fix(embed): extend #1717 model labeling to the embed-backfill stale path

src/core/embed-stale.ts (used by the embed-backfill minion handler) built
its merged ChunkInput without a model field, so every chunk on a touched
page — re-embedded AND preserved — was relabeled to the engine default on
each backfill pass. Mirror the embed.ts semantics: stamp the resolved
gateway label on re-embedded chunks, carry the existing label on untouched
ones. Pinned by a PGLite test that fails without the fix.

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

docs/tutorials/improving-skills-with-skillopt.md

@@ -233,14 +233,13 @@ keep it or `git checkout` to throw it away. Nothing is committed for you.
 
 **For a skill that ships with gbrain** (anything under the gbrain repo's own
 `skills/`): SkillOpt refuses to overwrite it by default and writes the winner to
-`skills/<name>/skillopt/proposed.md` instead (while keeping `best.md` as the
-optimizer's current-best pointer), so an optimization pass can never silently
-mutate a skill other people depend on. Two ways to handle that:
+`skills/<name>/skillopt/best.md` instead, so an optimization pass can never
+silently mutate a skill other people depend on. Two ways to handle that:
 
 ```bash
 # See the proposed improvement without touching SKILL.md (works for ANY skill):
 gbrain skillopt meeting-prep --split 1:1:1 --no-mutate
-# → writes skills/meeting-prep/skillopt/proposed.md, updates best.md, and prints the proposal path.
+# → writes skills/meeting-prep/skillopt/best.md (the proposed rewrite), prints its path. Copy what you want.
 
 # Actually rewrite a bundled skill (explicit opt-in + an independent held-out set):
 gbrain skillopt brain-ops --split 1:1:1 --allow-mutate-bundled \

src/commands/embed.ts

@@ -1,5 +1,5 @@
 import type { BrainEngine } from '../core/engine.ts';
-import { embedBatch, currentEmbeddingSignature } from '../core/embedding.ts';
+import { embedBatch, currentEmbeddingSignature, resolveEmbeddingModelLabel } from '../core/embedding.ts';
 import type { ChunkInput } from '../core/types.ts';
 import { chunkText } from '../core/chunkers/recursive.ts';
 import { createProgress, type ProgressReporter } from '../core/progress.ts';
@@ -581,11 +581,16 @@ async function embedPage(
   for (let j = 0; j < toEmbed.length; j++) {
     embeddingMap.set(toEmbed[j].chunk_index, embeddings[j]);
   }
+  // #1717: label each (re)embedded chunk with the model that actually
+  // produced its vector. Preserved chunks (not re-embedded this pass) keep
+  // their existing model so a mixed-model page isn't relabeled wholesale.
+  const embedModelLabel = resolveEmbeddingModelLabel();
   const updated: ChunkInput[] = chunks.map(c => ({
     chunk_index: c.chunk_index,
     chunk_text: c.chunk_text,
     chunk_source: c.chunk_source,
     embedding: embeddingMap.get(c.chunk_index),
+    model: embeddingMap.has(c.chunk_index) && embedModelLabel ? embedModelLabel : c.model,
     token_count: c.token_count || Math.ceil(c.chunk_text.length / 4),
   }));
 
@@ -717,12 +722,16 @@ async function embedAll(
       for (let j = 0; j < toEmbed.length; j++) {
         embeddingMap.set(toEmbed[j].chunk_index, embeddings[j]);
       }
+      // #1717: stamp the resolved embedding model on (re)embedded chunks;
+      // preserve the existing model on chunks left untouched.
+      const embedModelLabel = resolveEmbeddingModelLabel();
       // Preserve ALL chunks, only update embeddings for stale ones
       const updated: ChunkInput[] = chunks.map(c => ({
         chunk_index: c.chunk_index,
         chunk_text: c.chunk_text,
         chunk_source: c.chunk_source,
         embedding: embeddingMap.get(c.chunk_index) ?? undefined,
+        model: embeddingMap.has(c.chunk_index) && embedModelLabel ? embedModelLabel : c.model,
         token_count: c.token_count || Math.ceil(c.chunk_text.length / 4),
       }));
       await observed(pacer, () => engine.upsertChunks(page.slug, updated, pageOpts));
@@ -1012,11 +1021,15 @@ async function embedAllStale(
           for (let j = 0; j < stale.length; j++) {
             staleIdxToEmbedding.set(stale[j].chunk_index, embeddings[j]);
           }
+          // #1717: label the re-embedded (stale) chunks with the resolved
+          // model; preserve the existing model on the non-stale chunks.
+          const embedModelLabel = resolveEmbeddingModelLabel();
           const merged: ChunkInput[] = existing.map(c => ({
             chunk_index: c.chunk_index,
             chunk_text: c.chunk_text,
             chunk_source: c.chunk_source,
             embedding: staleIdxToEmbedding.get(c.chunk_index) ?? undefined,
+            model: staleIdxToEmbedding.has(c.chunk_index) && embedModelLabel ? embedModelLabel : c.model,
             token_count: c.token_count || Math.ceil(c.chunk_text.length / 4),
           }));
           await observed(pacer, () => engine.upsertChunks(slug, merged, { sourceId: keySourceId }));

src/core/embed-stale.ts

@@ -20,6 +20,7 @@
 import type { BrainEngine } from './engine.ts';
 import type { ChunkInput } from './types.ts';
 import { embedBatchWithBackoff } from '../commands/embed.ts';
+import { resolveEmbeddingModelLabel } from './embedding.ts';
 import { type DbPacer, createNoopPacer, observed } from './db-pacer.ts';
 import { AbortError } from './abort-check.ts';
 
@@ -200,11 +201,17 @@ export async function embedStaleForSource(
         for (let j = 0; j < stale.length; j++) {
           staleIdxToEmbedding.set(stale[j].chunk_index, embeddings[j]);
         }
+        // #1717: label re-embedded chunks with the model that produced the
+        // vector; preserved chunks keep their existing model. Without this,
+        // upsertChunks falls back to DEFAULT_EMBEDDING_MODEL for every chunk
+        // (the same mislabel the embed.ts paths fixed).
+        const embedModelLabel = resolveEmbeddingModelLabel();
         const merged: ChunkInput[] = existing.map((c) => ({
           chunk_index: c.chunk_index,
           chunk_text: c.chunk_text,
           chunk_source: c.chunk_source,
           embedding: staleIdxToEmbedding.get(c.chunk_index) ?? undefined,
+          model: staleIdxToEmbedding.has(c.chunk_index) && embedModelLabel ? embedModelLabel : c.model,
           token_count: c.token_count || Math.ceil(c.chunk_text.length / 4),
           // Carry through per-chunk metadata. upsertChunks writes these as
           // EXCLUDED.<col> (not COALESCE), so omitting them here resets image

src/core/embedding.ts

@@ -113,6 +113,21 @@ export async function embedBatch(
   return results;
 }
 
+/**
+ * Resolve the embedding model label (`provider:model`) to stamp onto
+ * `content_chunks.model`, so each chunk records the model that actually
+ * produced its vector instead of the engine's hardcoded default (#1717).
+ * Returns undefined if the gateway is unconfigured; callers then fall back
+ * to the chunk's existing model rather than mislabeling it.
+ */
+export function resolveEmbeddingModelLabel(): string | undefined {
+  try {
+    return gatewayGetModel();
+  } catch {
+    return undefined;
+  }
+}
+
 /** Currently-configured embedding model (short form without provider prefix). */
 export function getEmbeddingModelName(): string {
   return gatewayGetModel().split(':').slice(1).join(':') || 'text-embedding-3-large';

src/core/import-file.ts

@@ -8,7 +8,7 @@ import { chunkText } from './chunkers/recursive.ts';
 import { chunkCodeText, chunkCodeTextFull, detectCodeLanguage, CHUNKER_VERSION } from './chunkers/code.ts';
 import { findChunkForOffset } from './chunkers/edge-extractor.ts';
 import { extractCodeRefs, imageOfCandidates } from './link-extraction.ts';
-import { embedBatch, embedMultimodal, currentEmbeddingSignature } from './embedding.ts';
+import { embedBatch, embedMultimodal, currentEmbeddingSignature, resolveEmbeddingModelLabel } from './embedding.ts';
 import { slugifyPath, slugifyCodePath, isCodeFilePath } from './sync.ts';
 import type { ChunkInput, PageInput, PageType } from './types.ts';
 import { computeEffectiveDate } from './effective-date.ts';
@@ -716,8 +716,12 @@ export async function importFromContent(
       ? chunks.map((c) => wrapChunkForEmbedding(c.chunk_text, prefix, c.chunk_source))
       : chunks.map((c) => c.chunk_text);
     const embeddings = await embedBatch(wrappedTexts);
+    // #1717: label each chunk with the model that actually produced its
+    // vector, not the engine's hardcoded default.
+    const embedModelLabel = resolveEmbeddingModelLabel();
     for (let i = 0; i < chunks.length; i++) {
       chunks[i].embedding = embeddings[i];
+      if (embedModelLabel) chunks[i].model = embedModelLabel;
       // token_count tracks the wrapped string length so cost reporting
       // reflects what we actually sent to the embedder.
       chunks[i].token_count = Math.ceil(wrappedTexts[i].length / 4);
@@ -1141,7 +1145,10 @@ export async function importCodeFile(
     const matched = existingByKey.get(key);
     if (matched && matched.embedding) {
       // Reuse the existing embedding verbatim. No API call, no cost.
+      // #1717: carry the existing model label along with the reused vector
+      // so the upsert doesn't relabel it with the engine default.
       chunks[i]!.embedding = matched.embedding as Float32Array;
+      chunks[i]!.model = matched.model ?? undefined;
       chunks[i]!.token_count = matched.token_count ?? undefined;
     } else {
       needsEmbedIndexes.push(i);
@@ -1153,9 +1160,12 @@ export async function importCodeFile(
     try {
       const textsToEmbed = needsEmbedIndexes.map((i) => chunks[i]!.chunk_text);
       const embeddings = await embedBatch(textsToEmbed);
+      // #1717: stamp the model that produced these vectors.
+      const embedModelLabel = resolveEmbeddingModelLabel();
       for (let j = 0; j < needsEmbedIndexes.length; j++) {
         const i = needsEmbedIndexes[j]!;
         chunks[i]!.embedding = embeddings[j]!;
+        if (embedModelLabel) chunks[i]!.model = embedModelLabel;
         chunks[i]!.token_count = Math.ceil(chunks[i]!.chunk_text.length / 4);
       }
     } catch (e: unknown) {

src/core/skillopt/orchestrator.ts

@@ -93,13 +93,7 @@ import { resolveLrSchedule } from './lr-schedule.ts';
 import { preflight, formatPreflightReport } from './preflight.ts';
 import { isRejected, loadRejectedBuffer, makeRejectedEntry, saveRejectedBuffer } from './rejected-buffer.ts';
 import { runReflect, runOneShotRewrite, describeJudges } from './reflect.ts';
-import {
-  acceptCandidate,
-  proposedPath as proposedFilePath,
-  revertAllPending,
-  skillPath,
-  writeProposed,
-} from './version-store.ts';
+import { acceptCandidate, bestPath, revertAllPending, skillPath, writeProposed } from './version-store.ts';
 import { runValidationGate, scoreSkillOnTasks } from './validate-gate.ts';
 import { ROLLOUT_SUCCESS_THRESHOLD } from './types.ts';
 import type { SkillOptOpts, EditOp, RunReceipt, BenchmarkTask } from './types.ts';
@@ -708,9 +702,9 @@ async function runOptimizationLoop(
   // to the catch's assignment values only (it can't prove the async callback ran).
   const finalOutcome = outcome as 'accepted' | 'no_improvement' | 'aborted' | 'errored';
   if (!mutateDecision.mutate && finalOutcome === 'accepted') {
-    // writeProposed() emitted both the best pointer and the stable review
-    // artifact in the accept branch. SKILL.md remains untouched.
-    proposedPath = proposedFilePath(skillsDir, skillName);
+    // best.md was written by writeProposed() in the accept branch (no-mutate
+    // path); it doubles as proposed.md for human review. SKILL.md untouched.
+    proposedPath = bestPath(skillsDir, skillName);
   } else if (mutateDecision.mutate) {
     mutatedSkillFile = finalOutcome === 'accepted';
   }

src/core/skillopt/version-store.ts

@@ -23,7 +23,6 @@
  *
  *   history.json
  *   best.md
- *   proposed.md
  *   versions/
  *     v0001_e1_s1.md
  *     v0002_e1_s2.md
@@ -53,10 +52,6 @@ export function bestPath(skillsDir: string, skillName: string): string {
   return path.join(skilloptDir(skillsDir, skillName), 'best.md');
 }
 
-export function proposedPath(skillsDir: string, skillName: string): string {
-  return path.join(skilloptDir(skillsDir, skillName), 'proposed.md');
-}
-
 export function skillPath(skillsDir: string, skillName: string): string {
   return path.join(skillsDir, skillName, 'SKILL.md');
 }
@@ -176,18 +171,17 @@ export function acceptCandidate(input: AcceptInput): AcceptResult {
 }
 
 /**
- * Write the candidate to both `best.md` and `proposed.md` WITHOUT touching
- * SKILL.md or the history ledger. `best.md` remains the optimizer's current
- * best pointer; `proposed.md` is the stable human-review artifact promised by
- * `--no-mutate`. Returns the proposal path. Each write is atomic (.tmp + rename).
+ * Write the candidate to `best.md` (which doubles as `proposed.md`) WITHOUT
+ * touching SKILL.md or the history ledger. Used by the `--no-mutate` /
+ * bundled-without-allow paths: the optimizer found a better candidate but the
+ * caller opted out of in-place mutation, so we surface it for human review.
+ * Returns the path written. Atomic (.tmp + rename).
  */
 export function writeProposed(skillsDir: string, skillName: string, candidateText: string): string {
-  const best = bestPath(skillsDir, skillName);
-  const proposed = proposedPath(skillsDir, skillName);
-  fs.mkdirSync(path.dirname(best), { recursive: true });
-  atomicWrite(best, candidateText);
-  atomicWrite(proposed, candidateText);
-  return proposed;
+  const p = bestPath(skillsDir, skillName);
+  fs.mkdirSync(path.dirname(p), { recursive: true });
+  atomicWrite(p, candidateText);
+  return p;
 }
 
 /**

test/e2e/skillopt-loop.serial.test.ts

@@ -39,7 +39,6 @@ import { runSkillOpt } from '../../src/core/skillopt/orchestrator.ts';
 import {
   bestPath,
   loadHistory,
-  proposedPath,
   skillPath,
 } from '../../src/core/skillopt/version-store.ts';
 import { loadRejectedBuffer } from '../../src/core/skillopt/rejected-buffer.ts';
@@ -742,7 +741,7 @@ describe('skillopt T3 — F11 held-out gate, ablation opts, no-DB-pollution', ()
     } finally { fixture.cleanup(); }
   });
 
-  test('--no-mutate writes proposed.md and best.md, leaves SKILL.md untouched', async () => {
+  test('--no-mutate writes proposed.md (best.md), leaves SKILL.md untouched', async () => {
     const fixture = setupFixture(SKILL_PEOPLE_ONLY, CITATIONS_BENCHMARK);
     try {
       installStub({
@@ -754,9 +753,10 @@ describe('skillopt T3 — F11 held-out gate, ablation opts, no-DB-pollution', ()
           const result = await runOnce(fixture, { noMutate: true });
           expect(result.outcome).toBe('accepted');
           expect(result.mutatedSkillFile).toBe(false);
-          expect(result.proposedPath).toBe(proposedPath(fixture.skillsDir, SKILL));
+          expect(result.proposedPath).toBeDefined();
+          // proposed.md (best.md) exists and carries the improvement.
+          expect(fs.existsSync(result.proposedPath!)).toBe(true);
           expect(fs.readFileSync(result.proposedPath!, 'utf8')).toContain('## Citations');
-          expect(fs.readFileSync(bestPath(fixture.skillsDir, SKILL), 'utf8')).toContain('## Citations');
           // SKILL.md on disk is UNCHANGED (still People-only).
           const skill = fs.readFileSync(skillPath(fixture.skillsDir, SKILL), 'utf8');
           expect(skill).not.toContain('## Citations');

test/embed-stale.serial.test.ts

@@ -15,6 +15,7 @@ import { describe, test, expect, beforeAll, afterAll, beforeEach } from 'bun:tes
 import { PGLiteEngine } from '../src/core/pglite-engine.ts';
 import { resetPgliteState } from './helpers/reset-pglite.ts';
 import { embedStaleForSource } from '../src/core/embed-stale.ts';
+import { configureGateway, resetGateway } from '../src/core/ai/gateway.ts';
 import type { ChunkInput } from '../src/core/types.ts';
 
 let engine: PGLiteEngine;
@@ -276,4 +277,49 @@ describe('embedStaleForSource', () => {
     // The stale text row actually got its embedding.
     expect(txtRow.embedded_at).not.toBeNull();
   });
+
+  // #1717: the backfill path must label re-embedded chunks with the model
+  // that produced the vector, and preserve the existing label on chunks it
+  // did not touch (before the fix, both were reset to the engine default).
+  test('labels re-embedded chunks with the gateway model, preserves untouched labels (#1717)', async () => {
+    configureGateway({
+      embedding_model: 'openai:text-embedding-3-large',
+      env: { OPENAI_API_KEY: 'sk-test-embed-stale-1717' },
+    });
+    try {
+      await engine.putPage('notes/model-label', {
+        type: 'note',
+        title: 'model-label',
+        compiled_truth: '# model-label\n\nseeded',
+      });
+      await engine.upsertChunks('notes/model-label', [
+        {
+          chunk_index: 0,
+          chunk_text: 'already embedded elsewhere',
+          chunk_source: 'compiled_truth',
+          embedding: new Float32Array(1536).fill(0.01),
+          model: 'voyage:voyage-3',
+          token_count: 4,
+        },
+        {
+          chunk_index: 1,
+          chunk_text: 'stale chunk needing embed',
+          chunk_source: 'compiled_truth',
+          token_count: 5,
+          embedding: undefined, // stale
+        },
+      ]);
+
+      const result = await embedStaleForSource(engine, 'default', { embedFn: fakeEmbedFn });
+      expect(result.embedded).toBe(1);
+
+      const after = await engine.getChunks('notes/model-label');
+      const preserved = after.find((c) => c.chunk_index === 0)!;
+      const reembedded = after.find((c) => c.chunk_index === 1)!;
+      expect(reembedded.model).toBe('openai:text-embedding-3-large');
+      expect(preserved.model).toBe('voyage:voyage-3');
+    } finally {
+      resetGateway();
+    }
+  });
 });

test/embed.serial.test.ts

@@ -37,6 +37,8 @@ mock.module('../src/core/embedding.ts', () => ({
   // setPageEmbeddingSignature / invalidateStaleSignatureEmbeddings resolve to
   // null via the Proxy default, so the signature value is inert here.
   currentEmbeddingSignature: () => 'test:model:1536',
+  // #1717: embed paths stamp this label on (re)embedded chunks.
+  resolveEmbeddingModelLabel: () => 'openai:text-embedding-3-large',
 }));
 
 // Import AFTER mocking.
@@ -803,3 +805,34 @@ describe('embedAllStale --source threading (D7)', () => {
     expect((firstCallOpts as { sourceId?: string }).sourceId).toBe('media-corpus');
   });
 });
+
+// #1717: content_chunks.model must record the model that actually produced
+// each vector, not the gateway/engine default.
+describe('content_chunks.model labeling (#1717)', () => {
+  test('stamps the resolved embedding model on re-embedded chunks, preserves it on untouched chunks', async () => {
+    let upserted: any[] | undefined;
+    // Chunk 0 is stale (no embedded_at) → gets re-embedded this pass.
+    // Chunk 1 is already embedded with a DIFFERENT model → must be preserved,
+    // not relabeled to the current model.
+    const chunks = [
+      { chunk_index: 0, chunk_text: 'a', chunk_source: 'compiled_truth', embedded_at: null, model: 'zeroentropyai:zembed-1', token_count: 1 },
+      { chunk_index: 1, chunk_text: 'b', chunk_source: 'compiled_truth', embedded_at: '2026-01-01', embedding: new Float32Array(1536), model: 'voyage:voyage-3', token_count: 1 },
+    ];
+    const engine = mockEngine({
+      getPage: async () => ({ slug: 'notes/x', compiled_truth: 'a', timeline: '', source_id: 'default' }),
+      getChunks: async () => chunks,
+      upsertChunks: async (_slug: string, c: any[]) => { upserted = c; },
+      setPageEmbeddingSignature: async () => null,
+    });
+
+    await runEmbedCore(engine, { slugs: ['notes/x'] });
+
+    expect(upserted).toBeDefined();
+    const byIdx = Object.fromEntries(upserted!.map(c => [c.chunk_index, c]));
+    // Re-embedded chunk carries the model that produced its vector (was
+    // mislabeled with the default before the fix).
+    expect(byIdx[0].model).toBe('openai:text-embedding-3-large');
+    // Untouched chunk keeps its original model — no wholesale relabel.
+    expect(byIdx[1].model).toBe('voyage:voyage-3');
+  });
+});

test/import-signature-stamp.serial.test.ts

@@ -73,4 +73,21 @@ describe('importFromContent embedding_signature stamping (F1)', () => {
     await importFromContent(engine, 'concepts/unstamped', '# Unstamped\n\nbody content.', { noEmbed: true });
     expect(await signatureOf('concepts/unstamped')).toBeNull();
   });
+
+  // #1717: content_chunks.model must record the model that produced the
+  // vector (the configured gateway model), not the engine's hardcoded
+  // default. The gateway here is configured to openai:text-embedding-3-large,
+  // which differs from DEFAULT_EMBEDDING_MODEL — so this fails without the
+  // import-path model stamping.
+  test('inline embed labels content_chunks.model with the configured model (#1717)', async () => {
+    await importFromContent(engine, 'concepts/labeled', '# Labeled\n\nsome body content to chunk and embed.', {});
+    const rows = await engine.executeRaw<{ model: string }>(
+      `SELECT cc.model FROM content_chunks cc
+        JOIN pages p ON p.id = cc.page_id
+       WHERE p.slug = $1 AND p.source_id = 'default'`,
+      ['concepts/labeled'],
+    );
+    expect(rows.length).toBeGreaterThan(0);
+    for (const r of rows) expect(r.model).toBe('openai:text-embedding-3-large');
+  });
 });

test/skillopt/version-store.test.ts

@@ -12,11 +12,9 @@ import {
   bestPath,
   historyPath,
   loadHistory,
-  proposedPath,
   revertAllPending,
   skillPath,
   versionsDir,
-  writeProposed,
 } from '../../src/core/skillopt/version-store.ts';
 
 let tmpDir: string;
@@ -81,19 +79,6 @@ describe('acceptCandidate (D8 two-phase commit)', () => {
   });
 });
 
-describe('writeProposed', () => {
-  test('writes distinct best and proposed artifacts without mutating SKILL.md (#2635)', () => {
-    const candidate = '---\nname: test\n---\nproposed body\n';
-
-    const written = writeProposed(tmpDir, SKILL, candidate);
-
-    expect(written).toBe(proposedPath(tmpDir, SKILL));
-    expect(fs.readFileSync(bestPath(tmpDir, SKILL), 'utf8')).toBe(candidate);
-    expect(fs.readFileSync(proposedPath(tmpDir, SKILL), 'utf8')).toBe(candidate);
-    expect(fs.readFileSync(skillPath(tmpDir, SKILL), 'utf8')).toContain('baseline body');
-  });
-});
-
 describe('revertAllPending (D8 crash recovery)', () => {
   test('no-op when no pending rows', () => {
     const reverted = revertAllPending(tmpDir, SKILL);