Merge branch 'master' of http://192.168.0.200/knowledge-base-bj/deerflow-code into master-nj

frontend-web/src/core/llmwiki/source-links.ts

@@ -27,6 +27,16 @@ export type KnowledgeSourceLike = {
   slug?: string | null;
 };
 
+/** Build a HashRouter URL without dropping a reverse-proxy path such as /web/. */
+export function buildAppHashUrl(path: string): string {
+  const route = path.startsWith("/") ? path : `/${path}`;
+  if (typeof window !== "undefined") {
+    return `${window.location.pathname || "/"}#${route}`;
+  }
+  const base = import.meta.env.BASE_URL || "/";
+  return `${base}#${route}`;
+}
+
 export function parseWeKnoraPreviewUrl(url?: string | null): WeKnoraSourceTarget | null {
   if (!url?.startsWith("/weknora-source-preview?")) {
     return null;
@@ -92,7 +102,7 @@ export function buildWeKnoraSourceIframeUrl(target: WeKnoraSourceTarget): string
   if (target.documentId) params.set("documentId", target.documentId);
   if (target.wikiSlug) params.set("wikiSlug", target.wikiSlug);
   const query = params.toString();
-  return `/#/page/workspace/knowledge/${encodeURIComponent(target.knowledgeBaseId)}${query ? `?${query}` : ""}`;
+  return buildAppHashUrl(`/page/workspace/knowledge/${encodeURIComponent(target.knowledgeBaseId)}${query ? `?${query}` : ""}`);
 }
 
 /**
@@ -103,7 +113,7 @@ export function buildWeKnoraWikiPageUrl(
   knowledgeBaseId: string,
   wikiSlug: string,
 ): string {
-  return `/#${buildWeKnoraWikiPagePath(knowledgeBaseId, wikiSlug)}`;
+  return buildAppHashUrl(buildWeKnoraWikiPagePath(knowledgeBaseId, wikiSlug));
 }
 
 export function buildWeKnoraWikiPagePath(
@@ -129,7 +139,7 @@ export function buildPublicKnowledgeWikiUrl(
   knowledgeBaseId: string,
   wikiSlug?: string,
 ): string {
-  return `/#${buildPublicKnowledgeWikiPath(knowledgeBaseId, wikiSlug)}`;
+  return buildAppHashUrl(buildPublicKnowledgeWikiPath(knowledgeBaseId, wikiSlug));
 }
 
 export function buildLegacyKnowledgeSourceUrl(source: KnowledgeSourceLike): string {
@@ -160,6 +170,7 @@ export function normalizeMalformedWeKnoraHashRoute(): boolean {
   const { pathname, search, hash } = window.location;
   if (!pathname.startsWith("/page/workspace/knowledge/")) return false;
   if (hash && !hash.startsWith("#/")) return false;
-  window.location.replace(`/#${pathname}${search}`);
+  const base = import.meta.env.BASE_URL || "/";
+  window.location.replace(`${base}#${pathname}${search}`);
   return true;
 }

frontend-web/src/core/llmwiki/wiki-page-markdown.ts

@@ -37,6 +37,34 @@ const WIKI_INDEX_TYPE_LABELS: Record<string, string> = {
 
 const WIKI_INDEX_TYPE_ORDER = ["summary", "entity", "concept"];
 
+export const WIKI_PROTECTED_IMAGE_MARKER = "/__llmwiki-protected-image__";
+
+const PROTECTED_WIKI_IMAGE_SOURCE =
+  /(?:storage:\/\/[0-9A-Za-z_-]+\/)?(?:resource|local|minio|cos|tos|s3|oss|ks3|obs):\/\/[^\s"'<>\])]+/gi;
+
+/**
+ * Browsers cannot load WeKnora's protected resource:// and object-storage
+ * schemes directly. Replace them with a safe marker before Markdown
+ * sanitization; the image component resolves that marker through DeerFlow.
+ */
+export function rewriteProtectedWikiImageSources(markdown: string): string {
+  return markdown.replace(
+    PROTECTED_WIKI_IMAGE_SOURCE,
+    (filePath) => `${WIKI_PROTECTED_IMAGE_MARKER}?file_path=${encodeURIComponent(filePath)}`,
+  );
+}
+
+export function parseProtectedWikiImageSource(source: string | undefined): string | null {
+  if (!source) return null;
+  try {
+    const url = new URL(source, window.location.origin);
+    if (url.pathname !== WIKI_PROTECTED_IMAGE_MARKER) return null;
+    return url.searchParams.get("file_path")?.trim() || null;
+  } catch {
+    return null;
+  }
+}
+
 export function stripWikiFrontmatter(value: string) {
   return value.replace(/^\s*---\s*\n[\s\S]*?\n---\s*(?:\n|$)/, "").trim();
 }

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

@@ -179,6 +179,28 @@ export interface LlmWikiWikiIndexStatus {
   last_error?: string | null;
   last_started_at?: string | null;
   last_completed_at?: string | null;
+  progress?: {
+    total_pages: number;
+    processed_pages: number;
+    vectorized_pages: number;
+    failed_pages: number;
+    vector_count: number;
+    completion_percent: number;
+  };
+  pages?: Array<{
+    id: string;
+    slug: string;
+    title: string;
+    page_type: string;
+    status: string;
+    index_status: string;
+    vector_count: number;
+    vectorized: boolean;
+    processed_in_current_sync: boolean;
+    indexed_at?: string | null;
+    updated_at?: string | null;
+    last_error?: string | null;
+  }>;
 }
 
 export interface LlmWikiSourceContext {
@@ -444,6 +466,24 @@ export async function getLlmWikiDocumentPreview(
   return response.blob();
 }
 
+export async function getLlmWikiProtectedFileUrl(
+  knowledgeBaseId: string,
+  filePath: string,
+  publicAccess = false,
+): Promise<string> {
+  const params = new URLSearchParams({ file_path: filePath });
+  if (publicAccess) {
+    return `${PUBLIC_API()}/knowledge-bases/${encodeURIComponent(knowledgeBaseId)}/files?${params.toString()}`;
+  }
+  const access = await readJson<{ token: string; expires_in: number }>(
+    await apiFetch(
+      `${API()}/knowledge-bases/${encodeURIComponent(knowledgeBaseId)}/files/access-url?${params.toString()}`,
+    ),
+    "读取 Wiki 图片",
+  );
+  return `${PUBLIC_API()}/files/${encodeURIComponent(access.token)}`;
+}
+
 export async function getLlmWikiGraph(knowledgeBaseId: string): Promise<LlmWikiGraph> {
   return readJson(
     await apiFetch(`${API()}/knowledge-bases/${encodeURIComponent(knowledgeBaseId)}/graph?limit=200`),

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

@@ -33,6 +33,7 @@ import { reasoningPlugins } from "@/core/streamdown";
 import {
   exportableWikiMarkdown,
   renderableWikiMarkdown,
+  rewriteProtectedWikiImageSources,
   safeWikiFileName,
   stripWikiFrontmatter,
   wikiLinkMarkdown,
@@ -49,6 +50,7 @@ import {
   type LlmWikiDocument,
   type LlmWikiWikiPage,
 } from "@/strategy-components/api/llmwiki";
+import { WikiProtectedImage } from "./WikiProtectedImage";
 
 type SourceContextResponse = {
   success?: boolean;
@@ -926,10 +928,18 @@ export function WeKnoraSourceDrawer({
                     </div>
                     <MarkdownContent
                       className="max-w-none"
-                      content={preview.text}
+                      content={rewriteProtectedWikiImageSources(preview.text)}
                       isLoading={false}
                       remarkPlugins={reasoningPlugins.remarkPlugins}
                       rehypePlugins={reasoningPlugins.rehypePlugins}
+                      components={{
+                        img: (props) => (
+                          <WikiProtectedImage
+                            {...props}
+                            knowledgeBaseId={target?.knowledgeBaseId || ""}
+                          />
+                        ),
+                      }}
                     />
                   </article>
                 </section>
@@ -964,10 +974,18 @@ export function WeKnoraSourceDrawer({
                   <div className="rounded-lg border bg-background p-5">
                     <MarkdownContent
                       className="max-w-none"
-                      content={wikiDisplayContent}
+                      content={rewriteProtectedWikiImageSources(wikiDisplayContent)}
                       isLoading={false}
                       remarkPlugins={reasoningPlugins.remarkPlugins}
                       rehypePlugins={reasoningPlugins.rehypePlugins}
+                      components={{
+                        img: (props) => (
+                          <WikiProtectedImage
+                            {...props}
+                            knowledgeBaseId={target?.knowledgeBaseId || ""}
+                          />
+                        ),
+                      }}
                       linkClassName={(href) =>
                         parseWeKnoraPreviewUrl(href)?.wikiSlug
                           ? "wiki-content-link font-medium text-[#be2323] decoration-[#be2323]/55 decoration-dashed hover:decoration-solid"

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

@@ -6,6 +6,7 @@ import { toast } from "sonner";
 import { Button } from "@/components/ui/button";
 import { Badge } from "@/components/ui/badge";
 import { Input } from "@/components/ui/input";
+import { Progress } from "@/components/ui/progress";
 import { Textarea } from "@/components/ui/textarea";
 import { MarkdownCodeEditor } from "@/components/wiki/MarkdownCodeEditor";
 import {
@@ -331,6 +332,10 @@ export function WeKnoraWikiManagementPage() {
     () => moveTargetOptions(pagesList),
     [pagesList],
   );
+  const pageIndexBySlug = useMemo(
+    () => new Map((indexStatus.data?.pages ?? []).map((item) => [item.slug, item])),
+    [indexStatus.data?.pages],
+  );
   const currentMoveTarget = moveTargetValue(draft.parentSlug, draft.categoryPathText, draft.folderId);
 
   useEffect(() => {
@@ -387,6 +392,17 @@ export function WeKnoraWikiManagementPage() {
     });
   };
 
+  const renderVectorBadge = (page: WikiPageListItem) => {
+    const item = pageIndexBySlug.get(page.slug);
+    if (item?.last_error || item?.index_status === "failed" || item?.index_status === "stale") {
+      return <Badge variant="destructive" className="ml-auto shrink-0 text-[10px]" title={item?.last_error || "向量化失败"}>失败</Badge>;
+    }
+    if (item?.vectorized) {
+      return <Badge variant="secondary" className="ml-auto shrink-0 text-[10px]" title={`最近向量化：${item.indexed_at || "未知"}`}>已向量化 · {item.vector_count} 段</Badge>;
+    }
+    return <Badge variant="outline" className="ml-auto shrink-0 text-[10px]">未向量化</Badge>;
+  };
+
   const renderWikiNode = (node: WikiTreeNode, depth = 0) => {
     const hasChildren = node.children.length > 0;
     const isExpanded = query.trim() ? true : expandedDirectories.has(node.key);
@@ -432,7 +448,8 @@ export function WeKnoraWikiManagementPage() {
                 >
                   <div className="flex min-w-0 items-center gap-2">
                     <FileText className="size-3.5 shrink-0 text-muted-foreground" />
-                    <span className="truncate font-medium">{page.title || page.slug}</span>
+                     <span className="truncate font-medium">{page.title || page.slug}</span>
+                     {renderVectorBadge(page)}
                   </div>
                   <div className="mt-1 truncate pl-5 text-xs text-muted-foreground">{page.slug}</div>
                 </button>
@@ -460,6 +477,7 @@ export function WeKnoraWikiManagementPage() {
         <div className="flex min-w-0 items-center gap-2">
           <FileText className="size-3.5 shrink-0 text-muted-foreground" />
           <span className="truncate font-medium">{page.title || page.slug}</span>
+          {renderVectorBadge(page)}
         </div>
         <div className="mt-1 truncate pl-5 text-xs text-muted-foreground">{page.slug}</div>
         {page.aliases?.length ? (
@@ -634,6 +652,26 @@ export function WeKnoraWikiManagementPage() {
       <main className="grid min-h-0 flex-1 grid-cols-[320px_1fr] overflow-hidden">
         <aside className="flex min-h-0 flex-col border-r bg-muted/20">
           <div className="border-b p-3">
+            {indexStatus.data && (
+              <div className="mb-3 rounded-lg border bg-background p-3">
+                <div className="flex items-center justify-between gap-2 text-xs">
+                  <span className="flex items-center gap-2 font-medium">
+                    向量化进度
+                    {!indexStatus.data.local_index_enabled && <Badge variant="outline" className="text-[10px]">当前未启用</Badge>}
+                  </span>
+                  <span className="tabular-nums text-muted-foreground">
+                    {indexStatus.data.progress?.processed_pages ?? 0}/{indexStatus.data.progress?.total_pages ?? 0} 篇 · {indexStatus.data.progress?.completion_percent ?? 0}%
+                  </span>
+                </div>
+                <Progress value={indexStatus.data.progress?.completion_percent ?? 0} className="mt-2 h-2" />
+                <div className="mt-2 flex flex-wrap gap-x-3 gap-y-1 text-[11px] text-muted-foreground">
+                  <span>已向量化 {indexStatus.data.progress?.vectorized_pages ?? 0} 篇</span>
+                  <span>向量章节 {indexStatus.data.progress?.vector_count ?? 0}</span>
+                  <span className={(indexStatus.data.progress?.failed_pages ?? 0) > 0 ? "text-destructive" : ""}>失败 {indexStatus.data.progress?.failed_pages ?? 0} 篇</span>
+                </div>
+                {indexStatus.data.last_error && <div className="mt-2 break-words text-[11px] text-destructive">{indexStatus.data.last_error}</div>}
+              </div>
+            )}
             <div className="relative">
               <Search className="absolute left-3 top-1/2 size-4 -translate-y-1/2 text-muted-foreground" />
               <Input

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

@@ -19,6 +19,7 @@ import { Button } from "@/components/ui/button";
 import { Input } from "@/components/ui/input";
 import { MarkdownContent } from "@/components/workspace/messages/markdown-content";
 import {
+  buildAppHashUrl,
   buildPublicKnowledgeWikiPath,
   buildWeKnoraWikiPagePath,
 } from "@/core/llmwiki/source-links";
@@ -35,6 +36,7 @@ import { markdownToDocxBlob } from "@/core/scheduled-tasks/markdown-docx";
 import {
   exportableWikiMarkdown,
   renderableWikiMarkdown,
+  rewriteProtectedWikiImageSources,
   safeWikiFileName,
   stripWikiFrontmatter,
   wikiLinkMarkdown,
@@ -49,6 +51,7 @@ import {
   usePublicLlmWikiWikiPage,
   usePublicLlmWikiWikiPages,
 } from "@/strategy-components/api/llmwiki";
+import { WikiProtectedImage } from "./WikiProtectedImage";
 
 const LIST_PATH = "/page/workspace/knowledge";
 
@@ -96,9 +99,11 @@ function downloadBlob(blob: Blob, fileName: string) {
 }
 
 function publicWikiTargetFromHref(href: string, knowledgeBaseId: string) {
-  const routePrefix = `/#${buildPublicKnowledgeWikiPath(knowledgeBaseId)}`;
-  if (!href.startsWith(`${routePrefix}?`)) return null;
-  const route = href.slice(2);
+  const hashIndex = href.indexOf("#");
+  if (hashIndex < 0) return null;
+  const route = href.slice(hashIndex + 1);
+  const routePrefix = buildPublicKnowledgeWikiPath(knowledgeBaseId);
+  if (!route.startsWith(`${routePrefix}?`)) return null;
   const queryIndex = route.indexOf("?");
   const slug = new URLSearchParams(route.slice(queryIndex + 1)).get("wikiSlug")?.trim();
   return slug ? { route, slug } : null;
@@ -209,7 +214,7 @@ export function WeKnoraWikiPageView({
       : pageData?.content,
   ));
   const displayMarkdown = pageData
-    ? renderableWikiMarkdown(
+    ? rewriteProtectedWikiImageSources(renderableWikiMarkdown(
         title,
         content,
         noteId,
@@ -217,12 +222,12 @@ export function WeKnoraWikiPageView({
           ? (knowledgeBaseId, slug) => {
               const next = new URLSearchParams(searchParams);
               next.set("wikiSlug", slug);
-              return `/#${buildPublicKnowledgeWikiPath(knowledgeBaseId)}?${next.toString()}`;
+              return buildAppHashUrl(`${buildPublicKnowledgeWikiPath(knowledgeBaseId)}?${next.toString()}`);
             }
           : undefined,
         publicEmbed ? pagesList : [],
         publicEmbed ? publicWikiIndex.data?.groups || [] : [],
-      )
+      ))
     : "";
   const exportMarkdown = pageData
     ? wikiLinkMarkdown(exportableWikiMarkdown(title, summary, content), noteId)
@@ -497,6 +502,15 @@ export function WeKnoraWikiPageView({
                 isLoading={false}
                 remarkPlugins={reasoningPlugins.remarkPlugins}
                 rehypePlugins={reasoningPlugins.rehypePlugins}
+                components={{
+                  img: (props) => (
+                    <WikiProtectedImage
+                      {...props}
+                      knowledgeBaseId={noteId}
+                      publicAccess={publicEmbed}
+                    />
+                  ),
+                }}
                 linkClassName={(href) =>
                   publicEmbed && publicWikiTargetFromHref(href, noteId)
                     ? "wiki-content-link border-b border-dashed !border-[var(--primary)] font-medium !text-primary !no-underline hover:border-solid focus-visible:rounded-sm focus-visible:outline-2 focus-visible:outline-offset-2 focus-visible:outline-primary"

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

@@ -0,0 +1,72 @@
+import { ImageOff, Loader2 } from "lucide-react";
+import { type ImgHTMLAttributes, useEffect, useState } from "react";
+
+import { parseProtectedWikiImageSource } from "@/core/llmwiki/wiki-page-markdown";
+import { getLlmWikiProtectedFileUrl } from "@/strategy-components/api/llmwiki";
+
+type WikiProtectedImageProps = ImgHTMLAttributes<HTMLImageElement> & {
+  knowledgeBaseId: string;
+  publicAccess?: boolean;
+};
+
+export function WikiProtectedImage({
+  knowledgeBaseId,
+  publicAccess = false,
+  src,
+  alt,
+  className,
+  ...props
+}: WikiProtectedImageProps) {
+  const filePath = parseProtectedWikiImageSource(src);
+  const [httpUrl, setHttpUrl] = useState("");
+  const [error, setError] = useState("");
+
+  useEffect(() => {
+    if (!filePath || !knowledgeBaseId) {
+      setHttpUrl("");
+      setError("");
+      return;
+    }
+    let cancelled = false;
+    setHttpUrl("");
+    setError("");
+    void getLlmWikiProtectedFileUrl(knowledgeBaseId, filePath, publicAccess)
+      .then((url) => {
+        if (cancelled) return;
+        setHttpUrl(url);
+      })
+      .catch((reason: unknown) => {
+        if (!cancelled) {
+          setError(reason instanceof Error ? reason.message : "Wiki 图片加载失败");
+        }
+      });
+    return () => {
+      cancelled = true;
+    };
+  }, [filePath, knowledgeBaseId, publicAccess]);
+
+  if (!filePath) {
+    return <img {...props} src={src} alt={alt} className={className} />;
+  }
+  if (error) {
+    return (
+      <span
+        className="my-3 flex min-h-24 items-center justify-center gap-2 rounded-lg border border-dashed bg-muted/40 px-4 text-sm text-muted-foreground"
+        data-wiki-image-error={error}
+        role="img"
+        aria-label={alt || "Wiki 图片加载失败"}
+      >
+        <ImageOff className="size-5" />
+        {alt || "Wiki 图片加载失败"}
+      </span>
+    );
+  }
+  if (!httpUrl) {
+    return (
+      <span className="my-3 flex min-h-24 items-center justify-center rounded-lg bg-muted/30 text-muted-foreground">
+        <Loader2 className="size-5 animate-spin" aria-label="正在加载 Wiki 图片" />
+      </span>
+    );
+  }
+  return <img {...props} src={httpUrl} alt={alt} className={className} />;
+}

offline-backend-20260512/backend/CLAUDE.md

@@ -638,13 +638,20 @@ Two `TypeDecorator`s keep ORM columns correct across the `sqlite` / `postgres` /
 
 ### Schema Migrations / Manual Column-Add SQL (`scripts/sql/`)
 
-Alembic auto-migration is **disabled** ? `_run_alembic_upgrade(url)` is commented out in `persistence/engine.py`. On startup only `Base.metadata.create_all` runs, which creates **new tables** but never patches **existing** ones (no new columns, no new indexes). So a column added to an ORM model appears automatically on a brand-new deployment, but is **missing on any existing MySQL database** ? and code reading/writing that column will error there until the column is added by hand.
+Gateway startup runs Alembic through `app.gateway.app._run_db_migrations` after ORM table creation. The migration graph must always have unique revision ids and exactly one head; `tests/test_migration_graph.py` enforces both. Historical duplicate revision `20260821_01` is repaired by the `20260821_03` / `20260821_04` fixed-question chain and merge revision `20260825_01`. Startup DDL has a bounded timeout and failures are logged without preventing the app from serving.
 
 **Convention (required):** whenever code adds a field to an existing table (a new `mapped_column` on an ORM model), you must, in the same change:
 1. Keep the matching alembic version file under `persistence/migrations/versions/` (guarded with `_has_table`/`_has_column`), and
 2. Append the idempotent `ADD COLUMN` statement to **`scripts/sql/column_additions_mysql.sql`**, with a definition matching the ORM column exactly (type, length, nullability).
 
-Operators upgrading an existing DB run the whole file (`USE <db>;` then execute) ? it is idempotent (existing columns are skipped via `information_schema.columns`), so it is safe to re-run. Index-only changes go in the sibling `scripts/sql/leaderboard_indexes_mysql.sql` (same idempotent pattern).
+Operators upgrading an existing DB may run the whole file (`USE <db>;` then execute) as a recovery path; it is idempotent (existing columns are skipped via `information_schema.columns`). Index-only changes go in the sibling `scripts/sql/leaderboard_indexes_mysql.sql` (same idempotent pattern) and stay out of startup migrations because online index creation on large metric tables can delay service boot.
+
+Leaderboard scheduled-run exclusion uses correlated `NOT EXISTS` against the
+indexed `scheduled_task_runs.agent_run_id`; do not restore `NOT IN`. Legacy
+scheduler detection for tool metrics is also a correlated existence check and
+must not outer-join the complete `runs` table into metric aggregations. Keep the
+created-at + dimension compound indexes in ORM metadata and the idempotent
+MySQL index script synchronized.
 
 ### Sandbox System (`packages/harness/deerflow/sandbox/`)
 
@@ -1018,6 +1025,11 @@ Both can be modified at runtime via Gateway API endpoints or `DeerFlowClient` me
 `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 ordinary bases. The conversation-deposit KB is system-owned, always published, shared by all DeerFlow users, excluded from My Knowledge, and immutable through user/admin detail operations; recognition requires both the reserved name and `system` owner. Existing same-name remote mappings are adopted in place as the system deposit, preserving mapping ids and document links. Its user-facing description, generated Markdown source and tags use `cmzs` branding; list/detail reads synchronize stale remote descriptions, and search results, source previews and embedded-detail JSON recursively replace legacy brand text in historical conversation-deposit payloads. Deposit requests must reference a thread accessible to the current account. Personal knowledge-base list/search views are scoped to the current DeerFlow account even for admins; frontend LLMWiki caches include the live account id, and admin write power does not make other users' private KBs appear as owned. Only `system_role === "admin"` sees the WeKnora connection-config tab. 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 sequential multi-file upload with an aggregate result for writable ordinary bases, hides it for public read-only/system-deposit bases, and uses a generic loading label while the embed session starts. It also supports 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. WeKnora iframe detail pages first call `/api/llmwiki/knowledge-bases/{id}/weknora/session` through the normal DeerFlow API auth path, then load `/platform/knowledge-bases/{weknora_id}` with a signed `deerflow_weknora_embed` cookie for `/platform`, `/assets`, `/locales`, root assets, and API calls. Injected fetch/XHR rewriting sends all iframe `/api/v1/*`, including colliding `/api/v1/auth/*`, through `/api/llmwiki/weknora-embed/api/v1/*`; Gateway auth/CSRF bypass only applies to those signed iframe proxy requests, the proxy still revalidates the mapped KB, and cookie security honors `X-Forwarded-Proto` for HTTPS reverse proxies. `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. Source previews prefer the cleaned Wiki/LLMWiki page text when a retrieved chunk maps to a Wiki page, with raw split chunks as the fallback. 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`.
 
 The embedded detail proxy hides WeKnora's outer `.main > .aside_box` application sidebar and expands the route outlet, but deliberately keeps the Wiki page's internal directory/index sidebar. Do not add an HTML `sandbox` attribute to this trusted, permission-filtered detail iframe: a sandboxed ancestor causes Chromium's built-in Blob PDF viewer to render the browser-blocked page. The Gateway follows upstream redirects only for `/knowledge/{id}/preview` and `/knowledge-bases/{id}/files`, keeping PDF bytes and protected Wiki images inside the signed proxy instead of sending the browser to WeKnora object storage.
+Native/private and public Wiki readers must rewrite protected Markdown image
+sources to their mapping-scoped DeerFlow `/files?file_path=` endpoints and load
+them as Blob URLs. The embedded guard also rewrites protected fetch/XHR values
+and hydrates late `img[data-protected-src]` placeholders; never expose the
+service token or relax the current-KB authorization to make images work.
 
 **Embedded native knowledge-base switching:** The original WeKnora breadcrumb picker remains available inside the detail iframe. A signed 15-minute embed cookie carries only the current DeerFlow user's visible mappings (their own plus published bases); the proxied `/api/v1/knowledge-bases` response is filtered to that allowlist. Choosing another entry performs a full iframe navigation, rechecks the DeerFlow mapping and read/write permission, then reissues the cookie for the selected base. Never expose the upstream service-admin list directly or permit an arbitrary remote knowledge-base ID.
 
@@ -1444,8 +1456,10 @@ detail. Never call `/knowledge-search` from internal search, RAG, or the
 `deerflow.integrations.weknora.local_index` and
 `deerflow.persistence.llmwiki_index`; preserve the harness → app import
 firewall. Gateway composition injects the store/search/sync services through
-`local_index.runtime`, and the lease-protected scheduler starts only when both
-`enabled` and `sync_enabled` are true.
+`local_index.runtime`. `enabled` turns on local vector search and manual
+vectorization; `auto_sync` is the separate opt-in for the lease-protected
+scheduler. The scheduler sleeps for one full configured interval before its
+first scan, so application startup must never trigger immediate vectorization.
 
 Vectors are normalized little-endian float32 BLOBs. A page mirror and all of
 its vectors are replaced in one transaction only after strict batched
@@ -1458,6 +1472,10 @@ input, not an outbound API field: fixed-dimension BGE/OpenAI-compatible
 endpoints can reject a `dimensions` request parameter with HTTP 400. When the
 value is omitted, the embedding client probes and freezes the first response's
 dimension before sync lease/rebuild decisions.
+Embedding errors must retain the upstream status and bounded response body in
+logs, without exposing API keys. SQL vector writes are split into statements
+of at most `database_batch_size` rows (default 50) inside the same atomic page
+replacement transaction.
 During a fingerprint rebuild the mapping uses the Wiki-page API until every
 page is successfully rebuilt. `WikiRetrievalService` partitions mixed requests
 per mapping, so ready libraries stay local while incomplete libraries use that
@@ -1485,6 +1503,16 @@ Wiki directory, reconstruct hierarchy from `parent_slug`, `wiki_path`,
 directory navigation on the same lightweight route rather than mounting the
 WeKnora iframe.
 
+Generated HashRouter URLs must retain `window.location.pathname` (for example
+the `/web/` reverse-proxy prefix); never emit root-absolute `/#/...` citation,
+reader, or public-embed links. Protected Wiki image rendering must use real HTTP
+URLs, never `URL.createObjectURL`. Public readers use the published file proxy;
+authenticated readers obtain a short-lived signed capability from
+`/files/access-url` and load `/api/public/llmwiki/files/{token}`. The signed
+payload is restricted to `kind=wiki_file`, mapping id, remote id, validated
+protected path, and expiry, and responses provide an inline UTF-8 filename for
+new-tab viewing and browser Save As.
+
 Third-party pages embed published Wiki content through the chrome-free route
 `/#/embed/knowledge/{mapping_id}` (optional exact-page `wikiSlug`). It defaults
 to a recognized Wiki index page (title `索引`/`index`, page type `index`, or an
@@ -1517,7 +1545,11 @@ mapping id is hard to guess.
 
 Owners manually force a full sync/vectorization through `POST
 /api/llmwiki/knowledge-bases/{id}/vectorize`; status is available through `GET
-/api/llmwiki/knowledge-bases/{id}/wiki-index-status`. The Wiki management page
+/api/llmwiki/knowledge-bases/{id}/wiki-index-status`. That response includes
+aggregate `progress` and per-page `pages` details; current-scan progress is
+derived from `active_sync_id` / `last_seen_sync_id`, while each page exposes its
+index state, vector count, timestamps, and bounded error. The Wiki management
+page must show the progress bar and a status badge beside every article. It
 also exports all remote processed Wiki pages via `GET
 /api/llmwiki/knowledge-bases/{id}/wiki/export.xlsx`. Directory export must keep
 the full path, current directory, parent path, and dynamic per-level columns;

offline-backend-20260512/backend/README.md

@@ -779,15 +779,24 @@ processed Wiki content. A knowledge base uses exact NumPy cosine similarity
 only after its whole Wiki has completed one coherent vectorization pass; until
 then internal search calls WeKnora's Wiki-page search/list API and reads the
 matched Wiki pages. It never falls back to raw documents or chunks. The local
-vector feature is disabled by default. Configure
-`llmwiki.local_wiki_index.embedding` first, then enable `enabled`,
-`sync_enabled`, and `internal_search_enabled` in `config.yaml`.
+vector feature is disabled by default. Configure the three
+`llmwiki.local_wiki_index.embedding` values (`model`, `base_url`, and
+`api_key`), then change only `enabled` to `true` in `config.yaml`. Manual
+vectorization is the default. Set `auto_sync: true` only when periodic
+background synchronization is wanted; the scheduler waits one full interval
+after startup before its first scan, so boot does not trigger a rebuild.
+`database_batch_size` caps each vector insert statement (default `50`) while
+the page replacement remains atomic, preventing a large Wiki from flooding
+the SQL connection in one statement.
 The optional `embedding.dimensions` value is used only to validate the returned
 vector size and build the index fingerprint; it is deliberately not sent in
 the embedding request because fixed-dimension BGE/OpenAI-compatible endpoints
 may reject that extra field with HTTP 400. Omit it to discover the dimension
 from the first successful response.
 
+Embedding failures log the upstream HTTP status and a bounded response body.
+The client never sends a `dimensions` request parameter.
+
 The migration creates `llmwiki_wiki_pages`, `llmwiki_wiki_vectors`, and
 `llmwiki_wiki_sync_states`. Wiki Markdown and normalized float32 vectors are
 stored in DeerFlow's configured SQL database. Multi-worker sync is protected by
@@ -795,8 +804,12 @@ database leases; searches use revision-checked per-knowledge-base LRU snapshots
 after the full-library readiness gate succeeds. Knowledge-base owners can click
 **手动向量化** in Wiki management (`POST
 /api/llmwiki/knowledge-bases/{id}/vectorize`) and follow its status at
-`GET /api/llmwiki/knowledge-bases/{id}/wiki-index-status`. Existing chunk
-citations remain readable while all new retrieval citations open Wiki pages.
+`GET /api/llmwiki/knowledge-bases/{id}/wiki-index-status`. The response includes
+aggregate `progress` plus a `pages` list with every Wiki page's index state,
+vector-section count, timestamps, and last error. The management page renders
+the percentage and labels each article as 已向量化 / 待向量化 / 失败, so it is
+clear exactly what is searchable. Existing chunk citations remain readable
+while all new retrieval citations open Wiki pages.
 
 The frontend labels these results simply as **Wiki** references; provider-brand
 wording is intentionally hidden from user-facing labels and status text. Clicking any
@@ -819,6 +832,29 @@ left. It reconstructs hierarchy from `parent_slug`, `wiki_path`,
 page, and switches directory entries through the same lightweight reader route
 without mounting the WeKnora iframe.
 
+All generated SPA deep links preserve the deployment document path (for
+example `/web/`) before the `#` route, including Q&A references and Wiki index
+links. Protected Wiki images no longer render from temporary browser `blob:`
+URLs. Public images use the published file proxy directly; private images first
+obtain a 24-hour signed capability and then load through
+`GET /api/public/llmwiki/files/{token}` with a real HTTP URL and inline
+filename, allowing zoom, new-tab viewing, and browser Save As without exposing
+the user's bearer token.
+
+## Leaderboard query and schema startup safety
+
+The leaderboard excludes scheduled runs with indexed correlated `NOT EXISTS`
+checks. Tool metrics no longer outer-join the full `runs` table merely to
+recognize legacy scheduler messages. New databases receive dimension-aware
+analytics indexes from ORM metadata; existing large MySQL deployments should
+run `scripts/sql/leaderboard_indexes_mysql.sql` during a low-traffic window so
+index creation never delays application startup.
+
+Startup Alembic migrations now have one unambiguous head. The historical
+duplicate `20260821_01` identifier is repaired by unique fixed-question
+revisions and the `20260825_01` merge revision, which also fills the
+report-structure column for databases stamped by either old variant.
+
 Published knowledge bases also have a login-free, chrome-free reader at
 `/#/embed/knowledge/{mapping_id}` for direct third-party iframe use. With no
 `wikiSlug`, it selects the Wiki **索引** page first (recognized by title, page
@@ -841,6 +877,13 @@ system conversation-deposit knowledge base. Private or unpublished bases are
 never exposed by this route. Anonymous consumers can discover the available
 public Wiki libraries through `GET /api/public/llmwiki/knowledge-bases`; each
 entry deliberately contains only the DeerFlow mapping `id` and display `name`.
+Protected `resource://` and supported object-storage images are fetched through
+the current mapping instead of being exposed as upstream URLs. Authenticated
+readers use `GET /api/llmwiki/knowledge-bases/{id}/files?file_path=...`; public
+readers use the corresponding `/api/public/llmwiki/...` route, which requires a
+published mapping. The embedded WeKnora guard applies the same mapping-scoped
+proxy to document previews and Markdown images, including images inserted after
+the drawer/page first renders.
 
 Wiki management also provides **导出 Excel**. `GET
 /api/llmwiki/knowledge-bases/{id}/wiki/export.xlsx` exports every processed Wiki

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

@@ -998,7 +998,7 @@ async def lifespan(app: FastAPI) -> AsyncGenerator[None, None]:
         local_wiki_index = app.state.config.llmwiki.local_wiki_index
         if (
             local_wiki_index.enabled
-            and local_wiki_index.sync_enabled
+            and local_wiki_index.auto_sync
             and getattr(app.state, "llmwiki_sync_service", None) is not None
         ):
             from app.gateway.llmwiki_index_scheduler import local_wiki_index_scheduler_loop

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

@@ -321,13 +321,16 @@ async def langgraph_runtime(app: FastAPI) -> AsyncGenerator[None, None]:
         from deerflow.integrations.weknora.local_index.runtime import register_local_wiki_index_runtime
         from deerflow.persistence.llmwiki_index import make_llmwiki_index_store
 
-        app.state.llmwiki_index_store = make_llmwiki_index_store(sf)
+        local_index_config = config.llmwiki.local_wiki_index
+        app.state.llmwiki_index_store = make_llmwiki_index_store(
+            sf,
+            write_batch_size=local_index_config.database_batch_size,
+        )
         app.state.llmwiki_embedding = None
         app.state.llmwiki_vector_cache = None
         app.state.llmwiki_vector_search = None
         app.state.llmwiki_sync_service = None
         app.state.llmwiki_retrieval_service = None
-        local_index_config = config.llmwiki.local_wiki_index
         from deerflow.integrations.weknora.runtime import build_weknora_client, get_resolved_llmwiki_runtime
 
         resolved_llmwiki = get_resolved_llmwiki_runtime(config)

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

@@ -48,6 +48,10 @@ async def _sync_once(app: FastAPI) -> None:
 
 async def local_wiki_index_scheduler_loop(app: FastAPI) -> None:
     interval = app.state.config.llmwiki.local_wiki_index.sync_interval_seconds
+    logger.info(
+        "Local Wiki auto-sync scheduler started; first scan delayed by %s seconds",
+        interval,
+    )
     while True:
-        await _sync_once(app)
         await asyncio.sleep(interval)
+        await _sync_once(app)

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

@@ -637,10 +637,16 @@ def _non_scheduled_run_filter():
 
 
 def _non_scheduled_tool_filter():
-    return or_(
-        ToolCallMetricRow.run_id.is_(None),
-        RunRow.run_id.is_(None),
-        _non_scheduled_run_filter(),
+    """Exclude legacy scheduled runs without joining runs into the metric scan."""
+
+    return ~(
+        select(1)
+        .select_from(RunRow)
+        .where(
+            RunRow.run_id == ToolCallMetricRow.run_id,
+            RunRow.first_human_message.like("【定时任务触发】%"),
+        )
+        .exists()
     )
 
 
@@ -667,8 +673,15 @@ def _admin_user_ids_stmt():
     return select(UserRow.id).where(UserRow.system_role == "admin")
 
 
-def _scheduled_run_ids_stmt():
-    return select(ScheduledTaskRunRow.agent_run_id).where(ScheduledTaskRunRow.agent_run_id.is_not(None))
+def _not_scheduled_run_id_filter(run_id_column):
+    """Use a NULL-safe, index-friendly anti-semi join instead of NOT IN."""
+
+    return ~(
+        select(1)
+        .select_from(ScheduledTaskRunRow)
+        .where(ScheduledTaskRunRow.agent_run_id == run_id_column)
+        .exists()
+    )
 
 
 def _apply_user_scope(stmt, column, *, exclude_user_ids: set[str], include_admins: bool):
@@ -692,7 +705,7 @@ def _apply_run_scope(
     if not include_failed:
         stmt = stmt.where(RunRow.status != "error")
     if not include_scheduled:
-        stmt = stmt.where(or_(run_id_column.is_(None), run_id_column.not_in(_scheduled_run_ids_stmt())))
+        stmt = stmt.where(_not_scheduled_run_id_filter(run_id_column))
         stmt = stmt.where(_non_scheduled_run_filter())
     return stmt
 
@@ -706,8 +719,8 @@ def _apply_tool_scope(
 ):
     stmt = _apply_user_scope(stmt, ToolCallMetricRow.user_id, exclude_user_ids=exclude_user_ids, include_admins=include_admins)
     if not include_scheduled:
-        stmt = stmt.where(or_(ToolCallMetricRow.run_id.is_(None), ToolCallMetricRow.run_id.not_in(_scheduled_run_ids_stmt())))
-        stmt = stmt.outerjoin(RunRow, RunRow.run_id == ToolCallMetricRow.run_id).where(_non_scheduled_tool_filter())
+        stmt = stmt.where(_not_scheduled_run_id_filter(ToolCallMetricRow.run_id))
+        stmt = stmt.where(_non_scheduled_tool_filter())
     return stmt
 
 
@@ -985,7 +998,7 @@ async def _query_model_rankings(
     if not include_failed:
         stmt = stmt.where(LlmCallMetricRow.status != "error")
     if not include_scheduled:
-        stmt = stmt.where(or_(LlmCallMetricRow.run_id.is_(None), LlmCallMetricRow.run_id.not_in(_scheduled_run_ids_stmt())))
+        stmt = stmt.where(_not_scheduled_run_id_filter(LlmCallMetricRow.run_id))
     rows = (await _execute_leaderboard_query(session, "model_rankings", stmt)).all()
     return [
         AdminLeaderboardDimension(
@@ -1080,8 +1093,8 @@ async def _query_overview(
     if not include_failed:
         run_where.append(RunRow.status != "error")
     if not include_scheduled:
-        run_where.append(or_(RunRow.run_id.is_(None), RunRow.run_id.not_in(_scheduled_run_ids_stmt())))
-        tool_where.append(or_(ToolCallMetricRow.run_id.is_(None), ToolCallMetricRow.run_id.not_in(_scheduled_run_ids_stmt())))
+        run_where.append(_not_scheduled_run_id_filter(RunRow.run_id))
+        tool_where.append(_not_scheduled_run_id_filter(ToolCallMetricRow.run_id))
         tool_where.append(_non_scheduled_tool_filter())
 
     total_users_stmt = select(func.count(UserRow.id))
@@ -1104,7 +1117,6 @@ async def _query_overview(
             func.count(ToolCallMetricRow.id),
             func.sum(case((func.nullif(ToolCallMetricRow.skill_name, "").is_not(None), 1), else_=0)),
         )
-        .outerjoin(RunRow, RunRow.run_id == ToolCallMetricRow.run_id)
         .where(*tool_where)
     )
     tool_row = (await _execute_leaderboard_query(session, "overview.tool_summary", tool_summary_stmt)).one()

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

@@ -6,6 +6,7 @@ import asyncio
 import hashlib
 import json
 import logging
+import mimetypes
 import re
 import time
 from dataclasses import replace
@@ -47,6 +48,9 @@ from app.gateway.weknora_embed import (
 from app.gateway.weknora_embed import (
     verify_weknora_embed_cookie as _verify_embed_cookie,
 )
+from app.gateway.weknora_embed import (
+    verify_weknora_signed_payload as _verify_signed_payload,
+)
 from deerflow.config.agents_config import load_agent_config
 from deerflow.config.llmwiki_config import LlmWikiRuntimeOverride, normalize_service_url
 from deerflow.config.system_settings import load_system_settings, save_system_settings
@@ -61,6 +65,11 @@ router = APIRouter(tags=["llmwiki"])
 proxy_router = APIRouter(tags=["llmwiki-proxy"])
 
 _MAX_UPLOAD_BYTES = 100 * 1024 * 1024
+_WIKI_FILE_ACCESS_TTL_SECONDS = 24 * 60 * 60
+_PROTECTED_WIKI_FILE_PATH = re.compile(
+    r"^(?:storage://[0-9A-Za-z_-]+/)?(?:resource|local|minio|cos|tos|s3|oss|ks3|obs)://",
+    re.IGNORECASE,
+)
 _WEKNORA_ADMIN_SESSION_CACHE: dict[str, dict[str, Any]] = {}
 _WEKNORA_PUBLIC_PREFIX = "/deerflow"
 _WEKNORA_PUBLIC_ROOT_PATHS = (
@@ -242,7 +251,7 @@ def _store(request: Request) -> LlmWikiStore:
 def _schedule_local_wiki_sync(request: Request, mapping: dict[str, Any]) -> None:
     service = getattr(request.app.state, "llmwiki_sync_service", None)
     config = request.app.state.config.llmwiki.local_wiki_index
-    if service is not None and config.enabled and mapping.get("wiki_index_enabled", True):
+    if service is not None and config.enabled and config.auto_sync and mapping.get("wiki_index_enabled", True):
         from app.gateway.llmwiki_index_scheduler import spawn_local_wiki_index_job
 
         spawn_local_wiki_index_job(request.app, service.sync_mapping(mapping, force=True))
@@ -1085,6 +1094,8 @@ html.deerflow-source-detail-open [data-deerflow-source-detail-panel="true"] [cla
 
   function rewriteApiUrl(value) {{
     try {{
+      const protectedUrl = protectedImageApiUrl(value);
+      if (protectedUrl) return protectedUrl;
       const next = new URL(value, location.origin);
       if (next.origin !== location.origin) return value;
       const publicPrefix = cfg.publicPrefix || "";
@@ -1104,6 +1115,19 @@ html.deerflow-source-detail-open [data-deerflow-source-detail-panel="true"] [cla
       return value;
     }}
   }}
+  function protectedImageApiUrl(value) {{
+    const raw = String(value || "").trim();
+    if (!/^(?:storage:\\/\\/[0-9A-Za-z_-]+\\/)?(?:resource|local|minio|cos|tos|s3|oss|ks3|obs):\\/\\//i.test(raw)) {{
+      return "";
+    }}
+    const endpoint =
+      cfg.apiPrefix +
+      "/api/v1/knowledge-bases/" +
+      encodeURIComponent(cfg.weknoraId) +
+      "/files?file_path=" +
+      encodeURIComponent(raw);
+    return new URL(endpoint, location.origin).href;
+  }}
   const nativeFetch = window.fetch.bind(window);
   window.fetch = (input, init) => {{
     if (input instanceof Request) {{
@@ -1116,6 +1140,67 @@ html.deerflow-source-detail-open [data-deerflow-source-detail-panel="true"] [cla
     return nativeXhrOpen.call(this, method, rewriteApiUrl(url), ...rest);
   }};
 
+  // WeKnora renders protected Markdown images as transparent placeholders and
+  // hydrates them asynchronously.  In an embedded page that hydration can run
+  // before its own API client has picked up our proxy prefix, leaving a 1x1
+  // GIF forever.  Resolve the same protected source through the scoped proxy
+  // as a fallback.  The backend validates the scheme and current KB again.
+  function hydrateProtectedImage(image) {{
+    const protectedSrc = String(image?.getAttribute?.("data-protected-src") || "").trim();
+    const proxyUrl = protectedImageApiUrl(protectedSrc);
+    if (!proxyUrl) return;
+    if (
+      image.dataset.deerflowProtectedSource === protectedSrc &&
+      String(image.getAttribute("src") || "").startsWith("blob:")
+    ) return;
+    const attempts = Number(image.dataset.deerflowProtectedAttempts || "0");
+    if (attempts >= 3 || image.dataset.deerflowProtectedLoading === "1") return;
+    image.dataset.deerflowProtectedLoading = "1";
+    image.dataset.deerflowProtectedAttempts = String(attempts + 1);
+    nativeFetch(proxyUrl, {{ credentials: "same-origin", headers: {{ Accept: "image/*" }} }})
+      .then((response) => {{
+        if (!response.ok) throw new Error(`protected image request failed (${{response.status}})`);
+        return response.blob();
+      }})
+      .then((blob) => {{
+        const objectUrl = URL.createObjectURL(blob);
+        image.dataset.deerflowProtectedSource = protectedSrc;
+        image.removeAttribute("data-img-loading");
+        image.removeAttribute("data-wiki-image-error");
+        image.src = objectUrl;
+        image.addEventListener("load", () => URL.revokeObjectURL(objectUrl), {{ once: true }});
+      }})
+      .catch(() => {{
+        image.dataset.wikiImageError = "true";
+        window.setTimeout(() => {{
+          delete image.dataset.deerflowProtectedLoading;
+          hydrateProtectedImage(image);
+        }}, 1200 * (attempts + 1));
+      }})
+      .finally(() => {{
+        delete image.dataset.deerflowProtectedLoading;
+      }});
+  }}
+  function hydrateProtectedImages(scope = document) {{
+    if (scope?.matches?.("img[data-protected-src]")) hydrateProtectedImage(scope);
+    scope?.querySelectorAll?.("img[data-protected-src]").forEach(hydrateProtectedImage);
+  }}
+  const protectedImageObserver = new MutationObserver((mutations) => {{
+    mutations.forEach((mutation) => {{
+      if (mutation.type === "attributes") hydrateProtectedImage(mutation.target);
+      mutation.addedNodes?.forEach((node) => {{
+        if (node?.nodeType === Node.ELEMENT_NODE) hydrateProtectedImages(node);
+      }});
+    }});
+  }});
+  protectedImageObserver.observe(document.documentElement, {{
+    childList: true,
+    subtree: true,
+    attributes: true,
+    attributeFilter: ["data-protected-src"]
+  }});
+  hydrateProtectedImages();
+
   function tone(color, alpha) {{
     return `color-mix(in srgb, ${{color}} ${{Math.round(alpha * 100)}}%, transparent)`;
   }}
@@ -1519,6 +1604,7 @@ html.deerflow-source-detail-open [data-deerflow-source-detail-panel="true"] [cla
     enforceLocation();
     purgeGuides();
     hideWeKnoraShell();
+    hydrateProtectedImages();
     if (tickSourceDetail() && sourceDetailMode) {{
       clearInterval(maintenanceTimer);
       try {{ readyObserver.disconnect(); }} catch (_) {{}}
@@ -1820,6 +1906,36 @@ async def _public_mapping(request: Request, mapping_id: str) -> dict[str, Any]:
     return row
 
 
+def _validate_protected_wiki_file_path(file_path: str) -> str:
+    value = file_path.strip()
+    if not value or "\x00" in value or not _PROTECTED_WIKI_FILE_PATH.match(value):
+        raise HTTPException(status_code=422, detail="Unsupported Wiki image path")
+    return value
+
+
+async def _wiki_file_response(row: dict[str, Any], file_path: str, *, public: bool) -> Response:
+    safe_file_path = _validate_protected_wiki_file_path(file_path)
+    try:
+        content, media_type = await _client_or_503().get_knowledge_base_file(
+            str(row["weknora_id"]),
+            safe_file_path,
+        )
+    except WeKnoraError as exc:
+        _raise_upstream(exc)
+    filename = unquote(safe_file_path.rstrip("/").rsplit("/", 1)[-1]).replace("\r", "").replace("\n", "") or "wiki-file"
+    if "." not in filename:
+        filename += mimetypes.guess_extension(str(media_type).split(";", 1)[0].strip()) or ""
+    return Response(
+        content=content,
+        media_type=media_type,
+        headers={
+            "Cache-Control": "public, max-age=300" if public else "private, max-age=3600",
+            "Content-Disposition": f"inline; filename*=UTF-8''{quote(filename, safe='')}",
+            "X-Content-Type-Options": "nosniff",
+        },
+    )
+
+
 async def _verify_document(client: WeKnoraClient, row: dict[str, Any], document_id: str) -> dict[str, Any]:
     document = await client.get_document(document_id)
     if str(document.get("knowledge_base_id") or "") != str(row["weknora_id"]):
@@ -2510,6 +2626,37 @@ async def list_public_wiki_pages(
     return {**result, "pages": pages}
 
 
+@router.get("/api/public/llmwiki/knowledge-bases/{mapping_id}/files")
+async def get_public_wiki_file(
+    request: Request,
+    mapping_id: str,
+    file_path: str = Query(..., min_length=1, max_length=8192),
+) -> Response:
+    """Serve one protected image belonging to a published Wiki."""
+
+    row = await _public_mapping(request, mapping_id)
+    return await _wiki_file_response(row, file_path, public=True)
+
+
+@router.get("/api/public/llmwiki/files/{token}")
+async def get_signed_wiki_file(token: str) -> Response:
+    """Serve a private Wiki image through a short-lived, signed HTTP URL."""
+
+    payload = _verify_signed_payload(token)
+    if payload.get("kind") != "wiki_file":
+        raise HTTPException(status_code=401, detail="Signed Wiki file token is invalid")
+    mapping_id = str(payload.get("mapping_id") or "")
+    weknora_id = str(payload.get("weknora_id") or "")
+    file_path = str(payload.get("file_path") or "")
+    if not mapping_id or not weknora_id or not file_path:
+        raise HTTPException(status_code=401, detail="Signed Wiki file token is invalid")
+    return await _wiki_file_response(
+        {"id": mapping_id, "weknora_id": weknora_id},
+        file_path,
+        public=False,
+    )
+
+
 _PUBLIC_WIKI_INDEX_TYPES = ("summary", "entity", "concept", "synthesis", "comparison")
 
 
@@ -2820,6 +2967,40 @@ async def preview_document(request: Request, mapping_id: str, document_id: str)
     )
 
 
+@router.get("/api/llmwiki/knowledge-bases/{mapping_id}/files")
+async def get_wiki_file(
+    request: Request,
+    mapping_id: str,
+    file_path: str = Query(..., min_length=1, max_length=8192),
+) -> Response:
+    """Serve one protected Wiki image after normal knowledge-base authorization."""
+
+    row, _, _ = await _authorized_mapping(request, mapping_id, write=False)
+    return await _wiki_file_response(row, file_path, public=False)
+
+
+@router.get("/api/llmwiki/knowledge-bases/{mapping_id}/files/access-url")
+async def create_wiki_file_access(
+    request: Request,
+    mapping_id: str,
+    file_path: str = Query(..., min_length=1, max_length=8192),
+) -> dict[str, Any]:
+    """Issue a browser-usable HTTP capability after normal KB authorization."""
+
+    row, _, _ = await _authorized_mapping(request, mapping_id, write=False)
+    safe_file_path = _validate_protected_wiki_file_path(file_path)
+    token = _sign_embed_payload(
+        {
+            "kind": "wiki_file",
+            "mapping_id": mapping_id,
+            "weknora_id": str(row["weknora_id"]),
+            "file_path": safe_file_path,
+            "exp": int(time.time()) + _WIKI_FILE_ACCESS_TTL_SECONDS,
+        }
+    )
+    return {"token": token, "expires_in": _WIKI_FILE_ACCESS_TTL_SECONDS}
+
+
 @router.get("/api/llmwiki/knowledge-bases/{mapping_id}/graph")
 async def get_knowledge_base_graph(
     request: Request,

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

@@ -2,6 +2,7 @@
 
 from __future__ import annotations
 
+import logging
 from typing import Any
 from uuid import uuid4
 
@@ -14,6 +15,7 @@ from deerflow.integrations.weknora.local_index.retrieval import is_fully_vectori
 from deerflow.integrations.weknora.local_index.search import WikiIndexError
 
 router = APIRouter(prefix="/api/llmwiki", tags=["llmwiki-index"])
+logger = logging.getLogger(__name__)
 
 
 class WikiVectorSearchRequest(BaseModel):
@@ -181,6 +183,7 @@ async def _run_sync(app, mapping_ids: list[str], force: bool) -> None:
             try:
                 await service.sync_mapping(mapping, force=force)
             except Exception:
+                logger.exception("Manual Wiki vectorization failed mapping_id=%s", mapping.get("id"))
                 continue
 
 
@@ -191,6 +194,7 @@ async def _run_one_sync(app, mapping: dict[str, Any], *, force: bool) -> None:
     try:
         await service.sync_mapping(mapping, force=force)
     except Exception:
+        logger.exception("Manual Wiki vectorization failed mapping_id=%s", mapping.get("id"))
         return
 
 
@@ -241,6 +245,33 @@ async def knowledge_base_wiki_index_status(request: Request, mapping_id: str) ->
         raise HTTPException(status_code=503, detail={"code": "WIKI_INDEX_UNAVAILABLE"})
     states = await index_store.list_index_status([mapping_id])
     state = states[0] if states else {"knowledge_base_mapping_id": mapping_id, "state": "not_synced", "index_revision": 0}
+    if (
+        state.get("state") == "idle"
+        and int(state.get("index_revision") or 0) == 0
+        and not state.get("last_completed_at")
+        and not state.get("last_success_at")
+    ):
+        state = {**state, "state": "not_synced"}
+    pages = await index_store.list_page_index_status(mapping_id)
+    active_sync_id = str(state.get("active_sync_id") or "")
+    is_running = str(state.get("state") or "") in {"running", "rebuilding"}
+    for page in pages:
+        page["vectorized"] = page.get("index_status") == "ready" and int(page.get("vector_count") or 0) > 0
+        page["processed_in_current_sync"] = bool(active_sync_id) and page.get("last_seen_sync_id") == active_sync_id
+    total_pages = max(int(state.get("remote_page_count") or 0), len(pages))
+    processed_pages = (
+        sum(bool(page["processed_in_current_sync"]) for page in pages)
+        if is_running
+        else total_pages if str(state.get("state") or "") in {"idle", "partial"} else 0
+    )
+    completion_percent = (
+        min(100, round((processed_pages / total_pages) * 100))
+        if total_pages
+        else 100
+        if str(state.get("state") or "") in {"idle", "partial"}
+        and bool(state.get("last_completed_at") or state.get("last_success_at"))
+        else 0
+    )
     embedding = getattr(request.app.state, "llmwiki_embedding", None)
     fingerprint = embedding.fingerprint if embedding is not None else None
     fully_vectorized = bool(mapping.get("wiki_index_enabled", True)) and is_fully_vectorized(state, fingerprint)
@@ -251,6 +282,15 @@ async def knowledge_base_wiki_index_status(request: Request, mapping_id: str) ->
         "wiki_index_enabled": mapping.get("wiki_index_enabled", True),
         "fully_vectorized": fully_vectorized,
         "retrieval_mode": "vector" if fully_vectorized else "wiki_api",
+        "progress": {
+            "total_pages": total_pages,
+            "processed_pages": processed_pages,
+            "vectorized_pages": sum(bool(page["vectorized"]) for page in pages),
+            "failed_pages": sum(bool(page.get("last_error")) or page.get("index_status") in {"failed", "stale"} for page in pages),
+            "vector_count": sum(int(page.get("vector_count") or 0) for page in pages),
+            "completion_percent": completion_percent,
+        },
+        "pages": pages,
     }
 
 

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

@@ -106,6 +106,26 @@ def sign_weknora_embed_payload(payload: dict[str, Any]) -> str:
     return f"{body}.{_b64url_encode(sig)}"
 
 
+def verify_weknora_signed_payload(raw: str | None) -> dict[str, Any]:
+    """Verify a short-lived capability signed with the iframe proxy secret."""
+
+    if not raw or "." not in raw:
+        raise HTTPException(status_code=401, detail="Signed resource token is missing")
+    body, sig = raw.rsplit(".", 1)
+    expected = _b64url_encode(hmac.new(WEKNORA_PROXY_SECRET.encode("utf-8"), body.encode("ascii"), hashlib.sha256).digest())
+    if not hmac.compare_digest(sig, expected):
+        raise HTTPException(status_code=401, detail="Signed resource token is invalid")
+    try:
+        payload = json.loads(_b64url_decode(body))
+    except (ValueError, TypeError, json.JSONDecodeError):
+        raise HTTPException(status_code=401, detail="Signed resource token is invalid") from None
+    if not isinstance(payload, dict):
+        raise HTTPException(status_code=401, detail="Signed resource token is invalid")
+    if int(payload.get("exp") or 0) < int(time.time()):
+        raise HTTPException(status_code=401, detail="Signed resource token has expired")
+    return payload
+
+
 def verify_weknora_embed_cookie(raw: str | None) -> WeKnoraEmbedContext:
     if not raw or "." not in raw:
         raise HTTPException(status_code=401, detail="WeKnora iframe session is missing")

offline-backend-20260512/backend/config.yaml

@@ -556,47 +556,17 @@ llmwiki:
     web_base_url: "http://119.254.86.251:8088"
     admin_email: "lqq@qq.com"
     admin_password: "1qaz@WSX"
-  # DeerFlow-local Wiki mirror/vector index. Keep disabled until the
-  # OpenAI-compatible embedding endpoint below is configured and reachable.
+  # 本地 Wiki 向量化：只需把 enabled 改为 true，并填写下面 3 个 embedding 值。
+  # 默认只在页面点击“向量化”时执行；auto_sync=true 才会后台自动增量同步，
+  # 且服务启动时只启动定时器，不会立刻全量向量化。
   local_wiki_index:
     enabled: false
-    sync_enabled: false
-    internal_search_enabled: false
-    reference_ui_enabled: false
-    sync_interval_seconds: 120
-    chunk_max_chars: 1000
-    chunk_overlap_chars: 120
-    chunk_min_chars: 80
+    auto_sync: false
+    database_batch_size: 50
     embedding:
-      enabled: true
-      provider: openai
       model: bge-m3
       base_url: ""
       api_key: ""
-      # Expected BGE-M3 response size for local validation/indexing only.
-      # This value is not sent as an embedding API request parameter.
-      dimensions: 1024
-      batch_size: 32
-      timeout_seconds: 30
-      max_retries: 3
-    engine: numpy_exact
-    strict_vector: true
-    top_k_sections: 30
-    top_k_pages: 8
-    max_sections_per_page: 2
-    min_similarity: 0.35
-    cache_max_knowledge_bases: 20
-    cache_revision_check_seconds: 5
-    external_api:
-      enabled: false
-      api_key: ""
-      previous_api_key: ""
-      max_query_chars: 1000
-      max_knowledge_bases: 20
-      max_results: 50
-      default_results: 10
-      include_content_max_chars: 12000
-      requests_per_minute: 60
 
 log_system:
   enabled: true

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

@@ -279,65 +279,25 @@ llmwiki:
 
   local_wiki_index:
     enabled: true
-
-    # 同步
-    sync_enabled: true
-    sync_interval_seconds: 120
-    sync_page_size: 100
-    sync_concurrency: 3
-    sync_lease_seconds: 600
-    sync_request_timeout_seconds: 30
-    sync_max_retries: 3
-    sync_retry_backoff_seconds: 5
-
-    # 切分
-    chunk_max_chars: 1000
-    chunk_overlap_chars: 120
-    chunk_min_chars: 80
-    title_prefix_enabled: true
-
-    # Embedding：OpenAI-compatible /v1/embeddings
+    # 默认只允许手动向量化；需要定时增量同步时再开启。
+    auto_sync: false
+    # 每条 SQL 最多插入的向量行数；整页替换仍在同一事务内完成。
+    database_batch_size: 50
     embedding:
-      enabled: true
-      provider: openai
       model: bge-m3
       base_url: "$WIKI_EMBEDDING_BASE_URL"
       api_key: "$WIKI_EMBEDDING_API_KEY"
-      dimensions: 1024
-      batch_size: 32
-      timeout_seconds: 30
-      max_retries: 3
-
-    # 检索
-    engine: numpy_exact
-    strict_vector: true
-    top_k_sections: 30
-    top_k_pages: 8
-    max_sections_per_page: 2
-    min_similarity: 0.35
-    keyword_boost_enabled: false
-    keyword_boost_weight: 0.15
-    cache_max_knowledge_bases: 20
-    cache_revision_check_seconds: 5
-
-    # 外部接口
-    external_api:
-      enabled: false
-      api_key: "$LLMWIKI_EXTERNAL_SEARCH_API_KEY"
-      max_query_chars: 1000
-      max_knowledge_bases: 20
-      max_results: 50
-      default_results: 10
-      include_content_max_chars: 12000
-      requests_per_minute: 60
 ```
 
+日常部署只需配置上面三个 Embedding 值并将 `enabled` 改为 `true`。
+`auto_sync=false` 时仅由页面“手动向量化”触发；即使设为 `true`，定时器也会先等待一个完整同步周期，不会在服务刚启动时立即重建索引。其余切分、检索、超时与重试参数继续使用代码默认值，只有确有调优需要时才覆盖。
+
 ### 6.1 配置校验规则
 
 启动时应执行：
 
 1. `enabled=true` 时，Embedding `base_url` 和 `model` 必须非空；
-2. `dimensions` 可省略，但第一次成功 Embedding 后必须固定并写入索引 profile；
+2. 不向上游发送 `dimensions` 参数；第一次成功 Embedding 后从响应自动发现维度，并固定到索引 profile；
 3. `chunk_overlap_chars < chunk_max_chars`；
 4. `top_k_pages <= top_k_sections`；
 5. `external_api.enabled=true` 时必须配置非空 API Key；
@@ -1885,9 +1845,7 @@ class WikiVectorEngine(ABC):
 
 ```text
 local_wiki_index.enabled
-local_wiki_index.sync_enabled
-local_wiki_index.internal_search_enabled
-local_wiki_index.reference_ui_enabled
+local_wiki_index.auto_sync
 local_wiki_index.external_api.enabled
 ```
 

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

@@ -59,7 +59,7 @@ class LocalWikiEmbeddingConfig(BaseModel):
     base_url: str = ""
     api_key: str = Field(default="", repr=False)
     dimensions: int | None = Field(
-        default=1024,
+        default=None,
         ge=1,
         le=65536,
         description="Expected response vector dimensions; not sent to the embedding endpoint.",
@@ -108,9 +108,7 @@ class LocalWikiIndexConfig(BaseModel):
     """DeerFlow-owned mirror and exact-vector index for WeKnora Wiki pages."""
 
     enabled: bool = False
-    sync_enabled: bool = False
-    internal_search_enabled: bool = False
-    reference_ui_enabled: bool = False
+    auto_sync: bool = False
     sync_interval_seconds: int = Field(default=120, ge=10, le=86400)
     sync_page_size: int = Field(default=100, ge=1, le=500)
     sync_concurrency: int = Field(default=3, ge=1, le=32)
@@ -133,9 +131,31 @@ class LocalWikiIndexConfig(BaseModel):
     keyword_boost_weight: float = Field(default=0.15, ge=0, le=1)
     cache_max_knowledge_bases: int = Field(default=20, ge=1, le=1000)
     cache_revision_check_seconds: float = Field(default=5, ge=0, le=3600)
+    database_batch_size: int = Field(default=50, ge=1, le=500)
     external_api: LocalWikiExternalApiConfig = Field(default_factory=LocalWikiExternalApiConfig)
     model_config = ConfigDict(extra="ignore")
 
+    @model_validator(mode="before")
+    @classmethod
+    def _migrate_legacy_flags(cls, value):
+        if isinstance(value, dict) and "auto_sync" not in value and "sync_enabled" in value:
+            value = {**value, "auto_sync": value.get("sync_enabled")}
+        return value
+
+    @property
+    def sync_enabled(self) -> bool:
+        """Backward-compatible status name; automatic work is one explicit switch."""
+
+        return self.auto_sync
+
+    @property
+    def internal_search_enabled(self) -> bool:
+        return self.enabled
+
+    @property
+    def reference_ui_enabled(self) -> bool:
+        return self.enabled
+
     @model_validator(mode="after")
     def _validate_index(self):
         if self.chunk_overlap_chars >= self.chunk_max_chars:

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

@@ -2,8 +2,8 @@
 
 from __future__ import annotations
 
-from collections.abc import Mapping, Sequence
 import time
+from collections.abc import Mapping, Sequence
 from typing import Any
 from urllib.parse import quote
 
@@ -430,6 +430,51 @@ class WeKnoraClient:
         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 get_knowledge_base_file(
+        self,
+        knowledge_base_id: str,
+        file_path: str,
+    ) -> tuple[bytes, str]:
+        """Return one protected Wiki attachment using DeerFlow's WeKnora identity."""
+
+        owns_client = self._http_client is None
+        client = self._http_client or httpx.AsyncClient(
+            timeout=httpx.Timeout(self.timeout_seconds),
+            follow_redirects=True,
+            trust_env=False,
+        )
+        try:
+            for attempt in range(2):
+                response = await client.request(
+                    "GET",
+                    f"{self.base_url}/api/v1/knowledge-bases/{quote(knowledge_base_id, safe='')}/files",
+                    params={"file_path": file_path},
+                    headers=await self._headers(force_refresh=attempt > 0, accept="*/*"),
+                    follow_redirects=True,
+                )
+                if response.status_code not in {401, 403} or not self._has_admin_credentials or attempt > 0:
+                    break
+        except httpx.TimeoutException as exc:
+            raise WeKnoraError("WeKnora Wiki image request timed out", status_code=504) from exc
+        except httpx.HTTPError as exc:
+            raise WeKnoraError("WeKnora Wiki image is unavailable", status_code=502) from exc
+        finally:
+            if owns_client:
+                await client.aclose()
+        if response.status_code == 404:
+            raise WeKnoraError("WeKnora Wiki image was not found", status_code=404)
+        if response.status_code in {401, 403}:
+            if self._has_admin_credentials:
+                raise WeKnoraError("WeKnora rejected DeerFlow's authenticated Wiki image request", status_code=502)
+            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 {}

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

@@ -15,6 +15,21 @@ from deerflow.integrations.weknora.local_index.vector_codec import normalize_vec
 logger = logging.getLogger(__name__)
 
 
+def _upstream_error_detail(response: httpx.Response) -> str:
+    """Return a bounded upstream reason without hiding the useful response body."""
+
+    try:
+        payload = response.json()
+        if isinstance(payload, dict):
+            detail = payload.get("detail") or payload.get("message") or payload.get("error") or payload
+        else:
+            detail = payload
+        value = str(detail)
+    except (ValueError, TypeError):
+        value = response.text
+    return " ".join(value.split())[:2000] or "no response body"
+
+
 class WikiEmbeddingError(RuntimeError):
     def __init__(self, message: str, *, category: str = "request", retryable: bool = True) -> None:
         super().__init__(message)
@@ -76,11 +91,15 @@ class StrictWikiEmbeddingClient:
             started = monotonic()
             try:
                 response = await (await self._get_client()).post(url, headers=headers, json=payload)
-                if response.status_code in {401, 403}:
-                    raise WikiEmbeddingError("Embedding authentication failed", category="authentication", retryable=False)
-                if response.status_code == 429 or response.status_code >= 500:
-                    raise WikiEmbeddingError(f"Embedding service returned HTTP {response.status_code}", category="upstream", retryable=True)
-                response.raise_for_status()
+                if response.status_code >= 400:
+                    detail = _upstream_error_detail(response)
+                    category = "authentication" if response.status_code in {401, 403} else "upstream"
+                    retryable = response.status_code == 429 or response.status_code >= 500
+                    raise WikiEmbeddingError(
+                        f"Embedding service returned HTTP {response.status_code}: {detail}",
+                        category=category,
+                        retryable=retryable,
+                    )
                 data = response.json()
                 items = sorted(data.get("data", []), key=lambda item: int(item.get("index", 0))) if isinstance(data, dict) else []
                 if len(items) != len(texts):
@@ -98,8 +117,20 @@ class StrictWikiEmbeddingClient:
             except httpx.TimeoutException:
                 last_error = WikiEmbeddingError("Embedding request timed out", category="timeout", retryable=True)
             except (httpx.HTTPError, ValueError, TypeError) as exc:
-                last_error = WikiEmbeddingError(f"Embedding request failed ({type(exc).__name__})", category="request", retryable=isinstance(exc, httpx.HTTPError))
+                reason = " ".join(str(exc).split())[:2000] or type(exc).__name__
+                last_error = WikiEmbeddingError(
+                    f"Embedding request failed ({type(exc).__name__}): {reason}",
+                    category="request",
+                    retryable=isinstance(exc, httpx.HTTPError),
+                )
             if last_error is not None and (not last_error.retryable or attempt >= self.config.max_retries):
+                logger.error(
+                    "Wiki embedding batch failed model=%s batch=%d category=%s reason=%s",
+                    self.config.model,
+                    len(texts),
+                    last_error.category,
+                    last_error,
+                )
                 raise last_error
             await asyncio.sleep(min(30.0, 0.5 * (2**attempt)))
         raise last_error or WikiEmbeddingError("Embedding request failed")

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

@@ -90,6 +90,13 @@ class WikiSyncService:
                 batch = await self.client.list_wiki_pages(remote_id, page=page_number, page_size=self.config.sync_page_size)
                 pages = [page for page in batch.get("pages", []) if isinstance(page, dict)]
                 remote_count += len(pages)
+                reported_total = int(batch.get("total") or batch.get("total_count") or remote_count)
+                await self.store.update_sync_progress(
+                    mapping_id,
+                    owner=self.owner,
+                    sync_id=sync_id,
+                    remote_page_count=max(remote_count, reported_total),
+                )
                 for summary in pages:
                     try:
                         _ensure_lease()

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

@@ -18,6 +18,7 @@ class LlmCallMetricRow(Base):
     __table_args__ = (
         Index("ix_llm_call_metrics_created_user", "created_at", "user_id"),
         Index("ix_llm_call_metrics_run_created", "run_id", "created_at"),
+        Index("ix_llm_call_metrics_created_model_user", "created_at", "model_name", "user_id"),
     )
 
     id: Mapped[str] = mapped_column(String(64), primary_key=True)

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

@@ -6,8 +6,12 @@ from deerflow.persistence.llmwiki_index.model import LlmWikiWikiPageRow, LlmWiki
 from deerflow.persistence.llmwiki_index.sql import SqlLlmWikiIndexStore
 
 
-def make_llmwiki_index_store(session_factory):
-    return MemoryLlmWikiIndexStore() if session_factory is None else SqlLlmWikiIndexStore(session_factory)
+def make_llmwiki_index_store(session_factory, *, write_batch_size: int = 50):
+    return (
+        MemoryLlmWikiIndexStore()
+        if session_factory is None
+        else SqlLlmWikiIndexStore(session_factory, write_batch_size=write_batch_size)
+    )
 
 
 __all__ = [

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

@@ -44,11 +44,24 @@ class LlmWikiIndexStore(ABC):
     @abstractmethod
     async def renew_sync_lease(self, mapping_id: str, *, owner: str, lease_seconds: int) -> bool: ...
 
+    @abstractmethod
+    async def update_sync_progress(
+        self,
+        mapping_id: str,
+        *,
+        owner: str,
+        sync_id: str,
+        remote_page_count: int,
+    ) -> bool: ...
+
     @abstractmethod
     async def finish_sync(self, mapping_id: str, *, owner: str, sync_id: str, success: bool, remote_page_count: int, error: str | None = None) -> None: ...
 
     @abstractmethod
     async def list_index_status(self, mapping_ids: list[str] | None = None) -> list[dict[str, Any]]: ...
 
+    @abstractmethod
+    async def list_page_index_status(self, mapping_id: str) -> list[dict[str, Any]]: ...
+
     @abstractmethod
     async def mark_rebuild_required(self, mapping_ids: list[str], fingerprint: str) -> int: ...

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

@@ -123,6 +123,21 @@ class MemoryLlmWikiIndexStore(LlmWikiIndexStore):
         state["lease_expires_at"] = _now() + timedelta(seconds=lease_seconds)
         return True
 
+    async def update_sync_progress(
+        self,
+        mapping_id: str,
+        *,
+        owner: str,
+        sync_id: str,
+        remote_page_count: int,
+    ) -> bool:
+        state = self._state(mapping_id)
+        if state.get("lease_owner") != owner or state.get("active_sync_id") != sync_id:
+            return False
+        state["remote_page_count"] = max(0, remote_page_count)
+        state["updated_at"] = _now()
+        return True
+
     async def finish_sync(self, mapping_id: str, *, owner: str, sync_id: str, success: bool, remote_page_count: int, error: str | None = None) -> None:
         state = self._state(mapping_id)
         if state.get("lease_owner") != owner or state.get("active_sync_id") != sync_id:
@@ -150,6 +165,33 @@ class MemoryLlmWikiIndexStore(LlmWikiIndexStore):
         wanted = set(mapping_ids or [])
         return [dict(value) for key, value in self.states.items() if not wanted or key in wanted]
 
+    async def list_page_index_status(self, mapping_id: str) -> list[dict[str, Any]]:
+        result = []
+        for page in sorted(
+            (
+                page
+                for page in self.pages.values()
+                if page["knowledge_base_mapping_id"] == mapping_id and not page.get("is_remote_deleted")
+            ),
+            key=lambda page: (str(page.get("wiki_path") or ""), str(page.get("slug") or "")),
+        ):
+            result.append(
+                {
+                    "id": page["id"],
+                    "slug": page["slug"],
+                    "title": page["title"],
+                    "page_type": page["page_type"],
+                    "status": page["status"],
+                    "index_status": page["index_status"],
+                    "vector_count": len(self.vectors.get(page["id"], [])),
+                    "indexed_at": page.get("indexed_at"),
+                    "last_error": page.get("last_error"),
+                    "last_seen_sync_id": page.get("last_seen_sync_id"),
+                    "updated_at": page.get("updated_at"),
+                }
+            )
+        return result
+
     async def mark_rebuild_required(self, mapping_ids: list[str], fingerprint: str) -> int:
         for mapping_id in mapping_ids:
             state = self._state(mapping_id)

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

@@ -6,7 +6,7 @@ from datetime import UTC, datetime, timedelta
 from typing import Any
 from uuid import uuid4
 
-from sqlalchemy import case, delete, func, or_, select, update
+from sqlalchemy import case, delete, func, insert, or_, select, update
 from sqlalchemy.exc import IntegrityError
 from sqlalchemy.ext.asyncio import AsyncSession, async_sessionmaker
 
@@ -27,8 +27,9 @@ def _dict(row) -> dict[str, Any]:
 
 
 class SqlLlmWikiIndexStore(LlmWikiIndexStore):
-    def __init__(self, session_factory: async_sessionmaker[AsyncSession]) -> None:
+    def __init__(self, session_factory: async_sessionmaker[AsyncSession], *, write_batch_size: int = 50) -> None:
         self._sf = session_factory
+        self.write_batch_size = max(1, min(int(write_batch_size), 500))
 
     async def _ensure_state(self, session: AsyncSession, mapping_id: str) -> LlmWikiWikiSyncStateRow:
         row = await session.get(LlmWikiWikiSyncStateRow, mapping_id)
@@ -82,24 +83,29 @@ class SqlLlmWikiIndexStore(LlmWikiIndexStore):
             row.updated_at = now
             await session.flush()
             await session.execute(delete(LlmWikiWikiVectorRow).where(LlmWikiWikiVectorRow.page_id == row.id))
-            session.add_all(
-                [
-                    LlmWikiWikiVectorRow(
-                        id=str(uuid4()),
-                        page_id=row.id,
-                        knowledge_base_mapping_id=mapping_id,
-                        section_index=int(item["section_index"]),
-                        heading=item.get("heading"),
-                        section_content=str(item["section_content"]),
-                        content_hash=str(item["content_hash"]),
-                        embedding_fingerprint=fingerprint,
-                        embedding_dimensions=int(item["embedding_dimensions"]),
-                        vector_blob=bytes(item["vector_blob"]),
-                        created_at=now,
-                    )
-                    for item in vectors
-                ]
-            )
+            vector_rows = [
+                {
+                    "id": str(uuid4()),
+                    "page_id": row.id,
+                    "knowledge_base_mapping_id": mapping_id,
+                    "section_index": int(item["section_index"]),
+                    "heading": item.get("heading"),
+                    "section_content": str(item["section_content"]),
+                    "content_hash": str(item["content_hash"]),
+                    "embedding_fingerprint": fingerprint,
+                    "embedding_dimensions": int(item["embedding_dimensions"]),
+                    "vector_blob": bytes(item["vector_blob"]),
+                    "created_at": now,
+                }
+                for item in vectors
+            ]
+            # Keep the page replacement atomic, but cap each SQL statement so
+            # large Wiki pages cannot flood the database with one huge insert.
+            for start in range(0, len(vector_rows), self.write_batch_size):
+                await session.execute(
+                    insert(LlmWikiWikiVectorRow),
+                    vector_rows[start : start + self.write_batch_size],
+                )
             state = await self._ensure_state(session, mapping_id)
             state.index_revision = int(state.index_revision or 0) + 1
             state.embedding_fingerprint = fingerprint
@@ -229,6 +235,27 @@ class SqlLlmWikiIndexStore(LlmWikiIndexStore):
             await session.commit()
             return bool(result.rowcount)
 
+    async def update_sync_progress(
+        self,
+        mapping_id: str,
+        *,
+        owner: str,
+        sync_id: str,
+        remote_page_count: int,
+    ) -> bool:
+        async with self._sf() as session:
+            result = await session.execute(
+                update(LlmWikiWikiSyncStateRow)
+                .where(
+                    LlmWikiWikiSyncStateRow.knowledge_base_mapping_id == mapping_id,
+                    LlmWikiWikiSyncStateRow.lease_owner == owner,
+                    LlmWikiWikiSyncStateRow.active_sync_id == sync_id,
+                )
+                .values(remote_page_count=max(0, remote_page_count), updated_at=_now())
+            )
+            await session.commit()
+            return bool(result.rowcount)
+
     async def finish_sync(self, mapping_id: str, *, owner: str, sync_id: str, success: bool, remote_page_count: int, error: str | None = None) -> None:
         async with self._sf() as session:
             state = await session.get(LlmWikiWikiSyncStateRow, mapping_id)
@@ -280,6 +307,47 @@ class SqlLlmWikiIndexStore(LlmWikiIndexStore):
         async with self._sf() as session:
             return [_dict(row) for row in (await session.execute(stmt)).scalars()]
 
+    async def list_page_index_status(self, mapping_id: str) -> list[dict[str, Any]]:
+        vector_counts = (
+            select(
+                LlmWikiWikiVectorRow.page_id.label("page_id"),
+                func.count(LlmWikiWikiVectorRow.id).label("vector_count"),
+            )
+            .where(LlmWikiWikiVectorRow.knowledge_base_mapping_id == mapping_id)
+            .group_by(LlmWikiWikiVectorRow.page_id)
+            .subquery()
+        )
+        stmt = (
+            select(LlmWikiWikiPageRow, func.coalesce(vector_counts.c.vector_count, 0))
+            .outerjoin(vector_counts, vector_counts.c.page_id == LlmWikiWikiPageRow.id)
+            .where(
+                LlmWikiWikiPageRow.knowledge_base_mapping_id == mapping_id,
+                LlmWikiWikiPageRow.is_remote_deleted.is_(False),
+            )
+            .order_by(LlmWikiWikiPageRow.wiki_path, LlmWikiWikiPageRow.slug)
+        )
+        async with self._sf() as session:
+            rows = (await session.execute(stmt)).all()
+        result: list[dict[str, Any]] = []
+        for row, vector_count in rows:
+            page = _dict(row)
+            result.append(
+                {
+                    "id": page["id"],
+                    "slug": page["slug"],
+                    "title": page["title"],
+                    "page_type": page["page_type"],
+                    "status": page["status"],
+                    "index_status": page["index_status"],
+                    "vector_count": int(vector_count or 0),
+                    "indexed_at": page.get("indexed_at"),
+                    "last_error": page.get("last_error"),
+                    "last_seen_sync_id": page.get("last_seen_sync_id"),
+                    "updated_at": page.get("updated_at"),
+                }
+            )
+        return result
+
     async def mark_rebuild_required(self, mapping_ids: list[str], fingerprint: str) -> int:
         if not mapping_ids:
             return 0

offline-backend-20260512/backend/packages/harness/deerflow/persistence/migrations/versions/20260821_01_fixed_questions.py

@@ -1,7 +1,7 @@
 """Create administrator-managed exact-match fixed questions.
 
-Revision ID: 20260821_01
-Revises: 20260816_01
+Revision ID: 20260821_03
+Revises: 20260821_01
 Create Date: 2026-08-21
 """
 
@@ -11,8 +11,8 @@ import sqlalchemy as sa
 from alembic import op
 from sqlalchemy.dialects import mysql
 
-revision = "20260821_01"
-down_revision = "20260816_01"
+revision = "20260821_03"
+down_revision = "20260821_01"
 branch_labels = None
 depends_on = None
 

offline-backend-20260512/backend/packages/harness/deerflow/persistence/migrations/versions/20260821_02_fixed_question_stream_speed.py

@@ -1,7 +1,7 @@
 """Add configurable fixed-answer streaming speed.
 
-Revision ID: 20260821_02
-Revises: 20260821_01
+Revision ID: 20260821_04
+Revises: 20260821_03
 Create Date: 2026-08-21
 """
 
@@ -10,8 +10,8 @@ from __future__ import annotations
 import sqlalchemy as sa
 from alembic import op
 
-revision = "20260821_02"
-down_revision = "20260821_01"
+revision = "20260821_04"
+down_revision = "20260821_03"
 branch_labels = None
 depends_on = None
 

offline-backend-20260512/backend/packages/harness/deerflow/persistence/migrations/versions/20260823_01_llmwiki_local_wiki_index.py

@@ -1,7 +1,7 @@
 """Add DeerFlow-local WeKnora Wiki mirror and vector index.
 
 Revision ID: 20260823_01
-Revises: 20260821_02
+Revises: 20260821_04
 Create Date: 2026-08-23
 """
 
@@ -12,7 +12,7 @@ from alembic import op
 from sqlalchemy.dialects import mysql
 
 revision = "20260823_01"
-down_revision = "20260821_02"
+down_revision = "20260821_04"
 branch_labels = None
 depends_on = None
 

offline-backend-20260512/backend/packages/harness/deerflow/persistence/migrations/versions/20260825_01_merge_report_structure_and_llmwiki_heads.py

@@ -0,0 +1,34 @@
+"""Merge report-structure and local-Wiki heads, repairing the old duplicate stamp.
+
+Revision ID: 20260825_01
+Revises: 20260822_02, 20260823_01
+Create Date: 2026-08-25
+"""
+
+from __future__ import annotations
+
+import sqlalchemy as sa
+from alembic import op
+
+revision = "20260825_01"
+down_revision = ("20260822_02", "20260823_01")
+branch_labels = None
+depends_on = None
+
+
+def upgrade() -> None:
+    # An older build accidentally used 20260821_01 for two migrations. A DB
+    # stamped by the fixed-question variant may therefore still lack this
+    # report column even though Alembic considers the surviving 01 applied.
+    inspector = sa.inspect(op.get_bind())
+    if inspector.has_table("report_structures"):
+        columns = {column["name"] for column in inspector.get_columns("report_structures")}
+        if "structure_mode" not in columns:
+            op.add_column(
+                "report_structures",
+                sa.Column("structure_mode", sa.String(length=16), nullable=False, server_default="adaptive"),
+            )
+
+
+def downgrade() -> None:
+    pass

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

@@ -52,4 +52,6 @@ class RunRow(Base):
         Index("ix_runs_created_user_status", "created_at", "user_id", "status"),
         Index("ix_runs_user_created", "user_id", "created_at"),
         Index("ix_runs_thread_created", "thread_id", "created_at"),
+        Index("ix_runs_created_assistant_user", "created_at", "assistant_id", "user_id"),
+        Index("ix_runs_created_model_user", "created_at", "model_name", "user_id"),
     )

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

@@ -24,6 +24,8 @@ class ToolCallMetricRow(Base):
     __table_args__ = (
         Index("ix_tool_call_metrics_created_user", "created_at", "user_id"),
         Index("ix_tool_call_metrics_run_created", "run_id", "created_at"),
+        Index("ix_tool_call_metrics_created_skill_user", "created_at", "skill_name", "user_id"),
+        Index("ix_tool_call_metrics_created_tool_user", "created_at", "tool_name", "user_id"),
     )
 
     id: Mapped[str] = mapped_column(String(64), primary_key=True)

offline-backend-20260512/backend/scripts/sql/leaderboard_indexes_mysql.sql

@@ -4,9 +4,9 @@
 --
 -- 背景:
 --   这些索引声明在各 ORM model 的 __table_args__ 里,但 Base.metadata.create_all
---   对【已存在的表】不会补建索引,而 alembic 自动迁移在本项目里是关闭的
---   (engine.py 中 `# await _run_alembic_upgrade(url)` 被注释)。
---   因此在【已有数据的 MySQL 部署】上,这些索引很可能并未真正创建,排行榜的
+--   对【已存在的表】不会补建索引。本项目虽会在启动时运行 Alembic，但大表
+--   补索引刻意不放进启动迁移，避免在线建索引拖长或阻塞服务启动。
+--   因此在【已有数据的 MySQL 部署】上需要低峰期执行本脚本，否则排行榜的
 --   单日聚合查询会退化为全表扫描,拖慢并占用连接。
 --
 -- 用法:
@@ -55,19 +55,28 @@ DELIMITER ;
 CALL _create_index_if_missing('runs', 'ix_runs_created_user_status', '`created_at`, `user_id`, `status`');
 CALL _create_index_if_missing('runs', 'ix_runs_user_created',        '`user_id`, `created_at`');
 CALL _create_index_if_missing('runs', 'ix_runs_thread_created',      '`thread_id`, `created_at`');
+CALL _create_index_if_missing('runs', 'ix_runs_created_assistant_user', '`created_at`, `assistant_id`, `user_id`');
+CALL _create_index_if_missing('runs', 'ix_runs_created_model_user',     '`created_at`, `model_name`, `user_id`');
 
 -- ---- tool_call_metrics -------------------------------------------------------
 CALL _create_index_if_missing('tool_call_metrics', 'ix_tool_call_metrics_created_user', '`created_at`, `user_id`');
 CALL _create_index_if_missing('tool_call_metrics', 'ix_tool_call_metrics_run_created',  '`run_id`, `created_at`');
+CALL _create_index_if_missing('tool_call_metrics', 'ix_tool_call_metrics_created_skill_user', '`created_at`, `skill_name`, `user_id`');
+CALL _create_index_if_missing('tool_call_metrics', 'ix_tool_call_metrics_created_tool_user',  '`created_at`, `tool_name`, `user_id`');
 
 -- ---- llm_call_metrics --------------------------------------------------------
 CALL _create_index_if_missing('llm_call_metrics', 'ix_llm_call_metrics_created_user', '`created_at`, `user_id`');
 CALL _create_index_if_missing('llm_call_metrics', 'ix_llm_call_metrics_run_created',  '`run_id`, `created_at`');
+CALL _create_index_if_missing('llm_call_metrics', 'ix_llm_call_metrics_created_model_user', '`created_at`, `model_name`, `user_id`');
 
 -- ---- threads_meta ------------------------------------------------------------
 CALL _create_index_if_missing('threads_meta', 'ix_threads_meta_created_user', '`created_at`, `user_id`');
 CALL _create_index_if_missing('threads_meta', 'ix_threads_meta_user_created', '`user_id`, `created_at`');
 
+-- ---- scheduled_task_runs -----------------------------------------------------
+-- NOT EXISTS anti-join used when the leaderboard excludes scheduler runs.
+CALL _create_index_if_missing('scheduled_task_runs', 'ix_scheduled_task_runs_agent_run_id', '`agent_run_id`');
+
 -- ---- position_roundtable_sessions --------------------------------------------
 -- Explicit taskId history is loaded by taskId and newest update time.
 CALL _create_index_if_missing('position_roundtable_sessions', 'ix_position_roundtable_sessions_task_updated', '`external_task_id`, `updated_at`');
@@ -75,15 +84,19 @@ CALL _create_index_if_missing('position_roundtable_sessions', 'ix_position_round
 -- ---- 清理临时存储过程 --------------------------------------------------------
 DROP PROCEDURE IF EXISTS _create_index_if_missing;
 
--- ---- 执行后自检:确认 9 个索引都已存在 ---------------------------------------
+-- ---- 执行后自检:确认排行榜索引都已存在 ----------------------------------------
 SELECT table_name, index_name
   FROM information_schema.statistics
  WHERE table_schema = DATABASE()
    AND index_name IN (
        'ix_runs_created_user_status', 'ix_runs_user_created', 'ix_runs_thread_created',
+       'ix_runs_created_assistant_user', 'ix_runs_created_model_user',
        'ix_tool_call_metrics_created_user', 'ix_tool_call_metrics_run_created',
+       'ix_tool_call_metrics_created_skill_user', 'ix_tool_call_metrics_created_tool_user',
        'ix_llm_call_metrics_created_user', 'ix_llm_call_metrics_run_created',
-       'ix_threads_meta_created_user', 'ix_threads_meta_user_created'
+       'ix_llm_call_metrics_created_model_user',
+       'ix_threads_meta_created_user', 'ix_threads_meta_user_created',
+       'ix_scheduled_task_runs_agent_run_id'
    )
  GROUP BY table_name, index_name
  ORDER BY table_name, index_name;

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

@@ -0,0 +1,43 @@
+from __future__ import annotations
+
+from sqlalchemy import select
+from sqlalchemy.dialects import mysql
+
+from app.gateway.routers.admin_users import _apply_run_scope, _apply_tool_scope
+from deerflow.persistence.run.model import RunRow
+from deerflow.persistence.tool_metrics.model import ToolCallMetricRow
+
+
+def _mysql_sql(stmt) -> str:
+    return str(stmt.compile(dialect=mysql.dialect(), compile_kwargs={"literal_binds": True})).upper()
+
+
+def test_excluding_scheduled_runs_uses_index_friendly_not_exists() -> None:
+    stmt = _apply_run_scope(
+        select(RunRow.run_id),
+        user_column=RunRow.user_id,
+        run_id_column=RunRow.run_id,
+        exclude_user_ids=set(),
+        include_admins=True,
+        include_failed=True,
+        include_scheduled=False,
+    )
+    sql = _mysql_sql(stmt)
+
+    assert "NOT (EXISTS" in sql
+    assert "SCHEDULED_TASK_RUNS" in sql
+    assert " NOT IN " not in sql
+
+
+def test_tool_aggregation_does_not_join_the_full_runs_table() -> None:
+    stmt = _apply_tool_scope(
+        select(ToolCallMetricRow.tool_name),
+        exclude_user_ids=set(),
+        include_admins=True,
+        include_scheduled=False,
+    )
+    sql = _mysql_sql(stmt)
+
+    assert sql.count("NOT (EXISTS") == 2
+    assert "LEFT OUTER JOIN" not in sql
+    assert " NOT IN " not in sql

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

@@ -40,21 +40,48 @@ async def test_sql_store_atomically_replaces_vectors_and_marks_missing_deleted(t
         )
         await session.commit()
 
-    store = SqlLlmWikiIndexStore(session_factory)
+    store = SqlLlmWikiIndexStore(session_factory, write_batch_size=1)
     assert await store.try_acquire_sync_lease("mapping-1", owner="worker", sync_id="sync-1", lease_seconds=60, fingerprint="fp")
     assert not await store.try_acquire_sync_lease("mapping-1", owner="worker", sync_id="sync-2", lease_seconds=60, fingerprint="fp")
     page = normalize_wiki_page("mapping-1", {"id": "p1", "slug": "people/alice", "title": "Alice", "content": "Alice founded Acme.", "status": "published"})
     blob, dimensions = encode_vector([3.0, 4.0], dimensions=2)
     saved = await store.replace_page_vectors(
         page=page,
-        vectors=[{"section_index": 0, "heading": "History", "section_content": "Alice founded Acme.", "content_hash": "section", "embedding_dimensions": dimensions, "vector_blob": blob}],
+        vectors=[
+            {
+                "section_index": index,
+                "heading": "History",
+                "section_content": f"Alice founded Acme. Part {index}",
+                "content_hash": f"section-{index}",
+                "embedding_dimensions": dimensions,
+                "vector_blob": blob,
+            }
+            for index in range(3)
+        ],
         fingerprint="fp",
         sync_id="sync-1",
     )
     snapshot = await store.load_vector_snapshot("mapping-1", "fp")
     assert saved["slug"] == "people/alice"
-    assert len(snapshot["rows"]) == 1
+    assert store.write_batch_size == 1
+    assert len(snapshot["rows"]) == 3
     assert decode_vector(snapshot["rows"][0]["vector_blob"], 2).tolist() == pytest.approx([0.6, 0.8])
+    page_statuses = await store.list_page_index_status("mapping-1")
+    assert len(page_statuses) == 1
+    assert page_statuses[0] == {
+        **page_statuses[0],
+        "id": saved["id"],
+        "slug": "people/alice",
+        "title": "Alice",
+        "page_type": "article",
+        "status": "published",
+        "index_status": "ready",
+        "vector_count": 3,
+        "last_error": None,
+        "last_seen_sync_id": "sync-1",
+    }
+    assert page_statuses[0]["indexed_at"]
+    assert page_statuses[0]["updated_at"]
 
     assert await store.mark_missing_pages_deleted("mapping-1", "sync-2") == 1
     assert (await store.load_vector_snapshot("mapping-1", "fp"))["rows"] == []

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

@@ -690,6 +690,24 @@ async def test_weknora_client_proxies_original_document_preview_without_credenti
     assert media_type == "text/plain"
 
 
+@pytest.mark.asyncio
+async def test_weknora_client_fetches_protected_wiki_image() -> None:
+    async def handler(request: httpx.Request) -> httpx.Response:
+        assert request.url.path == "/api/v1/knowledge-bases/kb-one/files"
+        assert request.url.params["file_path"] == "resource://wiki/figure.png"
+        return httpx.Response(200, content=b"png-bytes", headers={"Content-Type": "image/png"})
+
+    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_knowledge_base_file(
+            "kb-one",
+            "resource://wiki/figure.png",
+        )
+
+    assert content == b"png-bytes"
+    assert media_type == "image/png"
+
+
 @pytest.mark.asyncio
 async def test_memory_store_enforces_owner_and_public_visibility() -> None:
     store = MemoryLlmWikiStore()
@@ -1358,6 +1376,10 @@ def test_weknora_iframe_html_rewrites_public_prefix_paths() -> None:
     assert '"allowedWeKnoraIds":["remote-kb","other-visible-kb"]' in injected
     assert ".breadcrumb-link.dropdown" not in injected
     assert "reloadForKnowledgeBaseSwitch" in injected
+    assert "function protectedImageApiUrl" in injected
+    assert 'img[data-protected-src]' in injected
+    assert 'encodeURIComponent(cfg.weknoraId)' in injected
+    assert 'credentials: "same-origin"' in injected
 
 
 def test_weknora_iframe_filters_kb_list_to_signed_visible_mappings() -> None:

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

@@ -1,10 +1,13 @@
 from __future__ import annotations
 
+import asyncio
 import json
+from types import SimpleNamespace
 
 import httpx
 import pytest
 
+from app.gateway import llmwiki_index_scheduler
 from deerflow.config.llmwiki_config import LocalWikiEmbeddingConfig, LocalWikiIndexConfig
 from deerflow.integrations.weknora.local_index.embedding import StrictWikiEmbeddingClient
 from deerflow.integrations.weknora.local_index.sync import WikiSyncService
@@ -51,7 +54,11 @@ class FakeWeKnora:
         return {"id": knowledge_base_id, "indexing_strategy": {"wiki_enabled": True}}
 
     async def list_wiki_pages(self, knowledge_base_id: str, *, page: int, page_size: int):
-        return {"pages": [{"slug": item["slug"], "title": item["title"]} for item in self.pages.values()], "total_pages": 1}
+        return {
+            "pages": [{"slug": item["slug"], "title": item["title"]} for item in self.pages.values()],
+            "total": len(self.pages),
+            "total_pages": 1,
+        }
 
     async def get_wiki_page(self, knowledge_base_id: str, slug: str):
         return dict(self.pages[slug])
@@ -85,6 +92,24 @@ async def test_configured_dimensions_are_validated_but_not_sent_to_bge() -> None
     assert requests == [{"model": "bge-embedding-m3", "input": ["Wiki page"]}]
 
 
+@pytest.mark.asyncio
+async def test_embedding_error_includes_upstream_reason(caplog) -> None:
+    def handler(request: httpx.Request) -> httpx.Response:
+        return httpx.Response(400, json={"detail": "unsupported request parameter: dimensions"})
+
+    config = LocalWikiEmbeddingConfig(
+        model="bge-embedding-m3",
+        base_url="http://embedding.test/v1",
+        max_retries=0,
+    )
+    async with httpx.AsyncClient(transport=httpx.MockTransport(handler)) as http_client:
+        client = StrictWikiEmbeddingClient(config, http_client=http_client)
+        with pytest.raises(RuntimeError, match="unsupported request parameter: dimensions"):
+            await client.embed_query("Wiki page")
+
+    assert "unsupported request parameter: dimensions" in caplog.text
+
+
 @pytest.mark.asyncio
 async def test_omitted_dimensions_are_discovered_from_first_response() -> None:
     def handler(request: httpx.Request) -> httpx.Response:
@@ -118,6 +143,10 @@ async def test_sync_is_incremental_and_full_scan_deletes_missing_pages() -> None
     first = await service.sync_mapping(mapping)
     assert first["changed_pages"] == 1
     assert embedding.calls == 1
+    statuses = await store.list_page_index_status("kb")
+    assert statuses[0]["title"] == "One"
+    assert statuses[0]["vector_count"] > 0
+    assert store.states["kb"]["remote_page_count"] == 1
     second = await service.sync_mapping(mapping)
     assert second["changed_pages"] == 0
     assert embedding.calls == 1
@@ -167,3 +196,39 @@ async def test_fingerprint_change_automatically_rebuilds_entire_mapping() -> Non
     assert store.states["kb"]["state"] == "idle"
     assert store.states["kb"]["embedding_fingerprint"] == "fp-new"
     assert {vector["embedding_fingerprint"] for vectors in store.vectors.values() for vector in vectors} == {"fp-new"}
+
+
+@pytest.mark.asyncio
+async def test_auto_sync_scheduler_waits_before_first_scan(monkeypatch) -> None:
+    calls: list[str] = []
+
+    async def fake_sleep(seconds: float) -> None:
+        calls.append(f"sleep:{seconds}")
+        raise asyncio.CancelledError
+
+    async def fake_sync_once(app) -> None:
+        calls.append("sync")
+
+    monkeypatch.setattr(llmwiki_index_scheduler.asyncio, "sleep", fake_sleep)
+    monkeypatch.setattr(llmwiki_index_scheduler, "_sync_once", fake_sync_once)
+    app = SimpleNamespace(
+        state=SimpleNamespace(
+            config=SimpleNamespace(
+                llmwiki=SimpleNamespace(
+                    local_wiki_index=SimpleNamespace(sync_interval_seconds=120),
+                )
+            )
+        )
+    )
+
+    with pytest.raises(asyncio.CancelledError):
+        await llmwiki_index_scheduler.local_wiki_index_scheduler_loop(app)
+
+    assert calls == ["sleep:120"]
+
+
+def test_local_wiki_index_accepts_legacy_sync_flag() -> None:
+    config = LocalWikiIndexConfig.model_validate({"sync_enabled": True})
+
+    assert config.auto_sync is True
+    assert config.sync_enabled is True

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

@@ -0,0 +1,17 @@
+from __future__ import annotations
+
+from pathlib import Path
+
+from alembic.config import Config
+from alembic.script import ScriptDirectory
+
+
+def test_migration_graph_has_unique_revisions_and_one_head() -> None:
+    migrations = Path(__file__).parents[1] / "packages" / "harness" / "deerflow" / "persistence" / "migrations"
+    config = Config()
+    config.set_main_option("script_location", str(migrations))
+    scripts = ScriptDirectory.from_config(config)
+
+    revisions = list(scripts.walk_revisions())
+    assert len({revision.revision for revision in revisions}) == len(revisions)
+    assert len(scripts.get_heads()) == 1

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

@@ -7,6 +7,7 @@ from fastapi import HTTPException
 from starlette.requests import Request
 
 from app.gateway.routers import llmwiki
+from app.gateway.weknora_embed import verify_weknora_signed_payload
 
 
 def _request(store: object) -> Request:
@@ -177,6 +178,86 @@ async def test_public_wiki_page_list_is_normalized(monkeypatch) -> None:
     assert payload["pages"][0]["aliases"] == []
 
 
+@pytest.mark.asyncio
+async def test_public_wiki_file_is_authorized_and_streamed(monkeypatch) -> None:
+    store = PublicStore(_published_row())
+
+    class Client:
+        async def get_knowledge_base_file(self, knowledge_base_id: str, file_path: str):
+            assert knowledge_base_id == "remote-one"
+            assert file_path == "storage://tenant/resource://wiki/figure.png"
+            return b"png-bytes", "image/png"
+
+    monkeypatch.setattr(llmwiki, "_client_or_503", lambda: Client())
+
+    response = await llmwiki.get_public_wiki_file(
+        _request(store),
+        "kb-one",
+        "storage://tenant/resource://wiki/figure.png",
+    )
+
+    assert store.calls == [("kb-one", "__public_wiki__", False, False)]
+    assert response.body == b"png-bytes"
+    assert response.media_type == "image/png"
+    assert response.headers["cache-control"] == "public, max-age=300"
+    assert response.headers["content-disposition"] == "inline; filename*=UTF-8''figure.png"
+
+
+@pytest.mark.asyncio
+async def test_private_wiki_file_uses_signed_http_access_instead_of_blob_url(monkeypatch) -> None:
+    async def authorized_mapping(*_args, **_kwargs):
+        return _published_row(), "user-one", False
+
+    class Client:
+        async def get_knowledge_base_file(self, knowledge_base_id: str, file_path: str):
+            assert knowledge_base_id == "remote-one"
+            assert file_path == "storage://tenant/resource://wiki/figure.png"
+            return b"png-bytes", "image/png"
+
+    monkeypatch.setattr(llmwiki, "_authorized_mapping", authorized_mapping)
+    monkeypatch.setattr(llmwiki, "_client_or_503", lambda: Client())
+
+    access = await llmwiki.create_wiki_file_access(
+        _request(PublicStore(_published_row())),
+        "kb-one",
+        "storage://tenant/resource://wiki/figure.png",
+    )
+    payload = verify_weknora_signed_payload(access["token"])
+    assert payload["kind"] == "wiki_file"
+    assert payload["mapping_id"] == "kb-one"
+    assert payload["weknora_id"] == "remote-one"
+    assert payload["file_path"] == "storage://tenant/resource://wiki/figure.png"
+
+    response = await llmwiki.get_signed_wiki_file(access["token"])
+    assert response.body == b"png-bytes"
+    assert response.headers["cache-control"] == "private, max-age=3600"
+    assert response.headers["content-disposition"] == "inline; filename*=UTF-8''figure.png"
+
+
+def test_wiki_file_proxy_rejects_non_weknora_urls() -> None:
+    with pytest.raises(HTTPException) as exc_info:
+        llmwiki._validate_protected_wiki_file_path("https://example.test/private.png")
+
+    assert exc_info.value.status_code == 422
+
+
+@pytest.mark.asyncio
+async def test_wiki_file_proxy_adds_image_extension_for_opaque_resource_ids(monkeypatch) -> None:
+    class Client:
+        async def get_knowledge_base_file(self, knowledge_base_id: str, file_path: str):
+            return b"png-bytes", "image/png"
+
+    monkeypatch.setattr(llmwiki, "_client_or_503", lambda: Client())
+
+    response = await llmwiki._wiki_file_response(
+        {"weknora_id": "remote-one"},
+        "resource://opaque-image-id",
+        public=True,
+    )
+
+    assert response.headers["content-disposition"] == "inline; filename*=UTF-8''opaque-image-id.png"
+
+
 @pytest.mark.asyncio
 async def test_public_wiki_index_preserves_weknora_group_and_item_order(monkeypatch) -> None:
     store = PublicStore(_published_row())