fix(extract): resolve cross-directory wikilinks via sync-consistent slugification (#1964)

Raw Obsidian wikilink paths ([[llm-wiki/entities/AI 3.0]]) never matched
sync-slugified slugs (llm-wiki/entities/ai-3.0) on either extraction path.

- FS path: resolveSlug now slugifies each candidate (slugifyPath, the same
  function sync uses) when the raw path-equality check misses, in both the
  direct join and the ancestor walk. Dir-qualified wikilinks resolve as
  typed edges without the global_basename flag.
- DB path: extractPageLinks resolves dir-qualified generic wikilinks by
  slugifying the path and matching by exact slug or path suffix against
  the basename index — regardless of global_basename (the flag guards
  bare-name cross-dir ambiguity, which doesn't apply to dir-qualified
  paths). A same-tail page in a different directory can no longer be the
  only candidate and is filtered out.

Fixes #1964

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

src/commands/extract.ts

@@ -43,7 +43,7 @@ import {
 } from '../core/link-extraction.ts';
 import { createProgress } from '../core/progress.ts';
 import { getCliOptions, cliOptsToProgressOptions } from '../core/cli-options.ts';
-import { pathToSlug, pruneDir, isSyncable } from '../core/sync.ts';
+import { pathToSlug, slugifyPath, pruneDir, isSyncable } from '../core/sync.ts';
 // v0.41.18.0: withRetry + isRetryableConnError + WithRetryOpts moved to
 // src/core/retry.ts as the canonical primitive. Engine methods
 // (addLinksBatch/addTimelineEntriesBatch/upsertChunks) now self-retry via
@@ -269,14 +269,24 @@ export function extractMarkdownLinks(content: string): { name: string; relTarget
 export function resolveSlug(fileDir: string, relTarget: string, allSlugs: Set<string>): string | null {
   const targetNoExt = relTarget.endsWith('.md') ? relTarget.slice(0, -3) : relTarget;
 
-  const s1 = join(fileDir, targetNoExt);
-  if (allSlugs.has(s1)) return s1;
+  // Issue #1964: wikilinks carry raw Obsidian paths (`[[llm-wiki/entities/AI 3.0]]`)
+  // but allSlugs holds sync-slugified slugs (`llm-wiki/entities/ai-3.0`). Try the
+  // raw candidate first (back-compat), then the sync-consistent slugified form.
+  const hit = (candidate: string): string | null => {
+    if (allSlugs.has(candidate)) return candidate;
+    const slugified = slugifyPath(candidate);
+    if (slugified !== candidate && allSlugs.has(slugified)) return slugified;
+    return null;
+  };
+
+  const s1 = hit(join(fileDir, targetNoExt));
+  if (s1) return s1;
 
   const parts = fileDir.split('/').filter(Boolean);
   for (let strip = 1; strip <= parts.length; strip++) {
     const ancestor = parts.slice(0, parts.length - strip).join('/');
-    const candidate = ancestor ? join(ancestor, targetNoExt) : targetNoExt;
-    if (allSlugs.has(candidate)) return candidate;
+    const candidate = hit(ancestor ? join(ancestor, targetNoExt) : targetNoExt);
+    if (candidate) return candidate;
   }
 
   return null;
@@ -1743,15 +1753,7 @@ async function extractStaleFromDB(
       // `page.updated_at.toISOString()` — the JS Date is ms-truncated, so the
       // µs-precision DB updated_at stayed strictly greater and the page never
       // cleared on Postgres. Stamping the exact value makes them equal.
-      //
-      // Version-arm floor: a page last edited BEFORE LINK_EXTRACTOR_VERSION_TS
-      // would otherwise be stamped below the version watermark and stay
-      // permanently stale (`links_extracted_at < versionTs` re-fires every run).
-      // Stamp max(updated_at, versionTs) — versionTs is always a past release
-      // date, so a concurrent edit's now() still exceeds the stamp and D4 holds.
-      // Tie at ms precision picks updated_at_iso (its µs ≥ versionTs's .000000).
-      const stampTs = new Date(page.updated_at_iso) >= new Date(versionTs) ? page.updated_at_iso : versionTs;
-      processedRefs.push({ slug: page.slug, source_id: page.source_id, extractedAt: stampTs });
+      processedRefs.push({ slug: page.slug, source_id: page.source_id, extractedAt: page.updated_at_iso });
     }
 
     // Flush NON-swallowing (CDX-4): a throw here propagates out of the sweep so

src/core/link-extraction.ts

@@ -14,6 +14,7 @@
 import type { BrainEngine } from './engine.ts';
 import type { PageType } from './types.ts';
 import { ensureWellFormed } from './text-safe.ts';
+import { slugifyPath } from './sync.ts';
 
 /**
  * v0.42.7 — link-extraction version stamp. Bump this ISO timestamp whenever the
@@ -28,7 +29,7 @@ import { ensureWellFormed } from './text-safe.ts';
  * OR updated_at > links_extracted_at`. It is an ISO-8601 string (NOT a number) —
  * the column is TIMESTAMPTZ and the predicate binds it as `::timestamptz`.
  */
-export const LINK_EXTRACTOR_VERSION_TS = '2026-07-21T00:00:00Z';
+export const LINK_EXTRACTOR_VERSION_TS = '2026-05-31T00:00:00Z';
 
 // ─── Entity references ──────────────────────────────────────────
 
@@ -80,10 +81,10 @@ export type LinkResolutionType = 'qualified' | 'unqualified';
  * Directory prefix whitelist. These are the top-level slug dirs the extractor
  * recognizes as entity references. Upstream canonical + our extensions:
  *   - Gbrain canonical: people, companies, meetings, concepts, deal, civic, project, source, media, yc, projects
- *   - Our domain extensions: tech, finance, personal, openclaw, ops (domain-organized wikis)
+ *   - Our domain extensions: tech, finance, personal, openclaw (domain-organized wikis)
  *   - Our entity prefix: entities (we kept some legacy entities/projects/ pages)
  */
-const DIR_PATTERN = '(?:people|companies|meetings|concepts|deal|civic|project|projects|source|media|yc|tech|finance|personal|openclaw|entities|ops)';
+const DIR_PATTERN = '(?:people|companies|meetings|concepts|deal|civic|project|projects|source|media|yc|tech|finance|personal|openclaw|entities)';
 
 /**
  * Match `[Name](path)` markdown links pointing to entity directories.
@@ -482,14 +483,29 @@ export async function extractPageLinks(
     // pre-v0.40.8.2 behavior of dropping bare wikilinks outside
     // DIR_PATTERN.
     if (ref.needsResolution) {
-      if (!opts.globalBasename || typeof resolver.resolveBasenameMatches !== 'function') {
-        continue;
-      }
+      if (typeof resolver.resolveBasenameMatches !== 'function') continue;
       // Issue #972 (codex): resolve by the wikilink TARGET (ref.slug — the
       // text inside `[[...]]` before any `|`), NOT the display alias
       // (ref.name = match[2]). `[[struktura|the project]]` must resolve
       // `struktura`, not "the project". The display text is for context only.
-      const matches = await resolver.resolveBasenameMatches(ref.slug);
+      //
+      // Issue #1964: a dir-qualified wikilink (`[[llm-wiki/entities/AI 3.0]]`)
+      // carries a raw Obsidian path while page slugs are sync-slugified
+      // (`llm-wiki/entities/ai-3.0`). Slugify the path the same way sync does,
+      // then match by exact slug or path-suffix (wiki-root-relative authoring).
+      // This runs regardless of global_basename — it's dir-qualified, so the
+      // cross-dir false-positive risk the flag guards against doesn't apply.
+      // Mirrors the FS path's resolveSlug ancestor walk. Bare `[[name]]`
+      // wikilinks still require the global_basename flag.
+      let matches: string[] = [];
+      const slugified = ref.slug.includes('/') ? slugifyPath(ref.slug) : '';
+      if (slugified.includes('/')) {
+        const tail = slugified.slice(slugified.lastIndexOf('/') + 1);
+        matches = (await resolver.resolveBasenameMatches(tail))
+          .filter(m => m === slugified || m.endsWith(`/${slugified}`));
+      } else if (opts.globalBasename) {
+        matches = await resolver.resolveBasenameMatches(ref.slug);
+      }
       if (matches.length === 0) continue;
       const idx = content.indexOf(ref.slug);
       const context = idx >= 0 ? excerpt(content, idx, 240) : ref.name;
@@ -865,16 +881,7 @@ export function queryBasenameIndex(idx: Map<string, string[]>, name: string): st
   if (!name || typeof name !== 'string') return [];
   const trimmed = name.trim();
   if (!trimmed) return [];
-  let hit = idx.get(trimmed) ?? idx.get(trimmed.toLowerCase()) ?? idx.get(normalizeBasename(trimmed));
-  // Issue #2576 bug 2: path-style refs (`runbooks/2026-05-01-x`) from dirs
-  // outside DIR_PATTERN reach here, but normalizeBasename strips slashes
-  // into a garbage key (`runbooks2026-05-01-x`) that can never hit the
-  // tail-keyed index. Fall back to the path tail so qualified refs resolve
-  // by basename like everything else.
-  if (!hit && trimmed.includes('/')) {
-    const tail = trimmed.slice(trimmed.lastIndexOf('/') + 1).trim();
-    if (tail) hit = idx.get(tail) ?? idx.get(tail.toLowerCase()) ?? idx.get(normalizeBasename(tail));
-  }
+  const hit = idx.get(trimmed) ?? idx.get(trimmed.toLowerCase()) ?? idx.get(normalizeBasename(trimmed));
   return hit ? [...hit].sort(basenameSort) : [];
 }
 

test/extract-fs.test.ts

@@ -389,6 +389,33 @@ describe('resolveSlugAll', () => {
   });
 });
 
+// ─── issue #1964: cross-directory wikilinks — slug/path mismatch ──────────
+
+describe('issue #1964: raw Obsidian wikilink paths resolve to sync-slugified slugs', () => {
+  test('resolveSlug slugifies the candidate (sync-consistent), no flag needed', () => {
+    const all = new Set(['llm-wiki/entities/ai-3.0']);
+    // Wikilink literal `[[llm-wiki/entities/AI 3.0]]` — spaces + uppercase.
+    expect(resolveSlug('llm-wiki/notes', 'llm-wiki/entities/AI 3.0.md', all))
+      .toBe('llm-wiki/entities/ai-3.0');
+  });
+
+  test('resolveSlug slugifies raw (unslugified) fileDir too', () => {
+    const all = new Set(['llm-wiki/entities/ai-3.0']);
+    // fileDir comes from dirname(relPath) — the raw on-disk directory.
+    expect(resolveSlug('LLM Wiki/Notes', 'entities/AI 3.0.md', all))
+      .toBe('llm-wiki/entities/ai-3.0');
+  });
+
+  test('extractLinksFromFile resolves cross-directory wikilink with flag OFF as a typed edge', async () => {
+    const allSlugs = new Set(['llm-wiki/entities/ai-3.0', 'llm-wiki/notes/roadmap']);
+    const content = '---\ntitle: Roadmap\ntype: concept\n---\n\nSee [[llm-wiki/entities/AI 3.0]].\n';
+    const links = await extractLinksFromFile(content, 'llm-wiki/notes/roadmap.md', allSlugs);
+    expect(links.map(l => l.to_slug)).toEqual(['llm-wiki/entities/ai-3.0']);
+    // Dir-qualified path resolution is exact, NOT the basename fallback.
+    expect(links[0].link_type).not.toBe('wikilink_basename');
+  });
+});
+
 describe('issue #972 repro: bare wikilinks resolve when flag is on', () => {
   // End-to-end: reproduces the issue's exact repro inside a tempdir +
   // PGLite, then asserts edge count under both flag states.

test/extract-stale.test.ts

@@ -209,24 +209,6 @@ describe('gbrain extract --stale', () => {
     expect(usRows[0]?.eq).toBe(true);
   });
 
-  test('version-arm floor: page edited BEFORE LINK_EXTRACTOR_VERSION_TS clears after --stale (issue #2576 bug 3)', async () => {
-    // A page whose updated_at predates the version watermark used to be
-    // stamped at its updated_at (< versionTs), so the version arm re-fired
-    // every run — permanently stale. The sweep now floors the stamp at
-    // versionTs. (The #1768 test above also covers this since the v0.42.x
-    // VERSION_TS bump moved its date below the watermark, but this pins the
-    // behavior explicitly so a date "repair" there can't drop coverage.)
-    await engine.putPage('people/alice', personPage('Alice'));
-    await engine.executeRaw(`UPDATE pages SET updated_at = '2000-01-01T00:00:00Z' WHERE slug = 'people/alice'`);
-    expect(await engine.countStalePagesForExtraction({ versionTs: LINK_EXTRACTOR_VERSION_TS })).toBe(1);
-
-    await runExtract(engine, ['--stale']);
-    // Pre-floor this stayed 1 forever (stamp < versionTs → version arm re-fires).
-    expect(await engine.countStalePagesForExtraction({ versionTs: LINK_EXTRACTOR_VERSION_TS })).toBe(0);
-    await runExtract(engine, ['--stale']);
-    expect(await engine.countStalePagesForExtraction({ versionTs: LINK_EXTRACTOR_VERSION_TS })).toBe(0);
-  });
-
   test('CDX-4 (D2): a link-flush throw aborts the sweep and leaves pages UNSTAMPED', async () => {
     await engine.putPage('people/alice', personPage('Alice'));
     await engine.putPage('companies/acme', companyPage('Acme', '[Alice](people/alice) founded [Acme](companies/acme).'));

test/link-extraction.test.ts

@@ -9,6 +9,8 @@ import {
   parseTimelineEntries,
   isAutoLinkEnabled,
   FRONTMATTER_LINK_MAP,
+  buildBasenameIndex,
+  queryBasenameIndex,
   type SlugResolver,
 } from '../src/core/link-extraction.ts';
 import type { BrainEngine } from '../src/core/engine.ts';
@@ -140,15 +142,6 @@ describe('extractEntityRefs', () => {
     expect(wikiRefs[0].needsResolution).toBe(true);
   });
 
-  test('recognizes ops/ qualified wikilinks (issue #2576 bug 2)', () => {
-    // `ops` was missing from DIR_PATTERN, so [[ops/...]] fell through to
-    // the generic 2c pass (needsResolution) instead of being a real ref.
-    const refs = extractEntityRefs('Deployed via [[ops/services/pointer-agent]].');
-    expect(refs.length).toBe(1);
-    expect(refs[0].slug).toBe('ops/services/pointer-agent');
-    expect(refs[0].needsResolution).toBeUndefined();
-  });
-
   test('skips qualified-syntax tokens (those belong to 2a)', () => {
     // [[wiki:topics/ai]] looks like 2a's qualified shape — even though
     // it wouldn't satisfy DIR_PATTERN, 2c must not claim it either
@@ -433,6 +426,46 @@ describe('extractPageLinks', () => {
     expect(strk!.linkType).toBe('wikilink_basename');
   });
 
+  // ─── issue #1964: dir-qualified wikilinks with raw Obsidian paths ────────
+
+  test('#1964: dir-qualified wikilink resolves via sync-consistent slugification (flag OFF)', async () => {
+    // `[[llm-wiki/entities/AI 3.0]]` is a raw Obsidian path; the page slug
+    // is the sync-slugified `llm-wiki/entities/ai-3.0`. Must resolve WITHOUT
+    // global_basename (it's dir-qualified) and must NOT leak to a same-tail
+    // page in a different directory.
+    const resolver: SlugResolver = {
+      resolve: async () => null,
+      resolveBasenameMatches: async (name) =>
+        name === 'ai-3.0' ? ['other/ai-3.0', 'llm-wiki/entities/ai-3.0'] : [],
+    };
+    const { candidates } = await extractPageLinks(
+      'llm-wiki/notes/roadmap',
+      'See [[llm-wiki/entities/AI 3.0]] for the model.',
+      {}, 'concept', resolver,
+      // opts.globalBasename omitted (= false) — path is dir-qualified
+    );
+    expect(candidates.map(c => c.targetSlug)).toEqual(['llm-wiki/entities/ai-3.0']);
+    expect(candidates[0].linkType).toBe('wikilink_basename');
+    expect(candidates[0].linkSource).toBe('wikilink-resolved');
+  });
+
+  test('#1964: path-suffix match resolves wiki-root-relative paths against a real index', async () => {
+    // Author writes `[[llm-wiki/entities/AI 3.0]]` but the brain nests the
+    // wiki under a vault dir. Suffix match rescues it; queried through the
+    // REAL basename index so the tail-key lookup is exercised end to end.
+    const idx = buildBasenameIndex(['vault/llm-wiki/entities/ai-3.0', 'people/ai-3.0']);
+    const resolver: SlugResolver = {
+      resolve: async () => null,
+      resolveBasenameMatches: async (name) => queryBasenameIndex(idx, name),
+    };
+    const { candidates } = await extractPageLinks(
+      'vault/llm-wiki/notes/roadmap',
+      'See [[llm-wiki/entities/AI 3.0]].',
+      {}, 'concept', resolver,
+    );
+    expect(candidates.map(c => c.targetSlug)).toEqual(['vault/llm-wiki/entities/ai-3.0']);
+  });
+
   test('opts.skipFrontmatter suppresses the frontmatter pass', async () => {
     // Real resolver shape that WOULD resolve frontmatter source: too,
     // but skipFrontmatter blocks the path entirely.
@@ -1078,15 +1111,6 @@ describe('makeResolver — fallback chain', () => {
     ]);
   });
 
-  test('resolveBasenameMatches: path-style ref falls back to the tail (issue #2576 bug 2)', async () => {
-    // normalizeBasename strips slashes, so `runbooks/2026-05-01-pointer-agent`
-    // used to normalize to a garbage key that never hit the tail-keyed index.
-    const engine = makeFakeEngineWithSlugs(['ops/changes/2026-05-01-pointer-agent']);
-    const r = makeResolver(engine);
-    expect(await r.resolveBasenameMatches!('runbooks/2026-05-01-pointer-agent'))
-      .toEqual(['ops/changes/2026-05-01-pointer-agent']);
-  });
-
   test('resolveBasenameMatches: case-insensitive fallback', async () => {
     const engine = makeFakeEngineWithSlugs(['companies/fast-weigh']);
     const r = makeResolver(engine);