
+        return ""
+    if isinstance(value, str):
+        return value
+    return str(value)
+
+
+def _error_result(query: str, message: str) -> str:
+    return json.dumps(
+        {
+            "query": query,
+            "total_results": 0,
+            "results": [],
+            "error": message,
+        },
+        ensure_ascii=False,
+    )
+
+
+def _normalize_results(data: Any, settings: ConfigurableSearchSettings) -> list[dict[str, str]]:
+    if not isinstance(data, dict):
+        raise ValueError("search API response must be a JSON object")
+
+    raw_results = data.get("results")
+    if not isinstance(raw_results, list):
+        raise ValueError("search API response field 'results' must be an array")
+
+    normalized: list[dict[str, str]] = []
+    for item in raw_results[: settings.max_results]:
+        if not isinstance(item, dict):
+            logger.warning("Ignoring non-object search result: %r", item)
+            continue
+
+        rec_uuid = _text(item.get("recUuid")).strip()
+        original_url = _text(item.get("url")).strip()
+        result_url = original_url
+        if rec_uuid and settings.result_url_template:
+            result_url = settings.result_url_template.replace(
+                "{recUuid}",
+                quote(rec_uuid, safe=""),
+            )
+
+        normalized.append(
+            {
+                "recUuid": rec_uuid,
+                "content": _text(item.get("content")),
+                "title": _text(item.get("title")),
+                "url": result_url,
+            }
+        )
+    return normalized
+
+
+async def execute_search(
+    query: str,
+    settings: ConfigurableSearchSettings,
+    *,
+    transport: httpx.AsyncBaseTransport | None = None,
+) -> str:
+    """Execute a search request; ``transport`` exists for network-free tests."""
+
+    normalized_query = query.strip()
+    if not normalized_query:
+        return _error_result(query, "Search query must not be empty")
+    if not settings.enabled:
+        return _error_result(normalized_query, "web_search is disabled")
+
+    payload = deepcopy(settings.payload)
+    payload["query"] = normalized_query
+
+    try:
+        async with httpx.AsyncClient(
+            timeout=settings.timeout,
+            verify=settings.verify_ssl,
+            transport=transport,
+        ) as client:
+            response = await client.post(settings.endpoint, json=payload)
+            response.raise_for_status()
+            data = response.json()
+        results = _normalize_results(data, settings)
+    except httpx.TimeoutException:
+        logger.warning("Configured web search timed out for endpoint %s", settings.endpoint)
+        return _error_result(normalized_query, "Search API request timed out")
+    except httpx.HTTPStatusError as exc:
+        logger.warning(
+            "Configured web search returned HTTP %s for endpoint %s",
+            exc.response.status_code,
+            settings.endpoint,
+        )
+        return _error_result(normalized_query, f"Search API returned HTTP {exc.response.status_code}")
+    except (httpx.HTTPError, json.JSONDecodeError) as exc:
+        logger.warning("Configured web search request failed: %s", exc)
+        return _error_result(normalized_query, f"Search API request failed: {type(exc).__name__}")
+    except ValueError as exc:
+        logger.warning("Configured web search returned an invalid response: %s", exc)
+        return _error_result(normalized_query, str(exc))
+
+    return json.dumps(
+        {
+            "query": normalized_query,
+            "total_results": len(results),
+            "results": results,
+        },
+        ensure_ascii=False,
+    )
+
+
+@tool("web_search", parse_docstring=True)
+async def web_search_tool(query: str) -> str:
+    """Search the configured information service for current facts and source passages.
+
+    Args:
+        query: Focused search keywords or a natural-language retrieval question.
+    """
+
+    try:
+        settings = ConfigurableSearchSettings.from_tool_config(get_app_config().get_tool_config("web_search"))
+    except ValueError as exc:
+        return _error_result(query, str(exc))
+    return await execute_search(query, settings)

offline-backend-20260512/backend/packages/harness/deerflow/config/agents_config.py

@@ -7,7 +7,7 @@ import threading
 from typing import Any
 
 import yaml
-from pydantic import BaseModel, model_validator
+from pydantic import BaseModel, field_validator, model_validator
 
 from deerflow.config.paths import get_paths
 
@@ -82,7 +82,7 @@ class CommonQuestion(BaseModel):
         return value
 
 
-class AgentConfig(BaseModel):
+class AgentConfig(BaseModel):
     """Configuration for a custom agent."""
 
     id: str | None = None
@@ -108,7 +108,20 @@ class AgentConfig(BaseModel):
     # 拼进本轮任务，避免弱模型无视系统提示里的技能块。None/False = 关（默认）；True = 开。
     # 与「业务链条配置」里的同名开关取 OR：任一打开即对该席位生效。见
     # ``app/gateway/roundtable_seat_skills.py``。
-    seat_skill_directive: bool | None = None
+    seat_skill_directive: bool | None = None
+    # DeerFlow mapping ids this agent may offer in its new-conversation
+    # knowledge selector. WeKnora ids and credentials never enter agent files.
+    llmwiki_knowledge_base_ids: list[str] | None = None
+
+    @field_validator("llmwiki_knowledge_base_ids")
+    @classmethod
+    def _normalize_llmwiki_ids(cls, value: list[str] | None) -> list[str] | None:
+        if value is None:
+            return None
+        normalized = list(dict.fromkeys(str(item).strip() for item in value if str(item).strip()))
+        if len(normalized) > 100:
+            raise ValueError("An agent can bind at most 100 LLMWiki knowledge bases")
+        return normalized
 
 
 def load_agent_config(agent_id: str | None) -> AgentConfig | None:

offline-backend-20260512/backend/packages/harness/deerflow/config/app_config.py

@@ -17,10 +17,11 @@ from deerflow.config.database_config import DatabaseConfig
 from deerflow.config.extensions_config import ExtensionsConfig
 from deerflow.config.gateway_config import GatewayConfig
 from deerflow.config.guardrails_config import GuardrailsConfig, load_guardrails_config_from_dict
-from deerflow.config.knowledge_config import KnowledgeConfig
-from deerflow.config.knowledge_ingest_config import KnowledgeIngestConfig
-from deerflow.config.log_system_config import LogSystemConfig
-from deerflow.config.login_config import AuthLoginConfig
+from deerflow.config.knowledge_config import KnowledgeConfig
+from deerflow.config.knowledge_ingest_config import KnowledgeIngestConfig
+from deerflow.config.llmwiki_config import LlmWikiConfig
+from deerflow.config.log_system_config import LogSystemConfig
+from deerflow.config.login_config import AuthLoginConfig
 from deerflow.config.memory_config import MemoryConfig, load_memory_config_from_dict
 from deerflow.config.model_config import ModelConfig
 from deerflow.config.run_events_config import RunEventsConfig
@@ -42,7 +43,8 @@ from deerflow.config.tool_search_config import ToolSearchConfig, load_tool_searc
 from deerflow.config.vision_config import VisionConfig
 from deerflow.config.y_log_config import YLogConfig
 
-load_dotenv()
+load_dotenv()
+load_dotenv(Path(__file__).resolve().parents[4] / ".env.local", override=True)
 
 logger = logging.getLogger(__name__)
 
@@ -115,7 +117,8 @@ class AppConfig(BaseModel):
     models: list[ModelConfig] = Field(default_factory=list, description="Available models")
     vision: VisionConfig = Field(default_factory=VisionConfig, description="Image recognition (识图) configuration")
     auth_login: AuthLoginConfig = Field(default_factory=AuthLoginConfig, description="Custom login modes (username password gate, token exchange)")
-    log_system: LogSystemConfig = Field(default_factory=LogSystemConfig, description="Admin-only external log-system entry shown in the settings menu")
+    log_system: LogSystemConfig = Field(default_factory=LogSystemConfig, description="Admin-only external log-system entry shown in the settings menu")
+    llmwiki: LlmWikiConfig = Field(default_factory=LlmWikiConfig, description="Conditional legacy/WeKnora LLMWiki integration")
     sandbox: SandboxConfig = Field(description="Sandbox configuration")
     tools: list[ToolConfig] = Field(default_factory=list, description="Available tools")
     tool_groups: list[ToolGroupConfig] = Field(default_factory=list, description="Available tool groups")

offline-backend-20260512/backend/packages/harness/deerflow/config/llmwiki_config.py

@@ -0,0 +1,117 @@
+"""Conditional LLMWiki provider configuration.
+
+DeerFlow owns only the connection to WeKnora. Database, Redis, document
+parsing, storage and model configuration stay entirely in WeKnora's own
+deployment files. An effective non-empty API URL is the sole provider switch.
+"""
+
+from __future__ import annotations
+
+from dataclasses import dataclass
+from urllib.parse import urlsplit, urlunsplit
+
+from pydantic import BaseModel, ConfigDict, Field, field_validator
+
+
+def normalize_service_url(value: str | None) -> str:
+    """Normalize an optional HTTP(S) service URL and reject embedded secrets."""
+
+    raw = str(value or "").strip()
+    if not raw:
+        return ""
+    parsed = urlsplit(raw)
+    if parsed.scheme not in {"http", "https"} or not parsed.hostname:
+        raise ValueError("WeKnora addresses must be absolute http(s) URLs")
+    if parsed.username is not None or parsed.password is not None:
+        raise ValueError("WeKnora addresses must not contain credentials")
+    if parsed.query or parsed.fragment:
+        raise ValueError("WeKnora addresses must not contain query strings or fragments")
+    path = parsed.path.rstrip("/")
+    return urlunsplit((parsed.scheme, parsed.netloc, path, "", ""))
+
+
+class WeKnoraConnectionConfig(BaseModel):
+    """The complete DeerFlow-owned connection surface for WeKnora."""
+
+    api_base_url: str = Field(default="", description="WeKnora app API base URL; non-empty enables WeKnora mode")
+    web_base_url: str = Field(default="", description="Optional internal WeKnora administration URL")
+    admin_email: str = Field(
+        default="",
+        description=(
+            "Optional WeKnora administrator email used only by DeerFlow's "
+            "server-side detail-page iframe proxy. Prefer $ENV references."
+        ),
+    )
+    admin_password: str = Field(
+        default="",
+        repr=False,
+        description=(
+            "Optional WeKnora administrator password used only by DeerFlow's "
+            "server-side detail-page iframe proxy. Prefer $ENV references."
+        ),
+    )
+    model_config = ConfigDict(extra="ignore")
+
+    @field_validator("api_base_url", "web_base_url")
+    @classmethod
+    def _validate_url(cls, value: str) -> str:
+        return normalize_service_url(value)
+
+
+class LlmWikiConfig(BaseModel):
+    """Top-level ``config.yaml`` section for conditional WeKnora mode."""
+
+    weknora: WeKnoraConnectionConfig = Field(default_factory=WeKnoraConnectionConfig)
+    model_config = ConfigDict(extra="ignore")
+
+
+class LlmWikiRuntimeOverride(BaseModel):
+    """Administrator-written URL override stored in system settings.
+
+    ``enabled=False`` means "use config.yaml". When enabled, an empty API URL
+    deliberately forces legacy mode even if the source file has an address.
+    Secrets are never stored here.
+    """
+
+    enabled: bool = False
+    api_base_url: str = ""
+    web_base_url: str = ""
+    model_config = ConfigDict(extra="ignore")
+
+    @field_validator("api_base_url", "web_base_url")
+    @classmethod
+    def _validate_url(cls, value: str) -> str:
+        return normalize_service_url(value)
+
+
+@dataclass(frozen=True)
+class ResolvedLlmWikiRuntime:
+    provider: str
+    api_base_url: str
+    web_base_url: str
+    admin_email: str
+    admin_password: str
+    override_enabled: bool
+
+    @property
+    def weknora_enabled(self) -> bool:
+        return self.provider == "weknora"
+
+
+def resolve_llmwiki_runtime(
+    config: LlmWikiConfig,
+    override: LlmWikiRuntimeOverride | None = None,
+) -> ResolvedLlmWikiRuntime:
+    """Resolve the deterministic provider from file config plus URL override."""
+
+    use_override = bool(override and override.enabled)
+    api_base_url = override.api_base_url if use_override and override else config.weknora.api_base_url
+    web_base_url = override.web_base_url if use_override and override else config.weknora.web_base_url
+    return ResolvedLlmWikiRuntime(
+        provider="weknora" if api_base_url else "legacy",
+        api_base_url=api_base_url,
+        web_base_url=web_base_url,
+        admin_email=config.weknora.admin_email,
+        admin_password=config.weknora.admin_password,
+        override_enabled=use_override,
+    )

offline-backend-20260512/backend/packages/harness/deerflow/config/system_settings.py

@@ -23,6 +23,7 @@ from uuid import uuid4
 
 from pydantic import BaseModel, ConfigDict, Field, field_validator
 
+from deerflow.config.llmwiki_config import LlmWikiRuntimeOverride
 from deerflow.config.runtime_paths import runtime_home
 
 logger = logging.getLogger(__name__)
@@ -499,6 +500,7 @@ class SystemSettings(BaseModel):
     agent_modules: AgentModuleSettings = Field(default_factory=AgentModuleSettings)
     concurrency_monitor: ConcurrencyMonitorSettings = Field(default_factory=ConcurrencyMonitorSettings)
     model_concurrency: ModelConcurrencySettings = Field(default_factory=ModelConcurrencySettings)
+    llmwiki: LlmWikiRuntimeOverride = Field(default_factory=LlmWikiRuntimeOverride)
     model_config = ConfigDict(extra="ignore")
 
 

offline-backend-20260512/backend/packages/harness/deerflow/config/tool_config.py

@@ -1,20 +1,21 @@
-from pydantic import BaseModel, ConfigDict, Field
-
-
-class ToolGroupConfig(BaseModel):
-    """Config section for a tool group"""
-
-    name: str = Field(..., description="Unique name for the tool group")
-    model_config = ConfigDict(extra="allow")
-
-
-class ToolConfig(BaseModel):
-    """Config section for a tool"""
-
-    name: str = Field(..., description="Unique name for the tool")
-    group: str = Field(..., description="Group name for the tool")
-    use: str = Field(
-        ...,
-        description="Variable name of the tool provider(e.g. deerflow.sandbox.tools:bash_tool)",
-    )
-    model_config = ConfigDict(extra="allow")
+from pydantic import BaseModel, ConfigDict, Field
+
+
+class ToolGroupConfig(BaseModel):
+    """Config section for a tool group"""
+
+    name: str = Field(..., description="Unique name for the tool group")
+    model_config = ConfigDict(extra="allow")
+
+
+class ToolConfig(BaseModel):
+    """Config section for a tool"""
+
+    name: str = Field(..., description="Unique name for the tool")
+    group: str = Field(..., description="Group name for the tool")
+    enabled: bool = Field(default=True, description="Whether the tool is registered with the agent")
+    use: str = Field(
+        ...,
+        description="Variable name of the tool provider(e.g. deerflow.sandbox.tools:bash_tool)",
+    )
+    model_config = ConfigDict(extra="allow")

offline-backend-20260512/backend/packages/harness/deerflow/integrations/__init__.py

@@ -0,0 +1,2 @@
+"""External service integrations owned by the DeerFlow harness."""
+

offline-backend-20260512/backend/packages/harness/deerflow/integrations/weknora/__init__.py

@@ -0,0 +1,5 @@
+"""WeKnora REST integration."""
+
+from deerflow.integrations.weknora.client import WeKnoraClient, WeKnoraError
+
+__all__ = ["WeKnoraClient", "WeKnoraError"]

offline-backend-20260512/backend/packages/harness/deerflow/integrations/weknora/client.py

@@ -0,0 +1,443 @@
+"""Small, version-pinned adapter over WeKnora's official REST API."""
+
+from __future__ import annotations
+
+from collections.abc import Mapping, Sequence
+from typing import Any
+from urllib.parse import quote
+
+import httpx
+
+
+class WeKnoraError(RuntimeError):
+    """Normalized upstream failure that is safe for DeerFlow to translate."""
+
+    def __init__(self, message: str, *, status_code: int = 502) -> None:
+        super().__init__(message)
+        self.status_code = status_code
+
+
+def _data(payload: Any) -> Any:
+    if not isinstance(payload, dict):
+        raise WeKnoraError("WeKnora returned an invalid response")
+    if payload.get("success") is False:
+        message = payload.get("message") or payload.get("error") or "WeKnora rejected the request"
+        raise WeKnoraError(str(message), status_code=400)
+    return payload.get("data", payload)
+
+
+def _as_list(value: Any) -> list[dict[str, Any]]:
+    if isinstance(value, list):
+        return [dict(item) for item in value if isinstance(item, Mapping)]
+    if isinstance(value, dict):
+        for key in ("items", "list", "knowledge_bases", "knowledge", "records"):
+            nested = value.get(key)
+            if isinstance(nested, list):
+                return [dict(item) for item in nested if isinstance(item, Mapping)]
+    return []
+
+
+class WeKnoraClient:
+    """Async client for a trusted-network WeKnora endpoint.
+
+    The caller may inject an ``httpx.AsyncClient`` for connection pooling and
+    tests. When omitted, each public method creates and closes a short-lived
+    client. DeerFlow deliberately sends no WeKnora credential; the independently
+    deployed service or its reverse proxy must allow DeerFlow's internal traffic.
+    """
+
+    def __init__(
+        self,
+        *,
+        base_url: str,
+        timeout_seconds: float = 30.0,
+        http_client: httpx.AsyncClient | None = None,
+    ) -> None:
+        self.base_url = base_url.rstrip("/")
+        self.timeout_seconds = timeout_seconds
+        self._http_client = http_client
+        if not self.base_url:
+            raise ValueError("WeKnora base URL is required")
+
+    async def _request(
+        self,
+        method: str,
+        path: str,
+        *,
+        json: Any = None,
+        params: Mapping[str, Any] | None = None,
+        files: Any = None,
+        data: Mapping[str, Any] | None = None,
+    ) -> Any:
+        headers = {"Accept": "application/json"}
+        owns_client = self._http_client is None
+        client = self._http_client or httpx.AsyncClient(timeout=httpx.Timeout(self.timeout_seconds), follow_redirects=False)
+        try:
+            response = await client.request(
+                method,
+                f"{self.base_url}{path}",
+                headers=headers,
+                json=json,
+                params=params,
+                files=files,
+                data=data,
+            )
+        except httpx.TimeoutException as exc:
+            raise WeKnoraError("WeKnora request timed out", status_code=504) from exc
+        except httpx.HTTPError as exc:
+            raise WeKnoraError("WeKnora is unavailable", status_code=502) from exc
+        finally:
+            if owns_client:
+                await client.aclose()
+        if response.status_code == 404:
+            raise WeKnoraError("WeKnora resource was not found", status_code=404)
+        if response.status_code in {401, 403}:
+            raise WeKnoraError(
+                "WeKnora rejected DeerFlow's anonymous internal request; allow DeerFlow in WeKnora or its reverse proxy",
+                status_code=502,
+            )
+        if response.status_code >= 400:
+            detail = ""
+            try:
+                body = response.json()
+                if isinstance(body, dict):
+                    detail = str(body.get("message") or body.get("error") or body.get("detail") or "")
+            except ValueError:
+                detail = ""
+            raise WeKnoraError(detail or f"WeKnora returned HTTP {response.status_code}", status_code=502)
+        if response.status_code == 204 or not response.content:
+            return {"success": True}
+        try:
+            return response.json()
+        except ValueError as exc:
+            raise WeKnoraError("WeKnora returned invalid JSON") from exc
+
+    async def health(self) -> dict[str, Any]:
+        payload = await self._request("GET", "/health")
+        return dict(payload) if isinstance(payload, dict) else {"status": "ok"}
+
+    async def list_knowledge_bases(self) -> list[dict[str, Any]]:
+        return _as_list(_data(await self._request("GET", "/api/v1/knowledge-bases")))
+
+    async def get_knowledge_base(self, knowledge_base_id: str) -> dict[str, Any]:
+        value = _data(await self._request("GET", f"/api/v1/knowledge-bases/{quote(knowledge_base_id, safe='')}"))
+        if not isinstance(value, dict):
+            raise WeKnoraError("WeKnora returned an invalid knowledge base")
+        return dict(value)
+
+    async def list_models(self) -> list[dict[str, Any]]:
+        """List models from WeKnora; model credentials are never returned."""
+
+        return _as_list(_data(await self._request("GET", "/api/v1/models")))
+
+    async def get_default_embedding_model_id(self) -> str:
+        models = [item for item in await self.list_models() if str(item.get("type") or "").strip().lower() == "embedding"]
+        preferred = next((item for item in models if item.get("is_default") is True), None)
+        selected = preferred or next(
+            (item for item in models if str(item.get("status") or "active").lower() == "active"),
+            None,
+        )
+        model_id = str((selected or {}).get("id") or "")
+        if not model_id:
+            raise WeKnoraError(
+                "WeKnora has no available embedding model; configure its default model in WeKnora",
+                status_code=400,
+            )
+        return model_id
+
+    async def get_default_knowledge_qa_model_id(self) -> str:
+        """Resolve WeKnora's own active synthesis model for Wiki indexing."""
+
+        models = [item for item in await self.list_models() if str(item.get("type") or "").strip().lower() == "knowledgeqa"]
+        preferred = next((item for item in models if item.get("is_default") is True), None)
+        selected = preferred or next(
+            (item for item in models if str(item.get("status") or "active").lower() == "active"),
+            None,
+        )
+        model_id = str((selected or {}).get("id") or "")
+        if not model_id:
+            raise WeKnoraError(
+                "WeKnora has no available KnowledgeQA model; configure one in WeKnora before creating a Wiki",
+                status_code=400,
+            )
+        return model_id
+
+    async def create_knowledge_base(
+        self,
+        *,
+        name: str,
+        description: str = "",
+        kb_type: str = "document",
+        embedding_model_id: str | None = None,
+        wiki_enabled: bool = False,
+    ) -> dict[str, Any]:
+        model_id = embedding_model_id or await self.get_default_embedding_model_id()
+        create_payload: dict[str, Any] = {
+            "name": name,
+            "description": description,
+            "type": kb_type,
+            "embedding_model_id": model_id,
+        }
+        if wiki_enabled:
+            synthesis_model_id = await self.get_default_knowledge_qa_model_id()
+            create_payload.update(
+                {
+                    "summary_model_id": synthesis_model_id,
+                    "indexing_strategy": {
+                        "vector_enabled": True,
+                        "keyword_enabled": True,
+                        "wiki_enabled": True,
+                        "graph_enabled": False,
+                    },
+                    "wiki_config": {
+                        "synthesis_model_id": synthesis_model_id,
+                        "extraction_granularity": "standard",
+                    },
+                }
+            )
+        payload = await self._request(
+            "POST",
+            "/api/v1/knowledge-bases",
+            json=create_payload,
+        )
+        value = _data(payload)
+        if not isinstance(value, dict) or not value.get("id"):
+            raise WeKnoraError("WeKnora did not return the created knowledge base")
+        return dict(value)
+
+    async def update_knowledge_base(self, knowledge_base_id: str, changes: Mapping[str, Any]) -> dict[str, Any]:
+        value = _data(
+            await self._request(
+                "PUT",
+                f"/api/v1/knowledge-bases/{quote(knowledge_base_id, safe='')}",
+                json=dict(changes),
+            )
+        )
+        return dict(value) if isinstance(value, dict) else {}
+
+    async def delete_knowledge_base(self, knowledge_base_id: str) -> None:
+        await self._request("DELETE", f"/api/v1/knowledge-bases/{quote(knowledge_base_id, safe='')}")
+
+    async def list_documents(self, knowledge_base_id: str, *, page: int = 1, page_size: int = 100) -> dict[str, Any]:
+        payload = await self._request(
+            "GET",
+            f"/api/v1/knowledge-bases/{quote(knowledge_base_id, safe='')}/knowledge",
+            params={"page": page, "page_size": page_size},
+        )
+        rows = _as_list(_data(payload))
+        total = payload.get("total", len(rows)) if isinstance(payload, dict) else len(rows)
+        return {"items": rows, "total": int(total or 0), "page": page, "page_size": page_size}
+
+    async def upload_document(
+        self,
+        knowledge_base_id: str,
+        *,
+        filename: str,
+        content: bytes,
+        content_type: str = "application/octet-stream",
+    ) -> dict[str, Any]:
+        value = _data(
+            await self._request(
+                "POST",
+                f"/api/v1/knowledge-bases/{quote(knowledge_base_id, safe='')}/knowledge/file",
+                files={"file": (filename, content, content_type)},
+            )
+        )
+        return dict(value) if isinstance(value, dict) else {}
+
+    async def import_document_url(self, knowledge_base_id: str, *, url: str) -> dict[str, Any]:
+        value = _data(
+            await self._request(
+                "POST",
+                f"/api/v1/knowledge-bases/{quote(knowledge_base_id, safe='')}/knowledge/url",
+                json={"url": url},
+            )
+        )
+        return dict(value) if isinstance(value, dict) else {}
+
+    async def create_manual_document(
+        self,
+        knowledge_base_id: str,
+        *,
+        title: str,
+        content: str,
+    ) -> dict[str, Any]:
+        value = _data(
+            await self._request(
+                "POST",
+                f"/api/v1/knowledge-bases/{quote(knowledge_base_id, safe='')}/knowledge/manual",
+                json={"title": title, "content": content, "status": "publish"},
+            )
+        )
+        return dict(value) if isinstance(value, dict) else {}
+
+    async def delete_document(self, knowledge_id: str) -> None:
+        await self._request("DELETE", f"/api/v1/knowledge/{quote(knowledge_id, safe='')}")
+
+    async def get_document(self, knowledge_id: str) -> dict[str, Any]:
+        value = _data(await self._request("GET", f"/api/v1/knowledge/{quote(knowledge_id, safe='')}"))
+        if not isinstance(value, dict):
+            raise WeKnoraError("WeKnora returned an invalid document")
+        return dict(value)
+
+    async def get_document_preview(self, knowledge_id: str) -> tuple[bytes, str]:
+        """Return the original document preview bytes and its media type."""
+
+        owns_client = self._http_client is None
+        client = self._http_client or httpx.AsyncClient(
+            timeout=httpx.Timeout(self.timeout_seconds),
+            follow_redirects=True,
+        )
+        try:
+            response = await client.request(
+                "GET",
+                f"{self.base_url}/api/v1/knowledge/{quote(knowledge_id, safe='')}/preview",
+                headers={"Accept": "*/*"},
+                follow_redirects=True,
+            )
+        except httpx.TimeoutException as exc:
+            raise WeKnoraError("WeKnora preview request timed out", status_code=504) from exc
+        except httpx.HTTPError as exc:
+            raise WeKnoraError("WeKnora preview is unavailable", status_code=502) from exc
+        finally:
+            if owns_client:
+                await client.aclose()
+        if response.status_code == 404:
+            raise WeKnoraError("WeKnora document preview was not found", status_code=404)
+        if response.status_code in {401, 403}:
+            raise WeKnoraError(
+                "WeKnora rejected DeerFlow's anonymous internal request; allow DeerFlow in WeKnora or its reverse proxy",
+                status_code=502,
+            )
+        if response.status_code >= 400:
+            raise WeKnoraError(f"WeKnora returned HTTP {response.status_code}", status_code=502)
+        media_type = response.headers.get("content-type", "application/octet-stream").split(";", 1)[0].strip()
+        return response.content, media_type or "application/octet-stream"
+
+    async def reprocess_document(self, knowledge_id: str) -> dict[str, Any]:
+        value = _data(await self._request("POST", f"/api/v1/knowledge/{quote(knowledge_id, safe='')}/reparse"))
+        return dict(value) if isinstance(value, dict) else {}
+
+    async def list_chunks(self, knowledge_id: str, *, page: int = 1, page_size: int = 20) -> dict[str, Any]:
+        payload = await self._request(
+            "GET",
+            f"/api/v1/chunks/{quote(knowledge_id, safe='')}",
+            params={"page": page, "page_size": page_size},
+        )
+        rows = _as_list(_data(payload))
+        total = payload.get("total", len(rows)) if isinstance(payload, dict) else len(rows)
+        return {"items": rows, "total": int(total or 0), "page": page, "page_size": page_size}
+
+    async def get_wiki_graph(self, knowledge_base_id: str, *, limit: int = 200) -> dict[str, Any]:
+        value = _data(
+            await self._request(
+                "GET",
+                f"/api/v1/knowledgebase/{quote(knowledge_base_id, safe='')}/wiki/graph",
+                params={"mode": "overview", "limit": max(1, min(limit, 500))},
+            )
+        )
+        if not isinstance(value, dict):
+            raise WeKnoraError("WeKnora returned an invalid wiki graph")
+        return dict(value)
+
+    async def list_wiki_pages(
+        self,
+        knowledge_base_id: str,
+        *,
+        page: int = 1,
+        page_size: int = 50,
+        query: str = "",
+    ) -> dict[str, Any]:
+        payload = await self._request(
+            "GET",
+            f"/api/v1/knowledgebase/{quote(knowledge_base_id, safe='')}/wiki/pages",
+            params={"page": page, "page_size": page_size, "query": query or None},
+        )
+        value = _data(payload)
+        source = value if isinstance(value, dict) else payload if isinstance(payload, dict) else {}
+        rows = _as_list(source.get("pages") if isinstance(source, dict) else value)
+        return {
+            "pages": rows,
+            "total": int(source.get("total") or len(rows)),
+            "page": int(source.get("page") or page),
+            "page_size": int(source.get("page_size") or page_size),
+            "total_pages": int(source.get("total_pages") or max(1, (int(source.get("total") or len(rows)) + page_size - 1) // page_size)),
+        }
+
+    async def get_wiki_page(self, knowledge_base_id: str, slug: str) -> dict[str, Any]:
+        safe_slug = "/".join(quote(part, safe="") for part in slug.split("/"))
+        value = _data(
+            await self._request(
+                "GET",
+                f"/api/v1/knowledgebase/{quote(knowledge_base_id, safe='')}/wiki/pages/{safe_slug}",
+            )
+        )
+        if not isinstance(value, dict):
+            raise WeKnoraError("WeKnora returned an invalid Wiki page")
+        return dict(value)
+
+    async def search(self, query: str, knowledge_base_ids: Sequence[str]) -> list[dict[str, Any]]:
+        ids = [str(value) for value in knowledge_base_ids if str(value)]
+        if not ids:
+            return []
+        rows = _as_list(
+            _data(
+                await self._request(
+                    "POST",
+                    "/api/v1/knowledge-search",
+                    json={"query": query, "knowledge_base_ids": ids},
+                )
+            )
+        )
+        normalized: list[dict[str, Any]] = []
+        for row in rows:
+            normalized.append(
+                {
+                    "chunk_id": str(row.get("id") or row.get("chunk_id") or ""),
+                    "content": str(row.get("content") or ""),
+                    "knowledge_id": str(row.get("knowledge_id") or ""),
+                    "knowledge_base_id": str(row.get("knowledge_base_id") or ""),
+                    "title": str(row.get("knowledge_title") or row.get("title") or row.get("knowledge_filename") or ""),
+                    "filename": str(row.get("knowledge_filename") or ""),
+                    "score": float(row.get("score") or 0),
+                    "chunk_index": int(row.get("chunk_index") or 0),
+                    "metadata": row.get("metadata") if isinstance(row.get("metadata"), dict) else {},
+                }
+            )
+        return normalized
+
+    async def get_chunk(self, chunk_id: str) -> dict[str, Any]:
+        value = _data(await self._request("GET", f"/api/v1/chunks/by-id/{quote(chunk_id, safe='')}"))
+        if not isinstance(value, dict):
+            raise WeKnoraError("WeKnora returned an invalid chunk")
+        return dict(value)
+
+    async def get_chunk_context(self, chunk_id: str, *, expected_knowledge_base_id: str, radius: int = 2) -> list[dict[str, Any]]:
+        center = await self.get_chunk(chunk_id)
+        if str(center.get("knowledge_base_id") or "") != expected_knowledge_base_id:
+            raise WeKnoraError("Chunk does not belong to the authorized knowledge base", status_code=404)
+
+        before: list[dict[str, Any]] = []
+        cursor = center
+        for _ in range(max(0, min(radius, 4))):
+            previous_id = str(cursor.get("pre_chunk_id") or "")
+            if not previous_id:
+                break
+            previous = await self.get_chunk(previous_id)
+            if str(previous.get("knowledge_base_id") or "") != expected_knowledge_base_id:
+                break
+            before.append(previous)
+            cursor = previous
+
+        after: list[dict[str, Any]] = []
+        cursor = center
+        for _ in range(max(0, min(radius, 4))):
+            next_id = str(cursor.get("next_chunk_id") or "")
+            if not next_id:
+                break
+            following = await self.get_chunk(next_id)
+            if str(following.get("knowledge_base_id") or "") != expected_knowledge_base_id:
+                break
+            after.append(following)
+            cursor = following
+        return [*reversed(before), center, *after]

offline-backend-20260512/backend/packages/harness/deerflow/integrations/weknora/runtime.py

@@ -0,0 +1,43 @@
+"""Runtime composition shared by Gateway routes and the read-only agent tool."""
+
+from __future__ import annotations
+
+import httpx
+
+from deerflow.config import get_app_config
+from deerflow.config.app_config import AppConfig
+from deerflow.config.llmwiki_config import ResolvedLlmWikiRuntime, resolve_llmwiki_runtime
+from deerflow.config.system_settings import load_system_settings
+from deerflow.integrations.weknora.client import WeKnoraClient
+from deerflow.persistence.llmwiki import LlmWikiStore
+
+_default_store: LlmWikiStore | None = None
+
+
+def register_llmwiki_store(store: LlmWikiStore | None) -> None:
+    global _default_store
+    _default_store = store
+
+
+def get_llmwiki_store() -> LlmWikiStore:
+    if _default_store is None:
+        raise RuntimeError("LLMWiki store is not initialized")
+    return _default_store
+
+
+def get_resolved_llmwiki_runtime(config: AppConfig | None = None) -> ResolvedLlmWikiRuntime:
+    current = config or get_app_config()
+    return resolve_llmwiki_runtime(current.llmwiki, load_system_settings().llmwiki)
+
+
+def build_weknora_client(
+    runtime: ResolvedLlmWikiRuntime,
+    *,
+    http_client: httpx.AsyncClient | None = None,
+) -> WeKnoraClient:
+    if not runtime.weknora_enabled:
+        raise RuntimeError("WeKnora mode is not enabled")
+    return WeKnoraClient(
+        base_url=runtime.api_base_url,
+        http_client=http_client,
+    )

offline-backend-20260512/backend/packages/harness/deerflow/persistence/llmwiki/__init__.py

@@ -0,0 +1,22 @@
+"""DeerFlow-owned LLMWiki authorization metadata."""
+
+from deerflow.persistence.llmwiki.base import LlmWikiScope, LlmWikiStore
+from deerflow.persistence.llmwiki.memory import MemoryLlmWikiStore
+from deerflow.persistence.llmwiki.model import LlmWikiConversationDepositRow, LlmWikiKnowledgeBaseRow
+from deerflow.persistence.llmwiki.sql import LlmWikiRepository
+
+__all__ = [
+    "LlmWikiKnowledgeBaseRow",
+    "LlmWikiConversationDepositRow",
+    "LlmWikiRepository",
+    "LlmWikiScope",
+    "LlmWikiStore",
+    "MemoryLlmWikiStore",
+    "make_llmwiki_store",
+]
+
+
+def make_llmwiki_store(session_factory):
+    if session_factory is None:
+        return MemoryLlmWikiStore()
+    return LlmWikiRepository(session_factory)

offline-backend-20260512/backend/packages/harness/deerflow/persistence/llmwiki/base.py

@@ -0,0 +1,72 @@
+"""Persistence contract for DeerFlow-owned LLMWiki authorization metadata."""
+
+from __future__ import annotations
+
+from abc import ABC, abstractmethod
+from typing import Any, Literal
+
+LlmWikiScope = Literal["personal", "public", "all"]
+
+
+class LlmWikiStore(ABC):
+    @abstractmethod
+    async def create_mapping(self, *, weknora_id: str, owner_user_id: str, name: str, description: str, kb_type: str) -> dict[str, Any]: ...
+
+    @abstractmethod
+    async def list_visible(self, user_id: str, *, scope: LlmWikiScope = "all", is_admin: bool = False) -> list[dict[str, Any]]: ...
+
+    @abstractmethod
+    async def get_authorized(self, mapping_id: str, user_id: str, *, write: bool, is_admin: bool) -> dict[str, Any] | None: ...
+
+    @abstractmethod
+    async def get_by_weknora_id(self, weknora_id: str) -> dict[str, Any] | None: ...
+
+    @abstractmethod
+    async def update_mapping(self, mapping_id: str, *, name: str | None = None, description: str | None = None) -> dict[str, Any] | None: ...
+
+    @abstractmethod
+    async def delete_mapping(self, mapping_id: str) -> bool: ...
+
+    @abstractmethod
+    async def request_publish(self, mapping_id: str, actor_user_id: str, *, is_admin: bool) -> dict[str, Any] | None: ...
+
+    @abstractmethod
+    async def review_publish(self, mapping_id: str, reviewer_user_id: str, *, approved: bool) -> dict[str, Any] | None: ...
+
+    @abstractmethod
+    async def unpublish(self, mapping_id: str, actor_user_id: str, *, is_admin: bool) -> dict[str, Any] | None: ...
+
+    @abstractmethod
+    async def list_publications(self, *, status: str | None = None) -> list[dict[str, Any]]: ...
+
+    @abstractmethod
+    async def get_conversation_deposit_mapping(self) -> dict[str, Any] | None: ...
+
+    @abstractmethod
+    async def get_conversation_deposit(
+        self,
+        *,
+        thread_id: str,
+        assistant_message_id: str,
+    ) -> dict[str, Any] | None: ...
+
+    @abstractmethod
+    async def upsert_conversation_deposit(
+        self,
+        *,
+        thread_id: str,
+        human_message_id: str | None,
+        assistant_message_id: str,
+        knowledge_base_mapping_id: str,
+        weknora_knowledge_id: str | None,
+        title: str,
+        created_by_user_id: str | None,
+    ) -> dict[str, Any]: ...
+
+    @abstractmethod
+    async def delete_conversation_deposits(
+        self,
+        *,
+        thread_id: str,
+        message_ids: list[str] | None = None,
+    ) -> list[dict[str, Any]]: ...

offline-backend-20260512/backend/packages/harness/deerflow/persistence/llmwiki/memory.py

@@ -0,0 +1,190 @@
+"""In-memory LLMWiki metadata store for tests and DB-less deployments."""
+
+from __future__ import annotations
+
+from datetime import UTC, datetime
+from typing import Any
+from uuid import uuid4
+
+from deerflow.persistence.llmwiki.base import LlmWikiScope, LlmWikiStore
+
+
+def _now() -> str:
+    return datetime.now(UTC).isoformat()
+
+
+class MemoryLlmWikiStore(LlmWikiStore):
+    def __init__(self) -> None:
+        self._rows: dict[str, dict[str, Any]] = {}
+        self._conversation_deposits: dict[str, dict[str, Any]] = {}
+
+    async def create_mapping(self, *, weknora_id: str, owner_user_id: str, name: str, description: str, kb_type: str) -> dict[str, Any]:
+        if any(row["weknora_id"] == weknora_id for row in self._rows.values()):
+            raise ValueError("WeKnora knowledge base is already mapped")
+        now = _now()
+        row = {
+            "id": str(uuid4()),
+            "weknora_id": weknora_id,
+            "owner_user_id": owner_user_id,
+            "name": name,
+            "description": description,
+            "kb_type": kb_type,
+            "publication_status": "private",
+            "published_by": None,
+            "reviewed_by": None,
+            "published_at": None,
+            "created_at": now,
+            "updated_at": now,
+        }
+        self._rows[row["id"]] = row
+        return dict(row)
+
+    async def list_visible(self, user_id: str, *, scope: LlmWikiScope = "all", is_admin: bool = False) -> list[dict[str, Any]]:
+        rows = list(self._rows.values())
+        if scope == "personal":
+            rows = rows if is_admin else [row for row in rows if row["owner_user_id"] == user_id]
+        elif scope == "public":
+            rows = [row for row in rows if row["publication_status"] == "published"]
+        else:
+            rows = rows if is_admin else [row for row in rows if row["owner_user_id"] == user_id or row["publication_status"] == "published"]
+        return [dict(row) for row in sorted(rows, key=lambda row: row["created_at"], reverse=True)]
+
+    async def get_authorized(self, mapping_id: str, user_id: str, *, write: bool, is_admin: bool) -> dict[str, Any] | None:
+        row = self._rows.get(mapping_id)
+        if row is None:
+            return None
+        if is_admin or row["owner_user_id"] == user_id:
+            return dict(row)
+        if not write and row["publication_status"] == "published":
+            return dict(row)
+        return None
+
+    async def get_by_weknora_id(self, weknora_id: str) -> dict[str, Any] | None:
+        for row in self._rows.values():
+            if row["weknora_id"] == weknora_id:
+                return dict(row)
+        return None
+
+    async def update_mapping(self, mapping_id: str, *, name: str | None = None, description: str | None = None) -> dict[str, Any] | None:
+        row = self._rows.get(mapping_id)
+        if row is None:
+            return None
+        if name is not None:
+            row["name"] = name
+        if description is not None:
+            row["description"] = description
+        row["updated_at"] = _now()
+        return dict(row)
+
+    async def delete_mapping(self, mapping_id: str) -> bool:
+        return self._rows.pop(mapping_id, None) is not None
+
+    async def request_publish(self, mapping_id: str, actor_user_id: str, *, is_admin: bool) -> dict[str, Any] | None:
+        row = self._rows.get(mapping_id)
+        if row is None or (not is_admin and row["owner_user_id"] != actor_user_id):
+            return None
+        row["publication_status"] = "published"
+        row["published_by"] = actor_user_id
+        row["reviewed_by"] = None
+        row["published_at"] = _now()
+        row["updated_at"] = _now()
+        return dict(row)
+
+    async def review_publish(self, mapping_id: str, reviewer_user_id: str, *, approved: bool) -> dict[str, Any] | None:
+        row = self._rows.get(mapping_id)
+        if row is None or row["publication_status"] != "pending":
+            return None
+        row["publication_status"] = "published" if approved else "rejected"
+        row["reviewed_by"] = reviewer_user_id
+        row["published_at"] = _now() if approved else None
+        row["updated_at"] = _now()
+        return dict(row)
+
+    async def unpublish(self, mapping_id: str, actor_user_id: str, *, is_admin: bool) -> dict[str, Any] | None:
+        row = self._rows.get(mapping_id)
+        if row is None or (not is_admin and row["owner_user_id"] != actor_user_id):
+            return None
+        row["publication_status"] = "private"
+        row["reviewed_by"] = actor_user_id if is_admin else row.get("reviewed_by")
+        row["published_at"] = None
+        row["updated_at"] = _now()
+        return dict(row)
+
+    async def list_publications(self, *, status: str | None = None) -> list[dict[str, Any]]:
+        rows = [row for row in self._rows.values() if row["publication_status"] != "private"]
+        if status:
+            rows = [row for row in rows if row["publication_status"] == status]
+        return [dict(row) for row in sorted(rows, key=lambda row: row["updated_at"], reverse=True)]
+
+    async def get_conversation_deposit_mapping(self) -> dict[str, Any] | None:
+        for row in self._rows.values():
+            if row.get("name") == "对话沉淀" and row.get("publication_status") == "published":
+                return dict(row)
+        return None
+
+    async def get_conversation_deposit(
+        self,
+        *,
+        thread_id: str,
+        assistant_message_id: str,
+    ) -> dict[str, Any] | None:
+        key = f"{thread_id}:{assistant_message_id}"
+        row = self._conversation_deposits.get(key)
+        return dict(row) if row else None
+
+    async def upsert_conversation_deposit(
+        self,
+        *,
+        thread_id: str,
+        human_message_id: str | None,
+        assistant_message_id: str,
+        knowledge_base_mapping_id: str,
+        weknora_knowledge_id: str | None,
+        title: str,
+        created_by_user_id: str | None,
+    ) -> dict[str, Any]:
+        key = f"{thread_id}:{assistant_message_id}"
+        now = _now()
+        row = self._conversation_deposits.get(key)
+        if row is None:
+            row = {
+                "id": str(uuid4()),
+                "thread_id": thread_id,
+                "human_message_id": human_message_id,
+                "assistant_message_id": assistant_message_id,
+                "knowledge_base_mapping_id": knowledge_base_mapping_id,
+                "weknora_knowledge_id": weknora_knowledge_id,
+                "title": title,
+                "created_by_user_id": created_by_user_id,
+                "created_at": now,
+                "updated_at": now,
+            }
+            self._conversation_deposits[key] = row
+        else:
+            row.update(
+                {
+                    "human_message_id": human_message_id,
+                    "knowledge_base_mapping_id": knowledge_base_mapping_id,
+                    "weknora_knowledge_id": weknora_knowledge_id,
+                    "title": title,
+                    "updated_at": now,
+                }
+            )
+        return dict(row)
+
+    async def delete_conversation_deposits(
+        self,
+        *,
+        thread_id: str,
+        message_ids: list[str] | None = None,
+    ) -> list[dict[str, Any]]:
+        ids = set(message_ids or [])
+        deleted: list[dict[str, Any]] = []
+        for key, row in list(self._conversation_deposits.items()):
+            if row.get("thread_id") != thread_id:
+                continue
+            if ids and row.get("human_message_id") not in ids and row.get("assistant_message_id") not in ids:
+                continue
+            deleted.append(dict(row))
+            self._conversation_deposits.pop(key, None)
+        return deleted

offline-backend-20260512/backend/packages/harness/deerflow/persistence/llmwiki/model.py

@@ -0,0 +1,63 @@
+"""ORM mapping between DeerFlow ownership and WeKnora knowledge bases."""
+
+from __future__ import annotations
+
+from datetime import UTC, datetime
+
+from sqlalchemy import String, UniqueConstraint
+from sqlalchemy.orm import Mapped, mapped_column
+
+from deerflow.persistence.base import Base
+from deerflow.persistence.types import BeijingDateTime
+
+
+class LlmWikiKnowledgeBaseRow(Base):
+    __tablename__ = "llmwiki_knowledge_bases"
+
+    id: Mapped[str] = mapped_column(String(36), primary_key=True)
+    weknora_id: Mapped[str] = mapped_column(String(128), unique=True, index=True)
+    owner_user_id: Mapped[str] = mapped_column(String(64), index=True)
+    name: Mapped[str] = mapped_column(String(255), default="")
+    description: Mapped[str] = mapped_column(String(2048), default="")
+    kb_type: Mapped[str] = mapped_column(String(32), default="document")
+    publication_status: Mapped[str] = mapped_column(String(16), default="private", index=True)
+    published_by: Mapped[str | None] = mapped_column(String(64), nullable=True)
+    reviewed_by: Mapped[str | None] = mapped_column(String(64), nullable=True)
+    published_at: Mapped[datetime | None] = mapped_column(BeijingDateTime(), nullable=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))
+
+    def to_dict(self) -> dict:
+        return {column.name: getattr(self, column.name) for column in self.__table__.columns}
+
+
+class LlmWikiConversationDepositRow(Base):
+    """Mapping from one DeerFlow answer turn to its WeKnora manual document.
+
+    The actual content lives in WeKnora.  DeerFlow keeps only the stable ids it
+    needs to rewrite/delete the document when the user regenerates or removes a
+    conversation turn.
+    """
+
+    __tablename__ = "llmwiki_conversation_deposits"
+    __table_args__ = (
+        UniqueConstraint(
+            "thread_id",
+            "assistant_message_id",
+            name="uq_llmwiki_conversation_deposit_answer",
+        ),
+    )
+
+    id: Mapped[str] = mapped_column(String(36), primary_key=True)
+    thread_id: Mapped[str] = mapped_column(String(128), index=True)
+    human_message_id: Mapped[str | None] = mapped_column(String(128), nullable=True, index=True)
+    assistant_message_id: Mapped[str] = mapped_column(String(128), index=True)
+    knowledge_base_mapping_id: Mapped[str] = mapped_column(String(36), index=True)
+    weknora_knowledge_id: Mapped[str | None] = mapped_column(String(128), nullable=True)
+    title: Mapped[str] = mapped_column(String(255), default="")
+    created_by_user_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))
+
+    def to_dict(self) -> dict:
+        return {column.name: getattr(self, column.name) for column in self.__table__.columns}

offline-backend-20260512/backend/packages/harness/deerflow/persistence/llmwiki/sql.py

@@ -0,0 +1,244 @@
+"""SQLAlchemy LLMWiki metadata repository."""
+
+from __future__ import annotations
+
+from datetime import UTC, datetime
+from typing import Any
+from uuid import uuid4
+
+from sqlalchemy import delete, or_, select
+from sqlalchemy.ext.asyncio import AsyncSession, async_sessionmaker
+
+from deerflow.persistence.llmwiki.base import LlmWikiScope, LlmWikiStore
+from deerflow.persistence.llmwiki.model import LlmWikiConversationDepositRow, LlmWikiKnowledgeBaseRow
+
+
+def _now() -> datetime:
+    return datetime.now(UTC)
+
+
+def _to_dict(row: LlmWikiKnowledgeBaseRow | LlmWikiConversationDepositRow) -> dict[str, Any]:
+    data = row.to_dict()
+    for key in ("published_at", "created_at", "updated_at"):
+        value = data.get(key)
+        if isinstance(value, datetime):
+            data[key] = value.isoformat()
+    return data
+
+
+class LlmWikiRepository(LlmWikiStore):
+    def __init__(self, session_factory: async_sessionmaker[AsyncSession]) -> None:
+        self._sf = session_factory
+
+    async def create_mapping(self, *, weknora_id: str, owner_user_id: str, name: str, description: str, kb_type: str) -> dict[str, Any]:
+        row = LlmWikiKnowledgeBaseRow(
+            id=str(uuid4()),
+            weknora_id=weknora_id,
+            owner_user_id=owner_user_id,
+            name=name,
+            description=description,
+            kb_type=kb_type,
+            publication_status="private",
+            created_at=_now(),
+            updated_at=_now(),
+        )
+        async with self._sf() as session:
+            session.add(row)
+            await session.commit()
+        return _to_dict(row)
+
+    async def list_visible(self, user_id: str, *, scope: LlmWikiScope = "all", is_admin: bool = False) -> list[dict[str, Any]]:
+        stmt = select(LlmWikiKnowledgeBaseRow)
+        if scope == "personal" and not is_admin:
+            stmt = stmt.where(LlmWikiKnowledgeBaseRow.owner_user_id == user_id)
+        elif scope == "public":
+            stmt = stmt.where(LlmWikiKnowledgeBaseRow.publication_status == "published")
+        elif scope == "all" and not is_admin:
+            stmt = stmt.where(
+                or_(
+                    LlmWikiKnowledgeBaseRow.owner_user_id == user_id,
+                    LlmWikiKnowledgeBaseRow.publication_status == "published",
+                )
+            )
+        stmt = stmt.order_by(LlmWikiKnowledgeBaseRow.created_at.desc())
+        async with self._sf() as session:
+            result = await session.execute(stmt)
+            return [_to_dict(row) for row in result.scalars()]
+
+    async def get_authorized(self, mapping_id: str, user_id: str, *, write: bool, is_admin: bool) -> dict[str, Any] | None:
+        stmt = select(LlmWikiKnowledgeBaseRow).where(LlmWikiKnowledgeBaseRow.id == mapping_id)
+        if not is_admin:
+            condition = LlmWikiKnowledgeBaseRow.owner_user_id == user_id
+            if not write:
+                condition = or_(condition, LlmWikiKnowledgeBaseRow.publication_status == "published")
+            stmt = stmt.where(condition)
+        async with self._sf() as session:
+            row = (await session.execute(stmt)).scalar_one_or_none()
+            return _to_dict(row) if row else None
+
+    async def get_by_weknora_id(self, weknora_id: str) -> dict[str, Any] | None:
+        stmt = select(LlmWikiKnowledgeBaseRow).where(LlmWikiKnowledgeBaseRow.weknora_id == weknora_id)
+        async with self._sf() as session:
+            row = (await session.execute(stmt)).scalar_one_or_none()
+            return _to_dict(row) if row else None
+
+    async def update_mapping(self, mapping_id: str, *, name: str | None = None, description: str | None = None) -> dict[str, Any] | None:
+        async with self._sf() as session:
+            row = await session.get(LlmWikiKnowledgeBaseRow, mapping_id)
+            if row is None:
+                return None
+            if name is not None:
+                row.name = name
+            if description is not None:
+                row.description = description
+            row.updated_at = _now()
+            await session.commit()
+            return _to_dict(row)
+
+    async def delete_mapping(self, mapping_id: str) -> bool:
+        async with self._sf() as session:
+            row = await session.get(LlmWikiKnowledgeBaseRow, mapping_id)
+            if row is None:
+                return False
+            await session.delete(row)
+            await session.commit()
+            return True
+
+    async def request_publish(self, mapping_id: str, actor_user_id: str, *, is_admin: bool) -> dict[str, Any] | None:
+        async with self._sf() as session:
+            row = await session.get(LlmWikiKnowledgeBaseRow, mapping_id)
+            if row is None or (not is_admin and row.owner_user_id != actor_user_id):
+                return None
+            row.publication_status = "published"
+            row.published_by = actor_user_id
+            row.reviewed_by = None
+            row.published_at = _now()
+            row.updated_at = _now()
+            await session.commit()
+            return _to_dict(row)
+
+    async def review_publish(self, mapping_id: str, reviewer_user_id: str, *, approved: bool) -> dict[str, Any] | None:
+        async with self._sf() as session:
+            row = await session.get(LlmWikiKnowledgeBaseRow, mapping_id)
+            if row is None or row.publication_status != "pending":
+                return None
+            row.publication_status = "published" if approved else "rejected"
+            row.reviewed_by = reviewer_user_id
+            row.published_at = _now() if approved else None
+            row.updated_at = _now()
+            await session.commit()
+            return _to_dict(row)
+
+    async def unpublish(self, mapping_id: str, actor_user_id: str, *, is_admin: bool) -> dict[str, Any] | None:
+        async with self._sf() as session:
+            row = await session.get(LlmWikiKnowledgeBaseRow, mapping_id)
+            if row is None or (not is_admin and row.owner_user_id != actor_user_id):
+                return None
+            row.publication_status = "private"
+            if is_admin:
+                row.reviewed_by = actor_user_id
+            row.published_at = None
+            row.updated_at = _now()
+            await session.commit()
+            return _to_dict(row)
+
+    async def list_publications(self, *, status: str | None = None) -> list[dict[str, Any]]:
+        stmt = select(LlmWikiKnowledgeBaseRow).where(LlmWikiKnowledgeBaseRow.publication_status != "private")
+        if status:
+            stmt = stmt.where(LlmWikiKnowledgeBaseRow.publication_status == status)
+        stmt = stmt.order_by(LlmWikiKnowledgeBaseRow.updated_at.desc())
+        async with self._sf() as session:
+            result = await session.execute(stmt)
+            return [_to_dict(row) for row in result.scalars()]
+
+    async def get_conversation_deposit_mapping(self) -> dict[str, Any] | None:
+        stmt = (
+            select(LlmWikiKnowledgeBaseRow)
+            .where(
+                LlmWikiKnowledgeBaseRow.name == "对话沉淀",
+                LlmWikiKnowledgeBaseRow.publication_status == "published",
+            )
+            .order_by(LlmWikiKnowledgeBaseRow.created_at.asc())
+            .limit(1)
+        )
+        async with self._sf() as session:
+            row = (await session.execute(stmt)).scalar_one_or_none()
+            return _to_dict(row) if row else None
+
+    async def get_conversation_deposit(
+        self,
+        *,
+        thread_id: str,
+        assistant_message_id: str,
+    ) -> dict[str, Any] | None:
+        stmt = select(LlmWikiConversationDepositRow).where(
+            LlmWikiConversationDepositRow.thread_id == thread_id,
+            LlmWikiConversationDepositRow.assistant_message_id == assistant_message_id,
+        )
+        async with self._sf() as session:
+            row = (await session.execute(stmt)).scalar_one_or_none()
+            return _to_dict(row) if row else None
+
+    async def upsert_conversation_deposit(
+        self,
+        *,
+        thread_id: str,
+        human_message_id: str | None,
+        assistant_message_id: str,
+        knowledge_base_mapping_id: str,
+        weknora_knowledge_id: str | None,
+        title: str,
+        created_by_user_id: str | None,
+    ) -> dict[str, Any]:
+        now = _now()
+        async with self._sf() as session:
+            stmt = select(LlmWikiConversationDepositRow).where(
+                LlmWikiConversationDepositRow.thread_id == thread_id,
+                LlmWikiConversationDepositRow.assistant_message_id == assistant_message_id,
+            )
+            row = (await session.execute(stmt)).scalar_one_or_none()
+            if row is None:
+                row = LlmWikiConversationDepositRow(
+                    id=str(uuid4()),
+                    thread_id=thread_id,
+                    human_message_id=human_message_id,
+                    assistant_message_id=assistant_message_id,
+                    knowledge_base_mapping_id=knowledge_base_mapping_id,
+                    weknora_knowledge_id=weknora_knowledge_id,
+                    title=title,
+                    created_by_user_id=created_by_user_id,
+                    created_at=now,
+                    updated_at=now,
+                )
+                session.add(row)
+            else:
+                row.human_message_id = human_message_id
+                row.knowledge_base_mapping_id = knowledge_base_mapping_id
+                row.weknora_knowledge_id = weknora_knowledge_id
+                row.title = title
+                row.updated_at = now
+            await session.commit()
+            return _to_dict(row)
+
+    async def delete_conversation_deposits(
+        self,
+        *,
+        thread_id: str,
+        message_ids: list[str] | None = None,
+    ) -> list[dict[str, Any]]:
+        conditions = [LlmWikiConversationDepositRow.thread_id == thread_id]
+        if message_ids:
+            conditions.append(
+                or_(
+                    LlmWikiConversationDepositRow.human_message_id.in_(message_ids),
+                    LlmWikiConversationDepositRow.assistant_message_id.in_(message_ids),
+                )
+            )
+        stmt = select(LlmWikiConversationDepositRow).where(*conditions)
+        async with self._sf() as session:
+            rows = (await session.execute(stmt)).scalars().all()
+            deleted = [_to_dict(row) for row in rows]
+            if rows:
+                await session.execute(delete(LlmWikiConversationDepositRow).where(LlmWikiConversationDepositRow.id.in_([row.id for row in rows])))
+                await session.commit()
+            return deleted

offline-backend-20260512/backend/packages/harness/deerflow/persistence/models/__init__.py

@@ -42,6 +42,7 @@ from deerflow.persistence.embed_sessions.model import EmbedSessionRow
 from deerflow.persistence.feedback.model import FeedbackRow
 from deerflow.persistence.light_apps.model import LightAppRow
 from deerflow.persistence.llm_metrics.model import LlmCallMetricRow
+from deerflow.persistence.llmwiki.model import LlmWikiConversationDepositRow, LlmWikiKnowledgeBaseRow
 from deerflow.persistence.memory_config.model import UserMemoryConfigRow
 from deerflow.persistence.menu_overrides.model import MenuOverrideRow
 from deerflow.persistence.models.run_event import RunEventRow
@@ -108,6 +109,8 @@ __all__ = [
     "KnowledgeTagRow",
     "LightAppRow",
     "LlmCallMetricRow",
+    "LlmWikiKnowledgeBaseRow",
+    "LlmWikiConversationDepositRow",
     "MenuOverrideRow",
     "NotificationRow",
     "ParallelAgentTaskRow",

offline-backend-20260512/backend/packages/harness/deerflow/runtime/references.py

@@ -13,7 +13,6 @@ from deerflow.config.extensions_config import (
     get_extensions_config,
 )
 
-
 _SKILL_PATH_RE = re.compile(r"[\\/]skills[\\/](?:public|custom)[\\/]([^\\/]+)(?:[\\/]|$)")
 
 
@@ -27,7 +26,11 @@ def build_reference_batches(messages: list[Any]) -> list[dict[str, Any]]:
             if _message_type(messages[cursor]) == "human":
                 end = cursor
                 break
-        sources = _collect_sources(messages[index + 1 : end])
+        preloaded_sources = _collect_preloaded_sources(message, start_index=1)
+        sources = [
+            *preloaded_sources,
+            *_collect_sources(messages[index + 1 : end], start_index=len(preloaded_sources) + 1),
+        ]
         if sources:
             anchor = _find_reference_batch_anchor(messages, index + 1, end)
             message_id = _message_id(anchor) or _message_id(message) or str(index)
@@ -46,7 +49,11 @@ def build_latest_reference_batch(messages: list[Any]) -> dict[str, Any] | None:
     human_index = _find_latest_human_index(messages)
     if human_index is None:
         return None
-    sources = _collect_sources(messages[human_index + 1 :])
+    preloaded_sources = _collect_preloaded_sources(messages[human_index], start_index=1)
+    sources = [
+        *preloaded_sources,
+        *_collect_sources(messages[human_index + 1 :], start_index=len(preloaded_sources) + 1),
+    ]
     if not sources:
         return None
     assistant = _find_reference_batch_anchor(messages, human_index + 1, len(messages))
@@ -365,7 +372,7 @@ def _source_debug_item(source: dict[str, Any]) -> dict[str, Any]:
     }
 
 
-def _collect_sources(messages: list[Any]) -> list[dict[str, Any]]:
+def _collect_sources(messages: list[Any], *, start_index: int = 1) -> list[dict[str, Any]]:
     sources: list[dict[str, Any]] = []
     pending_tool_contexts: list[dict[str, Any]] = []
     tool_contexts_by_id: dict[str, dict[str, Any]] = {}
@@ -393,7 +400,7 @@ def _collect_sources(messages: list[Any]) -> list[dict[str, Any]]:
         )
         if configured_sources is not None:
             for normalized in configured_sources:
-                normalized["index"] = len(sources) + 1
+                normalized["index"] = start_index + len(sources)
                 sources.append(normalized)
             continue
         for raw_item in _first_list(payload):
@@ -405,11 +412,37 @@ def _collect_sources(messages: list[Any]) -> list[dict[str, Any]]:
             )
             if not normalized:
                 continue
-            normalized["index"] = len(sources) + 1
+            normalized["index"] = start_index + len(sources)
             sources.append(normalized)
     return sources
 
 
+def _collect_preloaded_sources(message: Any, *, start_index: int = 1) -> list[dict[str, Any]]:
+    extra = _message_additional_kwargs(message)
+    raw_sources = extra.get("llmwiki_reference_sources")
+    if not isinstance(raw_sources, list):
+        return []
+    sources: list[dict[str, Any]] = []
+    for raw in raw_sources:
+        if not isinstance(raw, dict):
+            continue
+        source = dict(raw)
+        index = start_index + len(sources)
+        source["index"] = index
+        source.setdefault("skillName", "llmwiki_search")
+        source.setdefault("mode", "citation")
+        source.setdefault("type", "llmwiki")
+        title = str(source.get("title") or "").strip()
+        if not title:
+            source["title"] = f"LLMWiki 片段 {index}"
+        snippet = str(source.get("snippet") or source.get("content") or "").strip()
+        source["snippet"] = snippet[:360]
+        if "raw" not in source or not isinstance(source.get("raw"), dict):
+            source["raw"] = {key: value for key, value in source.items() if key != "raw"}
+        sources.append(source)
+    return sources
+
+
 def _collect_configured_sources(
     payload: Any,
     fallback_index: int,
@@ -486,6 +519,14 @@ def _message_content(message: Any) -> Any:
     return message.get("content") if isinstance(message, dict) else getattr(message, "content", None)
 
 
+def _message_additional_kwargs(message: Any) -> dict[str, Any]:
+    if isinstance(message, dict):
+        value = message.get("additional_kwargs") or message.get("additionalKwargs")
+    else:
+        value = getattr(message, "additional_kwargs", None)
+    return value if isinstance(value, dict) else {}
+
+
 def _message_name(message: Any) -> str:
     if isinstance(message, dict):
         return str(message.get("name") or (message.get("additional_kwargs") or {}).get("name") or "tool")

offline-backend-20260512/backend/packages/harness/deerflow/tools/builtins/llmwiki_search_tool.py

@@ -0,0 +1,132 @@
+"""Authorized read-only WeKnora retrieval for DeerFlow Q&A."""
+
+from __future__ import annotations
+
+import json
+from collections.abc import Mapping
+from typing import Any
+from urllib.parse import urlencode
+
+from langchain.tools import ToolRuntime, tool
+from langgraph.typing import ContextT
+
+from deerflow.agents.thread_state import ThreadState
+from deerflow.config.agents_config import load_agent_config
+from deerflow.integrations.weknora.client import WeKnoraError
+from deerflow.integrations.weknora.runtime import build_weknora_client, get_llmwiki_store, get_resolved_llmwiki_runtime
+from deerflow.runtime.user_context import get_current_user, get_effective_user_id
+
+
+def selected_knowledge_base_ids(context: Mapping[str, Any] | None) -> list[str] | None:
+    """Read the conversation selection; ``None`` means legacy/no selection key."""
+
+    if not context or "llmwiki_knowledge_base_ids" not in context:
+        return None
+    raw = context.get("llmwiki_knowledge_base_ids")
+    if not isinstance(raw, list):
+        return []
+    return list(dict.fromkeys(str(item).strip() for item in raw if str(item).strip()))[:100]
+
+
+def _agent_allowed_ids(context: Mapping[str, Any] | None) -> list[str] | None:
+    agent_id = str((context or {}).get("agent_id") or "").strip()
+    if not agent_id or agent_id in {"default", "lead_agent"}:
+        return None
+    try:
+        config = load_agent_config(agent_id)
+    except (FileNotFoundError, ValueError):
+        return []
+    return list(config.llmwiki_knowledge_base_ids or []) if config else []
+
+
+@tool("llmwiki_search", parse_docstring=True)
+async def llmwiki_search_tool(
+    query: str,
+    runtime: ToolRuntime[ContextT, ThreadState],
+) -> str:
+    """Search the LLMWiki knowledge selected for this conversation.
+
+    Use this tool when an answer can benefit from the user's managed knowledge
+    bases or the shared public LLMWiki. It only retrieves passages; it cannot
+    upload, edit, delete, publish or change permissions.
+
+    Args:
+        query: Focused natural-language retrieval query.
+    """
+
+    resolved = get_resolved_llmwiki_runtime()
+    if not resolved.weknora_enabled:
+        return "LLMWiki retrieval is unavailable because WeKnora mode is not enabled."
+
+    store = get_llmwiki_store()
+    user_id = get_effective_user_id()
+    is_admin = getattr(get_current_user(), "system_role", None) == "admin"
+    context = runtime.context if runtime.context else {}
+    selected_ids = selected_knowledge_base_ids(context)
+    allowed_ids = _agent_allowed_ids(context)
+    if selected_ids is None:
+        selected_ids = allowed_ids
+    elif allowed_ids is not None and not set(selected_ids).issubset(set(allowed_ids)):
+        return "The conversation requested knowledge outside this agent's configured LLMWiki scope."
+
+    if selected_ids == []:
+        return "No LLMWiki knowledge base is selected for this conversation."
+    if selected_ids is not None:
+        rows = []
+        for mapping_id in selected_ids:
+            row = await store.get_authorized(
+                mapping_id,
+                user_id,
+                write=False,
+                is_admin=is_admin,
+            )
+            if row is None:
+                return f"Knowledge base {mapping_id!r} is not available to the current user."
+            rows.append(row)
+    else:
+        rows = await store.list_visible(user_id, scope="all", is_admin=is_admin)
+    if not rows:
+        return "No authorized LLMWiki knowledge bases are available."
+
+    try:
+        results = await build_weknora_client(resolved).search(
+            query.strip(),
+            [str(row["weknora_id"]) for row in rows],
+        )
+    except (RuntimeError, WeKnoraError) as exc:
+        return f"LLMWiki retrieval failed: {exc}"
+
+    local_by_remote = {str(row["weknora_id"]): row for row in rows}
+    citations = []
+    for result in results:
+        mapping = local_by_remote.get(str(result.get("knowledge_base_id") or ""))
+        if mapping is None:
+            continue
+        params = urlencode(
+            {
+                "knowledge_base_id": mapping["id"],
+                "chunk_id": result.get("chunk_id") or "",
+            }
+        )
+        citations.append(
+            {
+                "title": result.get("title") or result.get("filename") or mapping.get("name") or "LLMWiki source",
+                "content": result.get("content") or "",
+                "score": result.get("score") or 0,
+                "knowledge_base_id": mapping["id"],
+                "knowledge_base_name": mapping.get("name") or "",
+                "source": mapping.get("name") or "LLMWiki",
+                "knowledge_id": result.get("knowledge_id") or "",
+                "chunk_id": result.get("chunk_id") or "",
+                "url": f"/weknora-source-preview?{params}",
+            }
+        )
+    return json.dumps(
+        {
+            "query": query,
+            "selection": [row["id"] for row in rows],
+            "knowledge_base_count": len(rows),
+            "results": citations,
+        },
+        ensure_ascii=False,
+    )

offline-backend-20260512/backend/packages/harness/deerflow/tools/tools.py

@@ -75,7 +75,11 @@ def get_available_tools(
         List of available tools.
     """
     config = app_config or get_app_config()
-    tool_configs = [tool for tool in config.tools if groups is None or tool.group in groups]
+    selected_tool_configs = [tool for tool in config.tools if groups is None or tool.group in groups]
+    disabled_tool_names = [tool.name for tool in selected_tool_configs if not tool.enabled]
+    if disabled_tool_names:
+        logger.info("Skipping disabled configured tool(s): %s", disabled_tool_names)
+    tool_configs = [tool for tool in selected_tool_configs if tool.enabled]
 
     browser_context_config = getattr(config, "browser_context", None)
     if not getattr(browser_context_config, "enabled", False):
@@ -154,9 +158,21 @@ def get_available_tools(
     # 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。
-    from deerflow.tools.builtins.roundtable_peers_tool import read_peer_delivery_tool
-
-    builtin_tools.append(read_peer_delivery_tool)
+    from deerflow.tools.builtins.roundtable_peers_tool import read_peer_delivery_tool
+
+    builtin_tools.append(read_peer_delivery_tool)
+
+    # Conditional LLMWiki mode: the tool is absent when no effective WeKnora
+    # address is configured, keeping legacy deployments behavior-identical.
+    try:
+        from deerflow.integrations.weknora.runtime import get_resolved_llmwiki_runtime
+
+        if get_resolved_llmwiki_runtime(config).weknora_enabled:
+            from deerflow.tools.builtins.llmwiki_search_tool import llmwiki_search_tool
+
+            builtin_tools.append(llmwiki_search_tool)
+    except Exception:  # noqa: BLE001 - configuration/store failures must not block agent creation
+        logger.debug("Failed to evaluate the conditional LLMWiki search tool", exc_info=True)
 
     # skill_manage 能创建/修改/删除技能,属于"自我进化"写能力,按 skill_evolution.enabled
     # 门控。该开关用户可在 per-user 配置里覆盖,所以按"生效配置(含 per-user 覆盖)"判断 ——

offline-backend-20260512/backend/packages/harness/tests/test_configurable_web_search.py

@@ -0,0 +1,229 @@
+from __future__ import annotations
+
+import json
+
+import httpx
+import pytest
+
+from deerflow.community.configurable_search.tools import (
+    ConfigurableSearchSettings,
+    execute_search,
+)
+from deerflow.config.app_config import AppConfig
+from deerflow.config.tool_config import ToolConfig
+from deerflow.tools.tools import get_available_tools
+
+
+def _settings(**overrides: object) -> ConfigurableSearchSettings:
+    values = {
+        "enabled": True,
+        "endpoint": "https://search.example.test/searchName",
+        "payload": {"query": "configured value", "xxx": "xxxx"},
+        "verify_ssl": False,
+        "timeout": 30.0,
+        "max_results": 5,
+        "result_url_template": "https://search.example.test/mindId/xxx?recUuid={recUuid}",
+    }
+    values.update(overrides)
+    return ConfigurableSearchSettings(**values)  # type: ignore[arg-type]
+
+
+@pytest.mark.asyncio
+async def test_configurable_search_overrides_query_and_normalizes_nullable_results() -> None:
+    requests: list[httpx.Request] = []
+
+    async def handler(request: httpx.Request) -> httpx.Response:
+        requests.append(request)
+        assert request.method == "POST"
+        assert json.loads(request.content) == {"query": "测试检索词", "xxx": "xxxx"}
+        return httpx.Response(
+            200,
+            json={
+                "results": [
+                    {
+                        "recUuid": "record/123",
+                        "content": "检索内容",
+                        "title": "标题",
+                        "url": "https://original.example.test/one",
+                    },
+                    {
+                        "recUuid": None,
+                        "content": None,
+                        "title": None,
+                        "url": "https://original.example.test/two",
+                    },
+                    "invalid row",
+                ]
+            },
+        )
+
+    settings = _settings()
+    result = json.loads(
+        await execute_search(
+            "  测试检索词  ",
+            settings,
+            transport=httpx.MockTransport(handler),
+        )
+    )
+
+    assert len(requests) == 1
+    assert settings.payload == {"query": "configured value", "xxx": "xxxx"}
+    assert result == {
+        "query": "测试检索词",
+        "total_results": 2,
+        "results": [
+            {
+                "recUuid": "record/123",
+                "content": "检索内容",
+                "title": "标题",
+                "url": "https://search.example.test/mindId/xxx?recUuid=record%2F123",
+            },
+            {
+                "recUuid": "",
+                "content": "",
+                "title": "",
+                "url": "https://original.example.test/two",
+            },
+        ],
+    }
+
+
+@pytest.mark.asyncio
+async def test_configurable_search_handles_empty_results() -> None:
+    async def handler(_: httpx.Request) -> httpx.Response:
+        return httpx.Response(200, json={"results": []})
+
+    result = json.loads(
+        await execute_search(
+            "nothing",
+            _settings(),
+            transport=httpx.MockTransport(handler),
+        )
+    )
+
+    assert result == {"query": "nothing", "total_results": 0, "results": []}
+
+
+@pytest.mark.asyncio
+@pytest.mark.parametrize(
+    ("response_factory", "expected_error"),
+    [
+        (lambda request: httpx.Response(503, request=request), "Search API returned HTTP 503"),
+        (
+            lambda request: httpx.Response(200, content=b"not-json", request=request),
+            "Search API request failed: JSONDecodeError",
+        ),
+        (
+            lambda request: httpx.Response(200, json={"results": None}, request=request),
+            "search API response field 'results' must be an array",
+        ),
+    ],
+)
+async def test_configurable_search_returns_safe_errors(
+    response_factory,
+    expected_error: str,
+) -> None:
+    async def handler(request: httpx.Request) -> httpx.Response:
+        return response_factory(request)
+
+    result = json.loads(
+        await execute_search(
+            "query",
+            _settings(),
+            transport=httpx.MockTransport(handler),
+        )
+    )
+
+    assert result["results"] == []
+    assert result["total_results"] == 0
+    assert result["error"] == expected_error
+
+
+@pytest.mark.asyncio
+async def test_configurable_search_handles_timeout() -> None:
+    async def handler(request: httpx.Request) -> httpx.Response:
+        raise httpx.ReadTimeout("timed out", request=request)
+
+    result = json.loads(
+        await execute_search(
+            "query",
+            _settings(),
+            transport=httpx.MockTransport(handler),
+        )
+    )
+
+    assert result["error"] == "Search API request timed out"
+
+
+@pytest.mark.asyncio
+async def test_configurable_search_passes_verify_ssl_to_httpx(monkeypatch: pytest.MonkeyPatch) -> None:
+    captured: dict[str, object] = {}
+
+    class FakeAsyncClient:
+        def __init__(self, **kwargs: object) -> None:
+            captured.update(kwargs)
+
+        async def __aenter__(self):
+            return self
+
+        async def __aexit__(self, *_: object) -> None:
+            return None
+
+        async def post(self, url: str, *, json: dict[str, object]) -> httpx.Response:
+            request = httpx.Request("POST", url, json=json)
+            return httpx.Response(200, json={"results": []}, request=request)
+
+    monkeypatch.setattr(
+        "deerflow.community.configurable_search.tools.httpx.AsyncClient",
+        FakeAsyncClient,
+    )
+
+    await execute_search("query", _settings(verify_ssl=False))
+
+    assert captured["verify"] is False
+    assert captured["timeout"] == 30.0
+
+
+def test_disabled_configured_tool_is_not_registered() -> None:
+    config = AppConfig(
+        sandbox={"use": "deerflow.sandbox.local:LocalSandboxProvider"},
+        tools=[
+            ToolConfig(
+                name="web_search",
+                group="web",
+                enabled=False,
+                use="deerflow.community.configurable_search.tools:web_search_tool",
+                endpoint="https://search.example.test/searchName",
+            )
+        ],
+    )
+
+    available = get_available_tools(
+        groups=["web"],
+        include_mcp=False,
+        app_config=config,
+    )
+
+    assert "web_search" not in {item.name for item in available}
+
+    config.tools[0].enabled = True
+    enabled_tools = get_available_tools(
+        groups=["web"],
+        include_mcp=False,
+        app_config=config,
+    )
+
+    assert "web_search" in {item.name for item in enabled_tools}
+
+
+def test_config_validation_requires_rec_uuid_url_placeholder() -> None:
+    config = ToolConfig(
+        name="web_search",
+        group="web",
+        use="deerflow.community.configurable_search.tools:web_search_tool",
+        endpoint="https://search.example.test/searchName",
+        result_url_template="https://search.example.test/mindId/xxx",
+    )
+
+    with pytest.raises(ValueError, match="must contain"):
+        ConfigurableSearchSettings.from_tool_config(config)

offline-backend-20260512/backend/tests/test_health_auth_bypass.py

@@ -0,0 +1,100 @@
+"""Regression coverage for the unauthenticated Gateway health probes."""
+
+from __future__ import annotations
+
+import pytest
+from fastapi import FastAPI
+from fastapi.testclient import TestClient
+
+from app.gateway import auth_middleware
+from app.gateway.app import app as gateway_app
+from app.gateway.auth_middleware import PUBLIC_HEALTH_PATHS, AuthMiddleware
+
+
+def _health_only_app() -> FastAPI:
+    app = FastAPI()
+
+    async def health_check() -> dict[str, str]:
+        return {"status": "healthy", "service": "deer-flow-gateway"}
+
+    for path in PUBLIC_HEALTH_PATHS:
+        app.add_api_route(path, health_check, methods=["GET", "HEAD"])
+    app.add_middleware(AuthMiddleware)
+    return app
+
+
+@pytest.mark.parametrize("path", sorted(PUBLIC_HEALTH_PATHS))
+@pytest.mark.parametrize("method", ["GET", "HEAD"])
+def test_health_probe_is_public_when_auth_is_enabled(
+    monkeypatch: pytest.MonkeyPatch,
+    path: str,
+    method: str,
+) -> None:
+    monkeypatch.setattr(auth_middleware, "is_auth_disabled", lambda: False)
+
+    with TestClient(_health_only_app()) as client:
+        client.cookies.set("access_token", "invalid-token")
+        response = client.request(
+            method,
+            path,
+            headers={"Authorization": "Bearer invalid-token"},
+        )
+
+    assert response.status_code == 200
+    if method == "GET":
+        assert response.json() == {"status": "healthy", "service": "deer-flow-gateway"}
+    else:
+        assert response.content == b""
+
+
+@pytest.mark.parametrize("path", sorted(PUBLIC_HEALTH_PATHS))
+def test_health_probe_allows_a_trailing_slash(
+    monkeypatch: pytest.MonkeyPatch,
+    path: str,
+) -> None:
+    monkeypatch.setattr(auth_middleware, "is_auth_disabled", lambda: False)
+
+    with TestClient(_health_only_app(), follow_redirects=False) as client:
+        response = client.get(f"{path}/")
+
+    # FastAPI redirects to the canonical route; importantly, auth must not turn
+    # the request into a 401 before routing can perform that redirect.
+    assert response.status_code == 307
+
+
+@pytest.mark.parametrize("path", sorted(PUBLIC_HEALTH_PATHS))
+def test_health_probe_is_public_behind_a_reverse_proxy_prefix(
+    monkeypatch: pytest.MonkeyPatch,
+    path: str,
+) -> None:
+    monkeypatch.setattr(auth_middleware, "is_auth_disabled", lambda: False)
+
+    with TestClient(_health_only_app(), root_path="/deerflow") as client:
+        response = client.get(f"/deerflow{path}")
+
+    assert response.status_code == 200
+
+
+def test_gateway_registers_every_public_health_path_for_get_and_head() -> None:
+    methods_by_path = {
+        route.path: route.methods
+        for route in gateway_app.routes
+        if hasattr(route, "methods")
+    }
+
+    for path in PUBLIC_HEALTH_PATHS:
+        assert {"GET", "HEAD"} <= methods_by_path[path]
+
+
+def test_non_health_route_still_requires_auth(monkeypatch: pytest.MonkeyPatch) -> None:
+    monkeypatch.setattr(auth_middleware, "is_auth_disabled", lambda: False)
+    app = _health_only_app()
+
+    @app.get("/private")
+    async def private() -> dict[str, bool]:
+        return {"ok": True}
+
+    with TestClient(app) as client:
+        response = client.get("/private")
+
+    assert response.status_code == 401

offline-backend-20260512/backend/tests/test_llmwiki_weknora.py

@@ -0,0 +1,418 @@
+from __future__ import annotations
+
+import json
+
+import httpx
+import pytest
+from fastapi import HTTPException
+from starlette.datastructures import Headers, QueryParams
+
+from deerflow.config.agents_config import AgentConfig
+from deerflow.config.llmwiki_config import (
+    LlmWikiConfig,
+    LlmWikiRuntimeOverride,
+    resolve_llmwiki_runtime,
+)
+from deerflow.integrations.weknora.client import WeKnoraClient, WeKnoraError
+from deerflow.persistence.llmwiki import MemoryLlmWikiStore
+from deerflow.tools.builtins.llmwiki_search_tool import selected_knowledge_base_ids
+
+
+def test_empty_weknora_address_keeps_legacy_mode() -> None:
+    runtime = resolve_llmwiki_runtime(LlmWikiConfig())
+
+    assert runtime.provider == "legacy"
+    assert runtime.api_base_url == ""
+
+
+def test_configured_weknora_address_enables_weknora_mode() -> None:
+    config = LlmWikiConfig.model_validate(
+        {
+            "weknora": {
+                "api_base_url": "http://weknora-app:8080/",
+                "web_base_url": "https://weknora.example.test/",
+                "admin_email": "admin@example.test",
+                "admin_password": "secret",
+            }
+        }
+    )
+
+    runtime = resolve_llmwiki_runtime(config)
+
+    assert runtime.provider == "weknora"
+    assert runtime.api_base_url == "http://weknora-app:8080"
+    assert runtime.web_base_url == "https://weknora.example.test"
+    assert runtime.admin_email == "admin@example.test"
+    assert runtime.admin_password == "secret"
+
+
+def test_runtime_override_can_force_legacy_or_reset_to_file() -> None:
+    config = LlmWikiConfig.model_validate({"weknora": {"api_base_url": "http://weknora:8080"}})
+
+    forced_legacy = resolve_llmwiki_runtime(
+        config,
+        LlmWikiRuntimeOverride(enabled=True, api_base_url=""),
+    )
+    reset_to_file = resolve_llmwiki_runtime(
+        config,
+        LlmWikiRuntimeOverride(enabled=False, api_base_url=""),
+    )
+
+    assert forced_legacy.provider == "legacy"
+    assert reset_to_file.provider == "weknora"
+
+
+def test_weknora_url_rejects_credentials_and_non_http_schemes() -> None:
+    with pytest.raises(ValueError):
+        LlmWikiConfig.model_validate({"weknora": {"api_base_url": "ftp://weknora.local"}})
+    with pytest.raises(ValueError):
+        LlmWikiConfig.model_validate({"weknora": {"api_base_url": "http://user:pass@weknora.local"}})
+
+
+def test_agent_config_tracks_allowed_llmwiki_knowledge_bases() -> None:
+    config = AgentConfig(
+        name="Research agent",
+        llmwiki_knowledge_base_ids=["kb-one", "kb-two", "kb-one"],
+    )
+
+    assert config.llmwiki_knowledge_base_ids == ["kb-one", "kb-two"]
+
+
+def test_conversation_selection_is_deduplicated_and_absence_is_distinct() -> None:
+    assert selected_knowledge_base_ids({}) is None
+    assert selected_knowledge_base_ids({"llmwiki_knowledge_base_ids": []}) == []
+    assert selected_knowledge_base_ids({"llmwiki_knowledge_base_ids": ["kb-one", "kb-one", "", "kb-two"]}) == ["kb-one", "kb-two"]
+
+
+@pytest.mark.asyncio
+async def test_weknora_client_sends_no_credentials_and_normalizes_search() -> None:
+    requests: list[httpx.Request] = []
+
+    async def handler(request: httpx.Request) -> httpx.Response:
+        requests.append(request)
+        assert "X-API-Key" not in request.headers
+        assert "Authorization" not in request.headers
+        assert json.loads(request.content) == {
+            "query": "deployment guide",
+            "knowledge_base_ids": ["kb-one", "kb-two"],
+        }
+        return httpx.Response(
+            200,
+            json={
+                "success": True,
+                "data": [
+                    {
+                        "id": "chunk-one",
+                        "content": "matched passage",
+                        "knowledge_id": "doc-one",
+                        "knowledge_base_id": "kb-one",
+                        "knowledge_title": "Guide",
+                        "score": 0.93,
+                    }
+                ],
+            },
+        )
+
+    async with httpx.AsyncClient(transport=httpx.MockTransport(handler)) as http_client:
+        client = WeKnoraClient(
+            base_url="http://weknora.local:8080",
+            http_client=http_client,
+        )
+        results = await client.search("deployment guide", ["kb-one", "kb-two"])
+
+    assert requests[0].url.path == "/api/v1/knowledge-search"
+    assert results[0]["chunk_id"] == "chunk-one"
+    assert results[0]["title"] == "Guide"
+
+
+@pytest.mark.asyncio
+async def test_weknora_client_resolves_default_embedding_model_inside_weknora() -> None:
+    async def handler(request: httpx.Request) -> httpx.Response:
+        if request.method == "GET" and request.url.path == "/api/v1/models":
+            return httpx.Response(
+                200,
+                json={
+                    "success": True,
+                    "data": [
+                        {"id": "embed-default", "type": "Embedding", "is_default": True},
+                    ],
+                },
+            )
+        assert request.method == "POST"
+        assert request.url.path == "/api/v1/knowledge-bases"
+        assert json.loads(request.content)["embedding_model_id"] == "embed-default"
+        return httpx.Response(201, json={"success": True, "data": {"id": "kb-created", "name": "Docs"}})
+
+    async with httpx.AsyncClient(transport=httpx.MockTransport(handler)) as http_client:
+        client = WeKnoraClient(
+            base_url="http://weknora.local:8080",
+            http_client=http_client,
+        )
+        created = await client.create_knowledge_base(name="Docs")
+
+    assert created["id"] == "kb-created"
+
+
+@pytest.mark.asyncio
+async def test_weknora_client_creates_wiki_with_weknora_owned_models() -> None:
+    model_reads = 0
+
+    async def handler(request: httpx.Request) -> httpx.Response:
+        nonlocal model_reads
+        if request.method == "GET" and request.url.path == "/api/v1/models":
+            model_reads += 1
+            return httpx.Response(
+                200,
+                json={
+                    "data": [
+                        {"id": "embed-one", "type": "Embedding", "is_default": True},
+                        {"id": "qa-one", "type": "KnowledgeQA", "status": "active"},
+                    ]
+                },
+            )
+        payload = json.loads(request.content)
+        assert payload["embedding_model_id"] == "embed-one"
+        assert payload["summary_model_id"] == "qa-one"
+        assert payload["indexing_strategy"] == {
+            "vector_enabled": True,
+            "keyword_enabled": True,
+            "wiki_enabled": True,
+            "graph_enabled": False,
+        }
+        assert payload["wiki_config"]["synthesis_model_id"] == "qa-one"
+        return httpx.Response(201, json={"data": {"id": "wiki-one", "name": "Wiki"}})
+
+    async with httpx.AsyncClient(transport=httpx.MockTransport(handler)) as http_client:
+        client = WeKnoraClient(base_url="http://weknora.local:8080", http_client=http_client)
+        created = await client.create_knowledge_base(name="Wiki", wiki_enabled=True)
+
+    assert created["id"] == "wiki-one"
+    assert model_reads == 2
+
+
+@pytest.mark.asyncio
+async def test_weknora_client_imports_url_manual_content_and_lists_wiki_pages() -> None:
+    async def handler(request: httpx.Request) -> httpx.Response:
+        if request.url.path.endswith("/knowledge/url"):
+            assert json.loads(request.content) == {"url": "https://example.test/guide"}
+            return httpx.Response(201, json={"data": {"id": "url-one"}})
+        if request.url.path.endswith("/knowledge/manual"):
+            assert json.loads(request.content) == {
+                "title": "Guide",
+                "content": "# Guide",
+                "status": "publish",
+            }
+            return httpx.Response(201, json={"data": {"id": "manual-one"}})
+        assert request.url.path == "/api/v1/knowledgebase/kb-one/wiki/pages"
+        assert request.url.params["query"] == "guide"
+        return httpx.Response(
+            200,
+            json={"pages": [{"slug": "guide", "title": "Guide"}], "total": 1, "page": 1, "page_size": 50},
+        )
+
+    async with httpx.AsyncClient(transport=httpx.MockTransport(handler)) as http_client:
+        client = WeKnoraClient(base_url="http://weknora.local:8080", http_client=http_client)
+        await client.import_document_url("kb-one", url="https://example.test/guide")
+        await client.create_manual_document("kb-one", title="Guide", content="# Guide")
+        pages = await client.list_wiki_pages("kb-one", query="guide")
+
+    assert pages["total"] == 1
+    assert pages["pages"][0]["slug"] == "guide"
+
+
+@pytest.mark.asyncio
+async def test_weknora_client_reads_document_chunks_and_wiki_graph() -> None:
+    async def handler(request: httpx.Request) -> httpx.Response:
+        if request.url.path == "/api/v1/chunks/doc-one":
+            assert request.url.params["page"] == "2"
+            assert request.url.params["page_size"] == "10"
+            return httpx.Response(
+                200,
+                json={
+                    "success": True,
+                    "data": [{"id": "chunk-one", "content": "parsed content", "chunk_index": 10}],
+                    "total": 31,
+                    "page": 2,
+                    "page_size": 10,
+                },
+            )
+        assert request.url.path == "/api/v1/knowledgebase/kb-one/wiki/graph"
+        assert request.url.params["mode"] == "overview"
+        assert request.url.params["limit"] == "50"
+        return httpx.Response(
+            200,
+            json={
+                "success": True,
+                "data": {
+                    "nodes": [{"slug": "entity/one", "title": "Entity One", "page_type": "entity"}],
+                    "edges": [],
+                    "meta": {"mode": "overview", "total": 1, "returned": 1},
+                },
+            },
+        )
+
+    async with httpx.AsyncClient(transport=httpx.MockTransport(handler)) as http_client:
+        client = WeKnoraClient(base_url="http://weknora.local:8080", http_client=http_client)
+        chunks = await client.list_chunks("doc-one", page=2, page_size=10)
+        graph = await client.get_wiki_graph("kb-one", limit=50)
+
+    assert chunks["total"] == 31
+    assert chunks["items"][0]["content"] == "parsed content"
+    assert graph["nodes"][0]["slug"] == "entity/one"
+
+
+@pytest.mark.asyncio
+async def test_weknora_client_proxies_original_document_preview_without_credentials() -> None:
+    async def handler(request: httpx.Request) -> httpx.Response:
+        assert request.url.path == "/api/v1/knowledge/doc-one/preview"
+        assert "X-API-Key" not in request.headers
+        assert "Authorization" not in request.headers
+        return httpx.Response(200, content=b"preview", headers={"Content-Type": "text/plain; charset=utf-8"})
+
+    async with httpx.AsyncClient(transport=httpx.MockTransport(handler)) as http_client:
+        client = WeKnoraClient(base_url="http://weknora.local:8080", http_client=http_client)
+        content, media_type = await client.get_document_preview("doc-one")
+
+    assert content == b"preview"
+    assert media_type == "text/plain"
+
+
+@pytest.mark.asyncio
+async def test_memory_store_enforces_owner_and_public_visibility() -> None:
+    store = MemoryLlmWikiStore()
+    created = await store.create_mapping(
+        weknora_id="wk-private",
+        owner_user_id="user-a",
+        name="Private notes",
+        description="",
+        kb_type="document",
+    )
+
+    assert [row["id"] for row in await store.list_visible("user-a", scope="personal")] == [created["id"]]
+    assert await store.list_visible("user-b", scope="personal") == []
+    assert await store.get_authorized(created["id"], "user-b", write=False, is_admin=False) is None
+
+    published = await store.request_publish(created["id"], "user-a", is_admin=False)
+    assert published and published["publication_status"] == "published"
+    assert [row["id"] for row in await store.list_visible("user-b", scope="public")] == [created["id"]]
+    assert await store.get_authorized(created["id"], "user-b", write=False, is_admin=False) is not None
+    assert await store.get_authorized(created["id"], "user-b", write=True, is_admin=False) is None
+
+
+@pytest.mark.asyncio
+async def test_delete_cleans_owner_mapping_when_remote_is_already_missing(monkeypatch: pytest.MonkeyPatch) -> None:
+    from app.gateway.routers import llmwiki as llmwiki_router
+
+    deleted: list[str] = []
+
+    class MissingClient:
+        async def delete_knowledge_base(self, _: str) -> None:
+            raise WeKnoraError("WeKnora resource was not found", status_code=404)
+
+    class Store:
+        async def delete_mapping(self, mapping_id: str) -> None:
+            deleted.append(mapping_id)
+
+    async def authorized_mapping(*_: object, **__: object) -> tuple[dict[str, str], str, bool]:
+        return {"weknora_id": "remote-missing"}, "owner", False
+
+    monkeypatch.setattr(llmwiki_router, "_authorized_mapping", authorized_mapping)
+    monkeypatch.setattr(llmwiki_router, "_client_or_503", lambda: MissingClient())
+    monkeypatch.setattr(llmwiki_router, "_store", lambda _: Store())
+
+    await llmwiki_router.delete_knowledge_base(object(), "mapping-one")
+
+    assert deleted == ["mapping-one"]
+
+
+def test_weknora_iframe_cookie_is_signed_and_short_lived() -> None:
+    from app.gateway.routers import llmwiki as llmwiki_router
+
+    raw = llmwiki_router._sign_embed_payload(
+        {
+            "mapping_id": "local-kb",
+            "weknora_id": "remote-kb",
+            "user_id": "user-a",
+            "is_admin": False,
+            "can_write": True,
+            "exp": 4_102_444_800,
+        }
+    )
+
+    ctx = llmwiki_router._verify_embed_cookie(raw)
+
+    assert ctx.mapping_id == "local-kb"
+    assert ctx.weknora_id == "remote-kb"
+    assert ctx.can_write is True
+
+    with pytest.raises(HTTPException):
+        llmwiki_router._verify_embed_cookie(raw + "tampered")
+
+
+def test_weknora_iframe_filters_kb_list_to_current_mapping() -> None:
+    from app.gateway.routers import llmwiki as llmwiki_router
+
+    ctx = llmwiki_router.WeKnoraEmbedContext(
+        mapping_id="local-kb",
+        weknora_id="remote-kb",
+        user_id="user-a",
+        is_admin=False,
+        can_write=False,
+        exp=4_102_444_800,
+    )
+    response = httpx.Response(
+        200,
+        json={
+            "success": True,
+            "data": [
+                {"id": "remote-kb", "name": "Allowed"},
+                {"id": "other-kb", "name": "Blocked"},
+            ],
+        },
+    )
+
+    filtered = llmwiki_router._filter_knowledge_base_list_response(response, ctx)
+
+    assert filtered is not None
+    payload = json.loads(filtered)
+    assert payload["data"] == [{"id": "remote-kb", "name": "Allowed"}]
+
+
+@pytest.mark.asyncio
+async def test_weknora_iframe_proxy_blocks_other_knowledge_bases() -> None:
+    from app.gateway.routers import llmwiki as llmwiki_router
+
+    class FakeRequest:
+        method = "GET"
+        headers = Headers({})
+        query_params = QueryParams("")
+
+    ctx = llmwiki_router.WeKnoraEmbedContext(
+        mapping_id="local-kb",
+        weknora_id="remote-kb",
+        user_id="user-a",
+        is_admin=False,
+        can_write=False,
+        exp=4_102_444_800,
+    )
+
+    await llmwiki_router._authorize_weknora_api_proxy(
+        FakeRequest(),
+        ctx,
+        object(),
+        "/api/v1/knowledge-bases/remote-kb",
+        b"",
+        "token",
+    )
+
+    with pytest.raises(HTTPException) as exc_info:
+        await llmwiki_router._authorize_weknora_api_proxy(
+            FakeRequest(),
+            ctx,
+            object(),
+            "/api/v1/knowledge-bases/other-kb",
+            b"",
+            "token",
+        )
+
+    assert exc_info.value.status_code == 403

offline-backend-20260512/offline-runtime-requirements.txt

@@ -1,8 +1,9 @@
 # Direct runtime packages layered into the prebuilt offline backend image.
-# Keep these three pins synchronized with start_backend.sh.
+# Keep these four pins synchronized with start_backend.sh.
 croniter==6.2.2
 hindsight-client==0.6.2
 redis==5.2.1
+aiomysql==0.3.2
 
 # Fully pinned dependency closure for CPython 3.12 on Linux. Keeping the
 # transitive packages here makes a newly generated wheelhouse reproducible.
@@ -25,3 +26,4 @@ typing-extensions==4.16.0
 typing-inspection==0.4.2
 urllib3==2.7.0
 yarl==1.24.5
+PyMySQL==1.2.0

offline-backend-20260512/prepare_wheelhouse.sh

@@ -59,6 +59,8 @@ echo "  Python/ABI:   ${TARGET_PYTHON_VERSION}/${TARGET_ABI}"
 require_wheel 'croniter-6.2.2-*.whl'
 require_wheel 'hindsight_client-0.6.2-*.whl'
 require_wheel 'redis-5.2.1-*.whl'
+require_wheel 'aiomysql-0.3.2-*.whl'
+require_wheel 'pymysql-1.2.0-*.whl'
 
 # Resolve the same requirements a second time using only wheelhouse. This
 # catches a missing transitive dependency before the bundle reaches offline.

offline-backend-20260512/start_backend.sh

@@ -130,6 +130,12 @@ if [ "${HINDSIGHT_INSTALL_DEPS:-1}" != "0" ]; then
     uv pip install --python "${PYTHON_BIN}" --no-index --find-links /app/wheelhouse ${HINDSIGHT_PIP_PACKAGES:-hindsight-client==0.6.2}
   }
 fi
+if [ "${AIOMYSQL_INSTALL_DEPS:-1}" != "0" ]; then
+  uv run --no-sync python -c "import aiomysql" >/dev/null 2>&1 || {
+    PYTHON_BIN="$(uv run --no-sync python -c "import sys; print(sys.executable)")"
+    uv pip install --python "${PYTHON_BIN}" --no-index --find-links /app/wheelhouse ${AIOMYSQL_PIP_PACKAGES:-aiomysql==0.3.2}
+  }
+fi
 uv run --no-sync python -c "import redis" >/dev/null 2>&1 || uv pip install --python "$(uv run --no-sync python -c "import sys; print(sys.executable)")" --no-index --find-links /app/wheelhouse ${REDIS_PIP_PACKAGES:-redis==5.2.1}
 PYTHONPATH=. uv run --no-sync python -m uvicorn app.gateway.app:app --host 0.0.0.0 --port 8001 --workers ${GATEWAY_WORKERS:-2}'