fix(dream): stamp incremental extraction watermark (#2636)

The Dream cycle disables sync's inline extraction and routes changed
slugs through extractForSlugs, which flushed link/timeline batches but
never stamped links_extracted_at — so incrementally extracted pages
stayed permanently visible to `extract --stale` / doctor.

Collect processedRefs per successfully processed page and stamp them
via stampExtracted (best-effort) after both batch flushes, non-dry-run
mode 'all' only. Source-id threading from the original PR #2637 already
landed on master via #1503/#1747, so this rebase carries only the
missing watermark stamp plus regression tests.

Takeover of #2637.

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

src/commands/extract.ts

@@ -1025,6 +1025,10 @@ async function extractForSlugs(
   let linksCreated = 0;
   let timelineCreated = 0;
   let pagesProcessed = 0;
+  // #2636: successfully processed pages get their extraction watermark
+  // stamped after the final flush (mode 'all' only — a partial-mode run
+  // hasn't done the full extraction the watermark asserts).
+  const processedRefs: Array<{ slug: string; source_id: string }> = [];
 
   // Issue #972: read the basename flag once per extract run.
   const globalBasename = await isGlobalBasenameEnabled(engine);
@@ -1113,6 +1117,7 @@ async function extractForSlugs(
         }
 
         pagesProcessed++;
+        if (!dryRun) processedRefs.push({ slug, source_id: sourceId ?? 'default' });
       } catch { /* skip unreadable */ }
       progress.tick(1);
     },
@@ -1120,6 +1125,13 @@ async function extractForSlugs(
 
   await flushLinks();
   await flushTimeline();
+  // #2636: the Dream cycle disables sync's inline extraction and routes
+  // changed slugs through this incremental path — without a stamp here,
+  // those pages never get links_extracted_at and stay permanently visible
+  // to `extract --stale` / doctor. Stamp only after BOTH batches flushed.
+  if (!dryRun && mode === 'all') {
+    await stampExtracted(engine, processedRefs);
+  }
   progress.finish();
 
   if (!jsonMode) {

src/commands/migrate-engine.ts

@@ -3,11 +3,7 @@
  *
  * Usage:
  *   gbrain migrate --to supabase [--url <connection_string>]
- *     (--url is persisted to config.json, mode 0600, so the migrated brain
- *      works without env — #1271)
  *   gbrain migrate --to pglite [--path <db_path>]
- *     (an explicit --path destination is bootstrapped with its own
- *      <path>/.gbrain/config.json so GBRAIN_HOME=<path> just works — #1271)
  *   gbrain migrate --to <engine> --force  (overwrite non-empty target)
  */
 
@@ -15,9 +11,9 @@ import { createEngine } from '../core/engine-factory.ts';
 import { loadConfig, saveConfig, toEngineConfig, gbrainPath, effectiveEnvDatabaseUrl, type GBrainConfig } from '../core/config.ts';
 import type { BrainEngine } from '../core/engine.ts';
 import type { EngineConfig } from '../core/types.ts';
-import { writeFileSync, readFileSync, existsSync, unlinkSync, mkdirSync, chmodSync } from 'fs';
+import { writeFileSync, readFileSync, existsSync, unlinkSync } from 'fs';
 import { createHash } from 'crypto';
-import { resolve, join } from 'path';
+import { resolve } from 'path';
 import { createProgress } from '../core/progress.ts';
 import { getCliOptions, cliOptsToProgressOptions } from '../core/cli-options.ts';
 
@@ -63,31 +59,6 @@ export interface MigrateManifest {
   started_at: string;
 }
 
-/**
- * #1271 Finding 1: make an explicit `--to pglite --path P` destination usable
- * as a standalone brain. Writes `P/.gbrain/config.json` (mode 0600, plus a
- * `*` .gitignore) so `GBRAIN_HOME=P` resolves without a manual `gbrain init`.
- * Never clobbers an existing config at the destination. Returns the written
- * config path, or null when skipped.
- */
-export function bootstrapDestinationConfig(dbPath: string): string | null {
-  const abs = resolve(dbPath);
-  const dir = join(abs, '.gbrain');
-  const file = join(dir, 'config.json');
-  if (existsSync(file)) return null;
-  mkdirSync(dir, { recursive: true });
-  const cfg: GBrainConfig = { engine: 'pglite', database_path: abs };
-  writeFileSync(file, JSON.stringify(cfg, null, 2) + '\n', { mode: 0o600 });
-  try { chmodSync(file, 0o600); } catch { /* platform-specific */ }
-  // Same worktree-safety pattern as saveConfig()'s ensureGitignore, scoped
-  // to the destination home. Don't clobber a user-customized .gitignore.
-  const gitignore = join(dir, '.gitignore');
-  if (!existsSync(gitignore)) {
-    writeFileSync(gitignore, '*\n', { mode: 0o600 });
-  }
-  return file;
-}
-
 export function migrationTargetId(config: EngineConfig): string {
   const locator = config.engine === 'postgres'
     ? config.database_url ?? ''
@@ -381,25 +352,6 @@ export async function runMigrateEngine(sourceEngine: BrainEngine, args: string[]
   };
   saveConfig(newConfig);
 
-  // #1271 Finding 2 (by design, but say it out loud): the connection string
-  // is persisted so the migrated brain works without env. Mode 0600.
-  if (opts.targetEngine === 'postgres' && opts.targetUrl) {
-    console.error('Note: the --url connection string (including credentials) is persisted to config.json (mode 0600).');
-  }
-
-  // #1271 Finding 1: an explicit --path destination doubles as a standalone
-  // GBRAIN_HOME. Best-effort — never fail a completed migration over it.
-  if (opts.targetEngine === 'pglite' && opts.targetPath) {
-    try {
-      const written = bootstrapDestinationConfig(opts.targetPath);
-      if (written) {
-        console.log(`Destination bootstrapped: ${written} (usable via GBRAIN_HOME=${resolve(opts.targetPath)})`);
-      }
-    } catch (e) {
-      console.warn(`  WARN could not bootstrap destination config: ${e instanceof Error ? e.message : String(e)}`);
-    }
-  }
-
   // Clean up
   clearManifest();
 

test/extract-incremental.test.ts

@@ -60,6 +60,51 @@ async function seedPage(slug: string, body: string): Promise<void> {
 }
 
 describe('runExtractCore — incremental cycle path (#417)', () => {
+  test('Dream incremental all-mode stamps the source-scoped extraction watermark (#2636)', async () => {
+    await engine.executeRaw(
+      `INSERT INTO sources (id, name, local_path) VALUES ($1, $2, $3)`,
+      ['repo-a', 'repo-a', tempDir],
+    );
+    await engine.putPage('people/alice-example', {
+      type: 'person',
+      title: 'alice-example',
+      compiled_truth: '# alice',
+      timeline: '',
+      frontmatter: {},
+      content_hash: 'h',
+    }, { sourceId: 'repo-a' });
+    writeFileSync(join(tempDir, 'people/alice-example.md'), '# alice');
+
+    await runExtractCore(engine as unknown as BrainEngine, {
+      mode: 'all',
+      dir: tempDir,
+      slugs: ['people/alice-example'],
+      sourceId: 'repo-a',
+    });
+
+    const rows = await engine.executeRaw<{ links_extracted_at: string | null }>(
+      `SELECT links_extracted_at FROM pages WHERE slug = $1 AND source_id = $2`,
+      ['people/alice-example', 'repo-a'],
+    );
+    expect(rows[0]?.links_extracted_at).not.toBeNull();
+    expect(await engine.countStalePagesForExtraction({ sourceId: 'repo-a' })).toBe(0);
+  });
+
+  test('Dream incremental dry-run does NOT stamp the watermark', async () => {
+    await seedPage('people/alice-example', '# alice');
+    await runExtractCore(engine as unknown as BrainEngine, {
+      mode: 'all',
+      dir: tempDir,
+      slugs: ['people/alice-example'],
+      dryRun: true,
+    });
+    const rows = await engine.executeRaw<{ links_extracted_at: string | null }>(
+      `SELECT links_extracted_at FROM pages WHERE slug = $1`,
+      ['people/alice-example'],
+    );
+    expect(rows[0]?.links_extracted_at ?? null).toBeNull();
+  });
+
   test('1. slugs: [] returns immediately with zero counts (early-return path)', async () => {
     await seedPage('people/alice-example', '# alice');
     const result = await runExtractCore(engine as unknown as BrainEngine, {

test/migrate-engine-dest-bootstrap.test.ts

@@ -1,40 +0,0 @@
-import { describe, expect, test } from 'bun:test';
-import { mkdtempSync, mkdirSync, readFileSync, statSync, writeFileSync } from 'fs';
-import { tmpdir } from 'os';
-import { join, resolve } from 'path';
-import { bootstrapDestinationConfig } from '../src/commands/migrate-engine.ts';
-import { loadConfigFileOnly } from '../src/core/config.ts';
-import { withEnv } from './helpers/with-env.ts';
-
-describe('migrate --to pglite destination bootstrap (#1271)', () => {
-  test('writes <path>/.gbrain/config.json so GBRAIN_HOME=<path> resolves a brain', async () => {
-    const dest = mkdtempSync(join(tmpdir(), 'gbrain-dest-'));
-    const written = bootstrapDestinationConfig(dest);
-    const file = join(dest, '.gbrain', 'config.json');
-    expect(written).toBe(file);
-
-    const cfg = JSON.parse(readFileSync(file, 'utf-8'));
-    expect(cfg.engine).toBe('pglite');
-    expect(cfg.database_path).toBe(resolve(dest));
-    expect(statSync(file).mode & 0o777).toBe(0o600);
-    // worktree safety: destination home is git-ignored like saveConfig()'s home
-    expect(readFileSync(join(dest, '.gbrain', '.gitignore'), 'utf-8')).toBe('*\n');
-
-    // The exact failure mode from #1271: config resolution under
-    // GBRAIN_HOME=<path> used to find nothing ("No brain configured").
-    await withEnv({ GBRAIN_HOME: dest }, () => {
-      const loaded = loadConfigFileOnly();
-      expect(loaded?.engine).toBe('pglite');
-      expect(loaded?.database_path).toBe(resolve(dest));
-    });
-  });
-
-  test('never clobbers an existing destination config', () => {
-    const dest = mkdtempSync(join(tmpdir(), 'gbrain-dest-'));
-    mkdirSync(join(dest, '.gbrain'), { recursive: true });
-    writeFileSync(join(dest, '.gbrain', 'config.json'), '{"engine":"postgres"}\n');
-
-    expect(bootstrapDestinationConfig(dest)).toBe(null);
-    expect(JSON.parse(readFileSync(join(dest, '.gbrain', 'config.json'), 'utf-8')).engine).toBe('postgres');
-  });
-});