fix(webhook): extract links for incremental push syncs (#2849)

GitHub webhook + sync-trigger 'sync' jobs never opted into extraction:
the standalone sync handler defaults noExtract to TRUE, so webhook-
imported pages advanced the source bookmark with links_extracted_at
unstamped — and nothing behind a standalone job ever swept them (the
autopilot cycle's extract phase only walks that cycle's changedSlugs).

Fix, both halves of the issue:
- webhook + sync-trigger payloads pass noExtract:false so normal
  incremental pushes get sync's inline link/timeline extraction.
- >100-file deferral branch: performSync surfaces extractDeferred, and
  the sync job handler queues a source-scoped extract{stale:true}
  follow-up job (idempotent sweep; extract handler gains a stale mode
  backed by the same extractStaleFromDB core as `gbrain extract --stale`).

Tests: new sync-large-deferral-extract.serial.test.ts drives a real
101-file incremental sync end-to-end (deferral flag → queued follow-up →
stale sweep recovers links + watermark); webhook payload contract test +
sync-trigger noExtract assertion salvaged from #2850.

Fixes #2849
Takeover of #2850

Co-authored-by: patentsong <patentsong@users.noreply.github.com>
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

src/commands/extract.ts

@@ -433,10 +433,7 @@ export async function extractLinksFromFile(
       async resolve(name: string, dirHint?: string | string[]): Promise<string | null> {
         if (!name) return null;
         const trimmed = name.trim();
-        // Same broadened slug-shape as makeResolver step 1: accepts
-        // digit-leading folders (`90-people/nicolai`) and nested paths.
-        // Exact Set membership guards it — no false positives.
-        if (/\//.test(trimmed) && /^[a-z0-9][a-z0-9/_-]*$/.test(trimmed) && allSlugs.has(trimmed)) {
+        if (/^[a-z][a-z0-9-]*\/[a-z0-9][a-z0-9-]*$/.test(trimmed) && allSlugs.has(trimmed)) {
           return trimmed;
         }
         const hints = Array.isArray(dirHint) ? dirHint : (dirHint ? [dirHint] : []);
@@ -585,17 +582,6 @@ export interface ExtractOpts {
    * before (single-'default'-source brains unaffected).
    */
   sourceId?: string;
-  /**
-   * v0.42 — also extract frontmatter links on the incremental (slugs) path.
-   * `extractForSlugs` extracts BODY links only by default; set this true to also
-   * parse each changed page's frontmatter so `sources:`/`related:` edges stay fresh
-   * when YAML is edited externally and synced in. Applied PER changed page, so the
-   * incremental walk stays bounded (no switch to a full DB scan). Only honored on
-   * the incremental path (`slugs` defined); the full-walk path already covers
-   * frontmatter via its own dispatch. Gated upstream by the config key
-   * `autopilot.incremental_extract_include_frontmatter` (default off).
-   */
-  includeFrontmatter?: boolean;
 }
 
 /**
@@ -634,7 +620,7 @@ export async function runExtractCore(engine: BrainEngine, opts: ExtractOpts): Pr
       // Nothing changed — skip entirely.
       return result;
     }
-    const r = await extractForSlugs(engine, opts.dir, opts.slugs, opts.mode, dryRun, jsonMode, workers, opts.signal, opts.sourceId, opts.includeFrontmatter);
+    const r = await extractForSlugs(engine, opts.dir, opts.slugs, opts.mode, dryRun, jsonMode, workers, opts.signal, opts.sourceId);
     result.links_created = r.links_created;
     result.timeline_entries_created = r.timeline_created;
     result.pages_processed = r.pages;
@@ -1025,11 +1011,6 @@ async function extractForSlugs(
   signal?: AbortSignal,
   // #1747/#1503: stamp resolved brain source id on batch rows (see ExtractOpts.sourceId).
   sourceId?: string,
-  // v0.42: when true, also extract frontmatter links per changed page so
-  // externally-edited YAML (`sources:`/`related:`) stays fresh on the cycle.
-  // Default false preserves the body-only incremental behavior. Gated upstream
-  // by `autopilot.incremental_extract_include_frontmatter`.
-  includeFrontmatter: boolean = false,
 ): Promise<{ links_created: number; timeline_created: number; pages: number }> {
   // Build the full slug set for link resolution (fast: just readdir, no file reads)
   const allFiles = walkMarkdownFiles(brainDir);
@@ -1104,7 +1085,7 @@ async function extractForSlugs(
         const content = readFileSync(fullPath, 'utf-8');
 
         if (doLinks) {
-          const links = await extractLinksFromFile(content, relPath, allSlugs, { globalBasename, includeFrontmatter });
+          const links = await extractLinksFromFile(content, relPath, allSlugs, { globalBasename });
           for (const link of links) {
             if (dryRun) {
               if (!jsonMode) console.log(`  ${link.from_slug} → ${link.to_slug} (${link.link_type})`);
@@ -1670,7 +1651,7 @@ async function extractTimelineFromDB(
  * make re-extraction idempotent). EVERY processed page is stamped, including
  * zero-link pages — they WERE processed.
  */
-async function extractStaleFromDB(
+export async function extractStaleFromDB(
   engine: BrainEngine,
   opts: {
     dryRun: boolean;

src/commands/jobs.ts

@@ -1479,7 +1479,31 @@ export async function registerBuiltinHandlers(
       embedSkipReason = 'auto_embed_disabled';
     }
 
-    return { ...result, embed_job_id: embedJobId, embed_skip_reason: embedSkipReason };
+    // #2849: large-sync extract deferral follow-up. performSync skips inline
+    // link/timeline extraction when totalChanges > 100, leaving
+    // links_extracted_at unstamped. A standalone sync job (webhook push,
+    // sync trigger) has no autopilot extract phase behind it, so the pages
+    // would stay extraction-stale until a manual `gbrain extract --stale`.
+    // Queue a source-scoped stale sweep instead. Best-effort + idempotent:
+    // a duplicate sweep finds 0 stale pages and no-ops.
+    let extractJobId: number | null = null;
+    if (result.extractDeferred) {
+      try {
+        const { MinionQueue } = await import('../core/minions/queue.ts');
+        const queue = new MinionQueue(engine);
+        const followUp = await queue.add(
+          'extract',
+          { stale: true, ...(sourceId ? { sourceId } : {}) },
+          {
+            idempotency_key: `sync-extract-stale:${sourceId ?? 'default'}:${Math.floor(Date.now() / 30_000)}`,
+            maxWaiting: 1,
+          },
+        );
+        extractJobId = followUp.id;
+      } catch { /* best-effort: extract --stale sweeps it later */ }
+    }
+
+    return { ...result, embed_job_id: embedJobId, embed_skip_reason: embedSkipReason, extract_stale_job_id: extractJobId };
   });
 
   registerBuiltinJob(worker, engine, 'embed', async (job) => {
@@ -1652,6 +1676,20 @@ export async function registerBuiltinHandlers(
   });
 
   worker.register('extract', async (job) => {
+    // #2849: stale-sweep mode — the sync handler's large-sync deferral
+    // follow-up. DB-source (reads page content from the DB, so it runs on
+    // checkout-less brains), source-scopable, idempotent. Same core as
+    // `gbrain extract --stale`.
+    if (job.data.stale === true) {
+      const { extractStaleFromDB } = await import('./extract.ts');
+      return await extractStaleFromDB(engine, {
+        dryRun: !!job.data.dryRun,
+        jsonMode: false,
+        includeFrontmatter: false,
+        sourceIdFilter: typeof job.data.sourceId === 'string' ? job.data.sourceId : undefined,
+        catchUp: false,
+      });
+    }
     const { runExtractCore } = await import('./extract.ts');
     const mode = (typeof job.data.mode === 'string' && ['links', 'timeline', 'all'].includes(job.data.mode))
       ? (job.data.mode as 'links' | 'timeline' | 'all')

src/commands/serve-http.ts

@@ -2146,8 +2146,13 @@ export async function runServeHttp(engine: BrainEngine, options: ServeHttpOption
   //     Other event types (ping, pull_request, etc.) return 202 'ignored'
   //     so GitHub doesn't retry.
   // D15.5: HMAC compare uses the shared safeHexEqual helper.
-  // D18: submits 'sync' job with auto_embed_backfill=true and priority -10
-  //     (above autopilot's 0).
+  // D18: submits 'sync' job with extraction + auto_embed_backfill enabled and
+  //     priority -10 (above autopilot's 0). noExtract:false opts normal
+  //     incremental pushes into sync's inline link/timeline extraction (#2849
+  //     — the standalone sync handler defaults noExtract to TRUE, which left
+  //     webhook-imported pages permanently stale). Large (>100 file) pushes
+  //     defer inline extract; the sync handler queues an extract --stale
+  //     follow-up job for that branch.
   // ---------------------------------------------------------------------------
   const githubWebhookLimiter = rateLimit({
     windowMs: 60_000,
@@ -2267,6 +2272,7 @@ export async function runServeHttp(engine: BrainEngine, options: ServeHttpOption
           'sync',
           {
             sourceId: source.id,
+            noExtract: false,
             auto_embed_backfill: true,
             embed_reason: 'webhook',
           },

src/commands/sync.ts

@@ -222,6 +222,14 @@ export interface SyncResult {
    * everything," the exact misdiagnosis in the #1794 recurrence report.
    */
   bankedFiles?: number;
+  /**
+   * #2849: true when extraction was REQUESTED (noExtract false) but this sync
+   * skipped inline link/timeline extraction because totalChanges > 100 (the
+   * #1794 large-sync deferral). links_extracted_at stays unstamped for the
+   * imported pages. The standalone `sync` job handler queues a source-scoped
+   * `extract --stale` follow-up when set; CLI runs print the manual hint.
+   */
+  extractDeferred?: boolean;
 }
 
 /**
@@ -1379,6 +1387,10 @@ See also:
     {
       sourceId: sourceIdArg,
       repoPath: source.local_path,
+      // #2849: opt in to inline extraction — the standalone sync handler
+      // defaults noExtract to TRUE (dedupe for doctor's [sync, extract]
+      // remediation plan), which would leave triggered syncs extraction-stale.
+      noExtract: false,
       auto_embed_backfill: true,
       embed_reason: 'sync_trigger',
     },
@@ -3287,11 +3299,16 @@ async function performSyncInner(engine: BrainEngine, opts: SyncOpts): Promise<Sy
   // the stale sweep scans the whole source, so banked-across-runs pages are
   // covered regardless.
   const extractOpts = opts.sourceId ? { sourceId: opts.sourceId } : undefined;
+  let extractDeferred = false;
   if (!opts.noExtract && totalChanges > 100 && pagesAffected.length > 0) {
+    // #2849: surface the deferral to callers. A standalone sync job (webhook
+    // push, sync trigger) has no autopilot extract phase behind it, so the
+    // job handler queues an `extract --stale` follow-up off this flag.
+    extractDeferred = true;
     slog(
       `  Large sync: deferring link/timeline extraction. ` +
       `Run 'gbrain extract --stale${opts.sourceId ? ` --source-id ${opts.sourceId}` : ''}' ` +
-      `(or let the autopilot cycle's extract phase sweep it).`,
+      `(sync jobs queue this follow-up automatically).`,
     );
   }
   if (!opts.noExtract && totalChanges <= 100 && pagesAffected.length > 0) {
@@ -3400,6 +3417,7 @@ async function performSyncInner(engine: BrainEngine, opts: SyncOpts): Promise<Sy
     chunksCreated,
     embedded,
     pagesAffected,
+    extractDeferred,
   };
 }
 

src/core/config.ts

@@ -121,18 +121,6 @@ export interface GBrainConfig {
       /** Daily spend cap (USD); bounds drains/day = floor(cap / ~$0.30). Default 2.0. */
       max_usd_per_day?: number;
     };
-    /**
-     * v0.42 — keep frontmatter links fresh on the incremental cycle. The cycle's
-     * extract phase re-extracts only the slugs a sync changed, but `extractForSlugs`
-     * extracts BODY links only — frontmatter (`sources:`/`related:` etc.) link edges
-     * silently drift stale when a page's YAML is edited externally and synced in.
-     * Set true to also extract frontmatter links per changed page each cycle, keeping
-     * externally-edited YAML edges fresh without a full rescan. Default false
-     * (preserves current behavior). Read via the file/env/DB plane in the cycle's
-     * extract dispatch. Disable/enable with
-     * `gbrain config set autopilot.incremental_extract_include_frontmatter <bool>`.
-     */
-    incremental_extract_include_frontmatter?: boolean;
   };
   eval?: {
     /** false disables capture entirely. Defaults to true. */

src/core/cycle.ts

@@ -996,21 +996,6 @@ async function runPhaseExtract(
 ): Promise<PhaseResult> {
   try {
     const { runExtractCore } = await import('../commands/extract.ts');
-    const { loadConfig } = await import('./config.ts');
-    // Default off: the incremental cycle extracts body links only unless the
-    // operator opts in to keeping externally-edited frontmatter links fresh too.
-    // Both planes, file wins (env > file > DB precedence, per loadConfigWithEngine):
-    // `gbrain config set autopilot.incremental_extract_include_frontmatter true`
-    // writes the DB plane (engine.setConfig), so a file-plane-only read here
-    // would make the documented enable command a silent no-op (#2120 class).
-    const fileVal = loadConfig()?.autopilot?.incremental_extract_include_frontmatter;
-    let includeFrontmatter = fileVal === true;
-    if (fileVal === undefined) {
-      try {
-        includeFrontmatter =
-          (await engine.getConfig('autopilot.incremental_extract_include_frontmatter')) === 'true';
-      } catch { /* config table unreadable → default off */ }
-    }
     // Extract is read-mostly against the filesystem + write to links table.
     // Honor dryRun by skipping with a 'skipped' entry: extract doesn't have
     // a clean dry-run mode today and runCycle should be honest about it.
@@ -1031,7 +1016,6 @@ async function runPhaseExtract(
       slugs: changedSlugs,  // undefined = full walk (first run / manual)
       signal,
       sourceId,
-      includeFrontmatter,  // honored on the incremental (slugs) path only
     });
     const linksCreated = result?.links_created ?? 0;
     const timelineCreated = result?.timeline_entries_created ?? 0;

src/core/link-extraction.ts

@@ -942,17 +942,8 @@ export function makeResolver(
 
       const hints = Array.isArray(dirHint) ? dirHint : (dirHint ? [dirHint] : []);
 
-      // Step 1: already a slug? Try an exact page lookup for any slug-shaped
-      // value (contains '/', slug charset). Broadened beyond the original
-      // single-segment lowercase-leading form (`^[a-z][a-z0-9-]*\/[a-z0-9]...`)
-      // to also accept digit-leading folders (`90-people/nicolai`,
-      // `01-trading/...`) and nested paths (`a/b/c`) — common in PARA-numbered
-      // vaults. This is an EXACT getPage match only — no fuzzy — so it never
-      // produces a false positive; a non-existent slug just falls through to
-      // the steps below. Fixes frontmatter `related: [[dir/slug]]` values
-      // (unwrapped by unwrapWikilink) that name a real page the strict regex
-      // could not reach and whose full-path fuzzy score is below threshold.
-      if (/\//.test(trimmed) && /^[a-z0-9][a-z0-9/_-]*$/.test(trimmed)) {
+      // Step 1: already a slug? (dir/name shape, lowercase, hyphenated)
+      if (/^[a-z][a-z0-9-]*\/[a-z0-9][a-z0-9-]*$/.test(trimmed)) {
         const page = await engine.getPage(trimmed);
         if (page) {
           cache.set(cacheKey, trimmed);
@@ -1012,25 +1003,6 @@ export function makeResolver(
 
 // ─── Frontmatter extractor ──────────────────────────────────────
 
-/**
- * Unwrap an Obsidian `[[wikilink]]` frontmatter value to its bare link
- * target so the resolver (which expects bare titles / dir slugs) can match
- * it. Mainstream Obsidian authors frontmatter links as `related: ["[[Page]]"]`;
- * without this, the resolver treats the brackets as part of the value and a
- * `[[90-people/nicolai]]` is normalized into `90peoplenicolai`, so it never
- * resolves. Strips a trailing `|alias`, `#heading`, or `^block` suffix — the
- * link target only. The regex is anchored to a wholly-wrapped value
- * (`^\s*\[\[…\]\]\s*$`), so bare titles and any value not fully wrapped pass
- * through unchanged and existing behavior is preserved exactly.
- */
-export function unwrapWikilink(value: string): string {
-  const match = /^\s*\[\[(.+?)\]\]\s*$/.exec(value);
-  if (!match) return value;
-  // Take the link target: drop |alias, then #heading / ^block suffixes.
-  const target = match[1].split('|')[0].split('#')[0].split('^')[0];
-  return target.trim();
-}
-
 export interface UnresolvedFrontmatterRef {
   /** The frontmatter field name. */
   field: string;
@@ -1088,12 +1060,7 @@ export async function extractFrontmatterLinks(
         }
         if (!name) continue;   // skip numbers, nulls, malformed objects
 
-        // Accept Obsidian `[[wikilink]]` values in frontmatter link fields by
-        // unwrapping to the bare target before resolution. Bare titles pass
-        // through unchanged; the original `name` is preserved for the
-        // unresolved report and edge context.
-        const linkTarget = unwrapWikilink(name);
-        const resolved = await resolver.resolve(linkTarget, mapping.dirHint);
+        const resolved = await resolver.resolve(name, mapping.dirHint);
         if (!resolved) {
           unresolved.push({ field, name });
           continue;

test/extract-incremental.test.ts

@@ -191,37 +191,3 @@ describe('runExtractCore — incremental cycle path (#417)', () => {
     expect(result.links_created).toBeGreaterThan(0);
   });
 });
-describe('runExtractCore — incremental frontmatter gate (includeFrontmatter)', () => {
-  // alice has a `source:` frontmatter edge but NO body links. The incremental
-  // path extracts body links only by default, so the frontmatter edge is the
-  // sole signal that distinguishes the gate off vs on.
-  const aliceFm = '---\nsource: companies/acme-example\n---\n# alice';
-
-  test('9. default (flag omitted) does NOT extract frontmatter links on the incremental path', async () => {
-    await seedPage('companies/acme-example', '# acme');
-    await seedPage('people/alice-example', aliceFm);
-    const result = await runExtractCore(engine as unknown as BrainEngine, {
-      mode: 'all',
-      dir: tempDir,
-      slugs: ['people/alice-example'],
-    });
-    // alice's only potential edge is her frontmatter `source:`; with the gate off
-    // it must not be extracted (preserves the body-only incremental behavior).
-    expect(result.pages_processed).toBe(1);
-    expect(result.links_created).toBe(0);
-  });
-
-  test('10. includeFrontmatter: true extracts the frontmatter link on the incremental path', async () => {
-    await seedPage('companies/acme-example', '# acme');
-    await seedPage('people/alice-example', aliceFm);
-    const result = await runExtractCore(engine as unknown as BrainEngine, {
-      mode: 'all',
-      dir: tempDir,
-      slugs: ['people/alice-example'],
-      includeFrontmatter: true,
-    });
-    // Same page, gate on → the `source:` frontmatter edge is now extracted.
-    expect(result.pages_processed).toBe(1);
-    expect(result.links_created).toBeGreaterThan(0);
-  });
-});

test/extract.test.ts

@@ -76,18 +76,6 @@ describe('extractLinksFromFile', () => {
     }
   });
 
-  it('resolves wrapped [[wikilink]] digit-leading slug-path in frontmatter (fs resolver, broadened step 1)', async () => {
-    // Same bug class as makeResolver step 1 (#1983): the fs resolver's strict
-    // `^[a-z]…` slug regex rejected digit-leading / nested paths, so a PARA-vault
-    // `related: "[[90-people/nicolai]]"` never resolved even though the page exists.
-    const content = '---\nrelated: "[[90-people/nicolai]]"\ntype: concept\n---\nContent.';
-    const allSlugs = new Set(['wiki/note', '90-people/nicolai']);
-    const links = await extractLinksFromFile(content, 'wiki/note.md', allSlugs, { includeFrontmatter: true });
-    const related = links.filter(l => l.link_type === 'related_to');
-    expect(related).toHaveLength(1);
-    expect(related[0].to_slug).toBe('90-people/nicolai');
-  });
-
   it('frontmatter extraction is default OFF (back-compat)', async () => {
     // Without includeFrontmatter, fs-source no longer auto-extracts frontmatter.
     // Matches db-source behavior. User opts in with --include-frontmatter flag.

test/link-extraction.test.ts

@@ -9,7 +9,6 @@ import {
   parseTimelineEntries,
   isAutoLinkEnabled,
   FRONTMATTER_LINK_MAP,
-  unwrapWikilink,
   type SlugResolver,
 } from '../src/core/link-extraction.ts';
 import type { BrainEngine } from '../src/core/engine.ts';
@@ -1292,175 +1291,3 @@ describe('parseTimelineEntries — Format 3: inline [Source: ..., YYYY-MM-DD] ci
     expect(parseTimelineEntries('[Source: import batch, 2025-07-01]')).toHaveLength(0);
   });
 });
-// ─── Frontmatter [[wikilink]] + slug-path resolution ──────────────────────
-// Mainstream Obsidian authors frontmatter links as `related: ["[[Page]]"]`,
-// and PARA-numbered vaults use digit-leading / nested slug paths like
-// `[[90-people/nicolai]]`. Both were silently dropped: brackets were treated
-// as part of the value and the step-1 slug regex (`^[a-z]…`) rejected
-// digit-leading / nested paths, while full-path fuzzy scored below threshold.
-// Fix: unwrapWikilink() before resolution + an exact getPage() for any
-// slug-shaped value (exact-match only → no false positives).
-
-describe('unwrapWikilink', () => {
-  test('wrapped title → bare title', () => {
-    expect(unwrapWikilink('[[Monday Range]]')).toBe('Monday Range');
-  });
-  test('wrapped slug-path (digit-leading folder) → bare slug', () => {
-    expect(unwrapWikilink('[[90-people/nicolai]]')).toBe('90-people/nicolai');
-  });
-  test('wrapped nested slug-path → bare slug', () => {
-    expect(unwrapWikilink('[[01-trading/wiki/strategies/opening-range-breakout]]'))
-      .toBe('01-trading/wiki/strategies/opening-range-breakout');
-  });
-  test('strips |alias', () => {
-    expect(unwrapWikilink('[[90-people/nicolai|Nicolai]]')).toBe('90-people/nicolai');
-  });
-  test('strips #heading', () => {
-    expect(unwrapWikilink('[[Page#Section]]')).toBe('Page');
-  });
-  test('strips ^block', () => {
-    expect(unwrapWikilink('[[Page^abc123]]')).toBe('Page');
-  });
-  test('surrounding whitespace tolerated', () => {
-    expect(unwrapWikilink('  [[Page]]  ')).toBe('Page');
-  });
-  test('bare title passes through unchanged', () => {
-    expect(unwrapWikilink('Monday Range')).toBe('Monday Range');
-  });
-  test('bare slug passes through unchanged', () => {
-    expect(unwrapWikilink('90-people/nicolai')).toBe('90-people/nicolai');
-  });
-  test('partially-wrapped value is NOT unwrapped (anchored)', () => {
-    // Not a wholly-wrapped value → left intact so existing behavior is exact.
-    expect(unwrapWikilink('see [[Page]] for detail')).toBe('see [[Page]] for detail');
-  });
-});
-
-describe('makeResolver — slug-path exact getPage (step 1 broadened)', () => {
-  function fakeEngine(
-    slugs: string[],
-    fuzzyMap: Map<string, { slug: string; similarity: number }> = new Map(),
-  ): BrainEngine {
-    const lookup = new Set(slugs);
-    return {
-      async getPage(slug: string) { return lookup.has(slug) ? { slug } as any : null; },
-      async findByTitleFuzzy(name: string) { return fuzzyMap.get(name) ?? null; },
-      async searchKeyword() { return []; },
-    } as unknown as BrainEngine;
-  }
-
-  test('digit-leading folder slug resolves via exact getPage', async () => {
-    const r = makeResolver(fakeEngine(['90-people/nicolai']));
-    expect(await r.resolve('90-people/nicolai')).toBe('90-people/nicolai');
-  });
-
-  test('nested (>2 segment) slug resolves via exact getPage', async () => {
-    const r = makeResolver(fakeEngine(['01-trading/wiki/strategies/opening-range-breakout']));
-    expect(await r.resolve('01-trading/wiki/strategies/opening-range-breakout'))
-      .toBe('01-trading/wiki/strategies/opening-range-breakout');
-  });
-
-  test('regression: single-segment lowercase slug still resolves', async () => {
-    const r = makeResolver(fakeEngine(['people/pedro']));
-    expect(await r.resolve('people/pedro')).toBe('people/pedro');
-  });
-
-  test('exact-only: slug-shaped value with no matching page falls through (no false positive)', async () => {
-    // `90-people/ghost` is slug-shaped but absent → step-1 getPage misses,
-    // no fuzzy hit → null. Never invents an edge.
-    const r = makeResolver(fakeEngine(['90-people/nicolai']));
-    expect(await r.resolve('90-people/ghost')).toBeNull();
-  });
-
-  test('non-slug value still routes to fuzzy', async () => {
-    const r = makeResolver(fakeEngine(
-      ['01-trading/monday-range'],
-      new Map([['Monday Range', { slug: '01-trading/monday-range', similarity: 1 }]]),
-    ));
-    expect(await r.resolve('Monday Range')).toBe('01-trading/monday-range');
-  });
-});
-
-describe('extractFrontmatterLinks — [[wikilink]] related: values (end-to-end)', () => {
-  function fakeEngine(
-    slugs: string[],
-    fuzzyMap: Map<string, { slug: string; similarity: number }> = new Map(),
-  ): BrainEngine {
-    const lookup = new Set(slugs);
-    return {
-      async getPage(slug: string) { return lookup.has(slug) ? { slug } as any : null; },
-      async findByTitleFuzzy(name: string) { return fuzzyMap.get(name) ?? null; },
-      async searchKeyword() { return []; },
-    } as unknown as BrainEngine;
-  }
-
-  test('wrapped slug-path related: resolves (the core win)', async () => {
-    const resolver = makeResolver(fakeEngine(['90-people/nicolai']));
-    const { candidates, unresolved } = await extractFrontmatterLinks(
-      'wiki/originals/ideas/note', 'note' as never,
-      { related: '[[90-people/nicolai]]' }, resolver,
-    );
-    expect(unresolved).toHaveLength(0);
-    expect(candidates).toHaveLength(1);
-    expect(candidates[0]).toMatchObject({
-      fromSlug: 'wiki/originals/ideas/note',
-      targetSlug: '90-people/nicolai',
-      linkType: 'related_to',
-      linkSource: 'frontmatter',
-    });
-  });
-
-  test('wrapped nested slug-path related: resolves', async () => {
-    const resolver = makeResolver(fakeEngine(['01-trading/wiki/strategies/opening-range-breakout']));
-    const { candidates } = await extractFrontmatterLinks(
-      'wiki/note', 'note' as never,
-      { related: ['[[01-trading/wiki/strategies/opening-range-breakout]]'] }, resolver,
-    );
-    expect(candidates).toHaveLength(1);
-    expect(candidates[0].targetSlug).toBe('01-trading/wiki/strategies/opening-range-breakout');
-  });
-
-  test('wrapped value with |alias resolves to the target', async () => {
-    const resolver = makeResolver(fakeEngine(['90-people/nicolai']));
-    const { candidates } = await extractFrontmatterLinks(
-      'wiki/note', 'note' as never,
-      { related: '[[90-people/nicolai|Nicolai]]' }, resolver,
-    );
-    expect(candidates).toHaveLength(1);
-    expect(candidates[0].targetSlug).toBe('90-people/nicolai');
-  });
-
-  test('regression: bare slug related: still resolves', async () => {
-    const resolver = makeResolver(fakeEngine(['90-people/nicolai']));
-    const { candidates } = await extractFrontmatterLinks(
-      'wiki/note', 'note' as never,
-      { related: '90-people/nicolai' }, resolver,
-    );
-    expect(candidates).toHaveLength(1);
-    expect(candidates[0].targetSlug).toBe('90-people/nicolai');
-  });
-
-  test('regression: wrapped title resolves via fuzzy (brackets harmless)', async () => {
-    const resolver = makeResolver(fakeEngine(
-      ['01-trading/monday-range'],
-      new Map([['Monday Range', { slug: '01-trading/monday-range', similarity: 1 }]]),
-    ));
-    const { candidates } = await extractFrontmatterLinks(
-      'wiki/note', 'note' as never,
-      { related: '[[Monday Range]]' }, resolver,
-    );
-    expect(candidates).toHaveLength(1);
-    expect(candidates[0].targetSlug).toBe('01-trading/monday-range');
-  });
-
-  test('unknown wrapped slug → unresolved (no crash), original value preserved', async () => {
-    const resolver = makeResolver(fakeEngine(['90-people/nicolai']));
-    const { candidates, unresolved } = await extractFrontmatterLinks(
-      'wiki/note', 'note' as never,
-      { related: '[[99-archive/does-not-exist]]' }, resolver,
-    );
-    expect(candidates).toHaveLength(0);
-    expect(unresolved).toHaveLength(1);
-    expect(unresolved[0]).toEqual({ field: 'related', name: '[[99-archive/does-not-exist]]' });
-  });
-});

test/sources-webhook.test.ts

@@ -16,6 +16,7 @@
  */
 import { describe, test, expect } from 'bun:test';
 import { createHmac } from 'node:crypto';
+import { readFileSync } from 'node:fs';
 import { safeHexEqual } from '../src/core/timing-safe.ts';
 
 const GITHUB_SECRET = 'super-secret-webhook-key';
@@ -123,3 +124,25 @@ describe('Branch ref construction (D5)', () => {
     expect(pushedRef === `refs/heads/${trackedBranch}`).toBe(false);
   });
 });
+
+describe('Webhook sync job extraction contract (#2849)', () => {
+  test('opts into extraction before the pushed commit is consumed', () => {
+    const serveSource = readFileSync(
+      new URL('../src/commands/serve-http.ts', import.meta.url),
+      'utf8',
+    );
+    const routeStart = serveSource.indexOf("'/webhooks/github'");
+    const queueStart = serveSource.indexOf('const job = await queue.add(', routeStart);
+    const responseStart = serveSource.indexOf('res.status(202)', queueStart);
+    expect(routeStart).toBeGreaterThanOrEqual(0);
+    expect(queueStart).toBeGreaterThan(routeStart);
+    expect(responseStart).toBeGreaterThan(queueStart);
+
+    const routeSource = serveSource.slice(queueStart, responseStart);
+    const payload = routeSource.match(
+      /queue\.add\(\s*'sync',\s*\{([\s\S]*?)\}\s*,\s*\{/,
+    );
+    expect(payload).not.toBeNull();
+    expect(payload?.[1]).toMatch(/\bnoExtract:\s*false\b/);
+  });
+});

test/sync-large-deferral-extract.serial.test.ts

@@ -0,0 +1,130 @@
+/**
+ * #2849 — large-sync extract deferral queues an `extract --stale` follow-up.
+ *
+ * performSync's incremental path skips inline link/timeline extraction when
+ * totalChanges > 100 (the #1794 large-sync deferral), leaving
+ * links_extracted_at unstamped. Pre-fix, a standalone sync job (webhook push,
+ * `gbrain sync trigger`) had NOTHING behind it to sweep those pages — the
+ * autopilot cycle's extract phase only walks that cycle's changedSlugs — so a
+ * large webhook push left extraction permanently stale until a manual
+ * `gbrain extract --stale`.
+ *
+ * Pins:
+ *   (a) performSync surfaces `extractDeferred: true` on the >100 branch and
+ *       leaves the pages unstamped/unlinked.
+ *   (b) the `sync` job handler queues an `extract` job with
+ *       { stale: true, sourceId? } when extractDeferred is set.
+ *   (c) the `extract` handler's stale mode actually sweeps: links created +
+ *       watermark stamped (end-to-end recovery, no manual step).
+ *
+ * Marked .serial.test.ts — spawns git subprocesses + shares one PGLite engine.
+ */
+
+import { describe, test, expect, beforeAll, afterAll } from 'bun:test';
+import { mkdtempSync, writeFileSync, rmSync, mkdirSync } from 'fs';
+import { execSync } from 'child_process';
+import { tmpdir } from 'os';
+import { join } from 'path';
+import { PGLiteEngine } from '../src/core/pglite-engine.ts';
+import { MinionWorker } from '../src/core/minions/worker.ts';
+import { MinionQueue } from '../src/core/minions/queue.ts';
+import { registerBuiltinHandlers } from '../src/commands/jobs.ts';
+
+let engine: PGLiteEngine;
+let worker: MinionWorker;
+let repoPath: string;
+
+function git(cmd: string): void { execSync(cmd, { cwd: repoPath, stdio: 'pipe' }); }
+
+describe('#2849 — large sync defers extract and queues a stale sweep', () => {
+  beforeAll(async () => {
+    engine = new PGLiteEngine();
+    await engine.connect({});
+    await engine.initSchema();
+    worker = new MinionWorker(engine, { queue: 'test' });
+    await registerBuiltinHandlers(worker, engine, { quiet: true });
+
+    repoPath = mkdtempSync(join(tmpdir(), 'gbrain-large-defer-'));
+    git('git init');
+    git('git config user.email "t@t.com"');
+    git('git config user.name "T"');
+    mkdirSync(join(repoPath, 'people'), { recursive: true });
+    mkdirSync(join(repoPath, 'notes'), { recursive: true });
+    writeFileSync(join(repoPath, 'people/alice.md'), [
+      '---', 'type: person', 'title: Alice', '---', '', 'Alice is a founder.',
+    ].join('\n'));
+    git('git add -A && git commit -m "initial"');
+
+    // Seed: full first sync imports the anchor page + sets last_commit.
+    const { performSync } = await import('../src/commands/sync.ts');
+    await performSync(engine, { repoPath, full: true, noPull: true, noEmbed: true });
+
+    // Second commit: 101 new pages → incremental totalChanges > 100.
+    for (let i = 0; i < 101; i++) {
+      writeFileSync(join(repoPath, `notes/n${i}.md`), [
+        '---', 'type: note', `title: Note ${i}`, '---', '',
+        `[Alice](people/alice) appears in note ${i}.`,
+      ].join('\n'));
+    }
+    git('git add -A && git commit -m "add 101 pages"');
+  }, 120_000);
+
+  afterAll(async () => {
+    if (repoPath) rmSync(repoPath, { recursive: true, force: true });
+    if (engine) await engine.disconnect();
+  }, 60_000);
+
+  test('sync handler defers inline extract and queues extract{stale} follow-up; stale sweep recovers', async () => {
+    const syncHandler = (worker as unknown as { handlers: Map<string, (job: unknown) => Promise<unknown>> })
+      .handlers.get('sync');
+    expect(syncHandler).toBeDefined();
+
+    // Same payload shape the webhook submits (minus embed backfill noise).
+    const result = await syncHandler!({
+      data: { repoPath, noExtract: false, noPull: true, auto_embed_backfill: false },
+      signal: { aborted: false },
+      updateProgress: async () => {},
+    }) as { status: string; extractDeferred?: boolean; extract_stale_job_id?: number | null };
+
+    expect(result.status).toBe('synced');
+    // (a) inline extract was deferred, pages left stale.
+    expect(result.extractDeferred).toBe(true);
+    const staleBefore = await engine.countStalePagesForExtraction();
+    expect(staleBefore).toBeGreaterThan(100);
+    expect(await engine.getLinks('notes/n0')).toHaveLength(0);
+
+    // (b) a follow-up extract job with stale:true was queued.
+    expect(result.extract_stale_job_id).toBeGreaterThan(0);
+    const queue = new MinionQueue(engine);
+    const extractJobs = await queue.getJobs({ name: 'extract', limit: 5 });
+    expect(extractJobs.length).toBe(1);
+    expect((extractJobs[0].data as { stale: boolean }).stale).toBe(true);
+
+    // (c) running the extract handler's stale mode recovers: links + stamps.
+    const extractHandler = (worker as unknown as { handlers: Map<string, (job: unknown) => Promise<unknown>> })
+      .handlers.get('extract');
+    await extractHandler!({
+      data: extractJobs[0].data,
+      signal: { aborted: false },
+      updateProgress: async () => {},
+    });
+    const links = await engine.getLinks('notes/n0');
+    expect(links.some(l => l.to_slug === 'people/alice')).toBe(true);
+    const rows = await engine.executeRaw<{ links_extracted_at: string | null }>(
+      `SELECT links_extracted_at FROM pages WHERE slug = 'notes/n0'`,
+    );
+    expect(rows[0]?.links_extracted_at).not.toBeNull();
+  }, 180_000);
+
+  test('sub-threshold sync does NOT set extractDeferred (no spurious follow-up)', async () => {
+    // One more small commit → inline extract path, no deferral.
+    writeFileSync(join(repoPath, 'notes/small.md'), [
+      '---', 'type: note', 'title: Small', '---', '', 'No big deal.',
+    ].join('\n'));
+    git('git add -A && git commit -m "one small page"');
+    const { performSync } = await import('../src/commands/sync.ts');
+    const result = await performSync(engine, { repoPath, noPull: true, noEmbed: true });
+    expect(result.status).toBe('synced');
+    expect(result.extractDeferred).toBeFalsy();
+  }, 60_000);
+});

test/sync-trigger-cli.test.ts

@@ -100,6 +100,10 @@ describe('runSyncTrigger', () => {
     const job = jobs[0];
     expect(job.priority).toBe(-10);
     expect((job.data as { sourceId: string }).sourceId).toBe('default');
+    // #2849: opt in to inline extraction — the standalone sync handler
+    // defaults noExtract to TRUE, which would leave triggered syncs
+    // extraction-stale.
+    expect((job.data as { noExtract: boolean }).noExtract).toBe(false);
     expect((job.data as { auto_embed_backfill: boolean }).auto_embed_backfill).toBe(true);
   });