Compare commits

..
Author SHA1 Message Date
06eeb2890b fix(skillopt): emit proposed.md in no-mutate mode (#2635)
writeProposed now writes both best.md (optimizer's current-best pointer)
and proposed.md (the stable review artifact documented by --no-mutate);
the orchestrator returns the real proposed.md path instead of aliasing
best.md. Fix lands in the shared helper so both accept-branch call sites
are covered.

Takeover of #2719.

Co-authored-by: RerankerGuo <RerankerGuo@users.noreply.github.com>
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-07-21 14:30:19 -07:00
8 changed files with 64 additions and 83 deletions
@@ -233,13 +233,14 @@ 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/best.md` instead, so an optimization pass can never
silently mutate a skill other people depend on. Two ways to handle that:
`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:
```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/best.md (the proposed rewrite), prints its path. Copy what you want.
# → writes skills/meeting-prep/skillopt/proposed.md, updates best.md, and prints the proposal path.
# Actually rewrite a bundled skill (explicit opt-in + an independent held-out set):
gbrain skillopt brain-ops --split 1:1:1 --allow-mutate-bundled \
+15 -33
View File
@@ -107,14 +107,6 @@ export interface EmbedOpts {
* runs lock every source in sorted order. dryRun skips it.
*/
singleFlight?: boolean;
/**
* #394: suppress human stdout summaries (the `[dry-run] Would embed ...` /
* `Embedded N chunks ...` slog lines). Set by structured-output callers —
* the cycle's embed phase (dream --json must keep stdout JSON-clean per
* docs/progress-events.md) reports counts via its own PhaseResult instead.
* Errors/warnings still go to stderr regardless.
*/
quiet?: boolean;
}
/**
@@ -261,7 +253,7 @@ export async function runEmbedCore(engine: BrainEngine, opts: EmbedOpts): Promis
for (const s of opts.slugs) {
if (isAborted(opts.signal)) break; // #1737: stop the per-slug loop on abort
try {
await embedPage(engine, s, !!opts.dryRun, result, opts.sourceId, opts.signal, opts.quiet);
await embedPage(engine, s, !!opts.dryRun, result, opts.sourceId, opts.signal);
} catch (e: unknown) {
serr(` Error embedding ${s}: ${e instanceof Error ? e.message : e}`);
}
@@ -355,7 +347,6 @@ export async function runEmbedCore(engine: BrainEngine, opts: EmbedOpts): Promis
catchUp: opts.catchUp,
pacer,
paceMaxConcurrency,
quiet: opts.quiet,
}, opts.signal);
} finally {
// E1: surface pacing telemetry (human + structured) when pacing was on.
@@ -385,7 +376,7 @@ export async function runEmbedCore(engine: BrainEngine, opts: EmbedOpts): Promis
return result;
}
if (opts.slug) {
await embedPage(engine, opts.slug, !!opts.dryRun, result, opts.sourceId, opts.signal, opts.quiet);
await embedPage(engine, opts.slug, !!opts.dryRun, result, opts.sourceId, opts.signal);
return result;
}
throw new Error('No embed target specified. Pass { slug }, { slugs }, { all }, or { stale }.');
@@ -530,7 +521,6 @@ async function embedPage(
result: EmbedResult,
sourceId?: string,
signal?: AbortSignal,
quiet?: boolean,
) {
const opts = sourceId ? { sourceId } : undefined;
const page = await engine.getPage(slug, opts);
@@ -575,7 +565,7 @@ async function embedPage(
result.skipped += chunks.length - toEmbed.length;
if (toEmbed.length === 0) {
if (!quiet) slog(`${slug}: all ${chunks.length} chunks already embedded`);
slog(`${slug}: all ${chunks.length} chunks already embedded`);
result.pages_processed++;
return;
}
@@ -612,7 +602,7 @@ async function embedPage(
}
result.embedded += toEmbed.length;
result.pages_processed++;
if (!quiet) slog(`${slug}: embedded ${toEmbed.length} chunks`);
slog(`${slug}: embedded ${toEmbed.length} chunks`);
}
async function embedAll(
@@ -630,8 +620,6 @@ async function embedAll(
pacer?: DbPacer;
/** Resolved concurrency cap (E-1: the worker count, no separate permit). */
paceMaxConcurrency?: number;
/** #394: suppress human stdout summaries (structured-output callers). */
quiet?: boolean;
},
signal?: AbortSignal,
) {
@@ -775,12 +763,10 @@ async function embedAll(
});
// Stdout summary preserved for scripts/tests that grep for counts.
if (!staleOpts?.quiet) {
if (dryRun) {
slog(`[dry-run] Would embed ${result.would_embed} chunks across ${pages.length} pages`);
} else {
slog(`Embedded ${result.embedded} chunks across ${pages.length} pages`);
}
if (dryRun) {
slog(`[dry-run] Would embed ${result.would_embed} chunks across ${pages.length} pages`);
} else {
slog(`Embedded ${result.embedded} chunks across ${pages.length} pages`);
}
}
@@ -816,8 +802,6 @@ async function embedAllStale(
pacer?: DbPacer;
/** Resolved concurrency cap (E-1: the worker count, no separate permit). */
paceMaxConcurrency?: number;
/** #394: suppress human stdout summaries (structured-output callers). */
quiet?: boolean;
},
signature?: string,
externalSignal?: AbortSignal,
@@ -835,7 +819,7 @@ async function embedAllStale(
signature,
...(sourceId && { sourceId }),
});
if (invalidated > 0 && !staleOpts?.quiet) {
if (invalidated > 0) {
slog(`[embed] invalidated ${invalidated} chunk(s) embedded under a prior model signature`);
}
}
@@ -846,12 +830,10 @@ async function embedAllStale(
dryRun && signature ? { ...sourceOpt, signature } : sourceOpt,
);
if (staleCount === 0) {
if (!staleOpts?.quiet) {
if (dryRun) {
slog('[dry-run] Would embed 0 chunks (0 stale found)');
} else {
slog('Embedded 0 chunks (0 stale found)');
}
if (dryRun) {
slog('[dry-run] Would embed 0 chunks (0 stale found)');
} else {
slog('Embedded 0 chunks (0 stale found)');
}
return;
}
@@ -860,7 +842,7 @@ async function embedAllStale(
result.would_embed += staleCount;
result.total_chunks += staleCount;
if (onProgress) onProgress(1, 1, 0);
if (!staleOpts?.quiet) slog(`[dry-run] Would embed ${staleCount} stale chunks`);
slog(`[dry-run] Would embed ${staleCount} stale chunks`);
return;
}
@@ -1100,7 +1082,7 @@ async function embedAllStale(
if (budgetTimer) clearTimeout(budgetTimer);
}
if (!staleOpts?.quiet) slog(`Embedded ${result.embedded} chunks across ${totalProcessedPages} pages`);
slog(`Embedded ${result.embedded} chunks across ${totalProcessedPages} pages`);
// #1946 (OV2a): a catch-up pass that completed without being aborted but left
// chunks unembedded means those chunks are stuck (a non-transient embed
+1 -3
View File
@@ -1214,9 +1214,7 @@ async function runPhaseEmbed(engine: BrainEngine, dryRun: boolean, signal?: Abor
// 10-15 min one) bails within a batch instead of running to completion
// after the job was killed — which left gbrain_cycle_locks held and
// wedged every subsequent autopilot cycle.
// #394: quiet — the cycle reports embed counts via its own PhaseResult;
// raw `[dry-run] Would embed ...` stdout lines would corrupt `dream --json`.
const result = await runEmbedCore(engine, { stale: true, dryRun, signal, quiet: true });
const result = await runEmbedCore(engine, { stale: true, dryRun, signal });
const embeddedCount = dryRun ? result.would_embed : result.embedded;
return {
phase: 'embed',
+10 -4
View File
@@ -93,7 +93,13 @@ 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, bestPath, revertAllPending, skillPath, writeProposed } from './version-store.ts';
import {
acceptCandidate,
proposedPath as proposedFilePath,
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';
@@ -702,9 +708,9 @@ 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') {
// 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);
// writeProposed() emitted both the best pointer and the stable review
// artifact in the accept branch. SKILL.md remains untouched.
proposedPath = proposedFilePath(skillsDir, skillName);
} else if (mutateDecision.mutate) {
mutatedSkillFile = finalOutcome === 'accepted';
}
+15 -9
View File
@@ -23,6 +23,7 @@
*
* history.json
* best.md
* proposed.md
* versions/
* v0001_e1_s1.md
* v0002_e1_s2.md
@@ -52,6 +53,10 @@ 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');
}
@@ -171,17 +176,18 @@ export function acceptCandidate(input: AcceptInput): AcceptResult {
}
/**
* 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).
* 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).
*/
export function writeProposed(skillsDir: string, skillName: string, candidateText: string): string {
const p = bestPath(skillsDir, skillName);
fs.mkdirSync(path.dirname(p), { recursive: true });
atomicWrite(p, candidateText);
return p;
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;
}
/**
-27
View File
@@ -292,33 +292,6 @@ describe('runDream — output format', () => {
expect(parsed).toHaveProperty('totals');
});
// #394 / takeover of #854: the embed phase's `[dry-run] Would embed ...`
// summary must not leak onto stdout ahead of the JSON CycleReport.
test('--dry-run --json emits only JSON even when embed has stale chunks', async () => {
await engine.putPage('concepts/testing', {
type: 'concept',
title: 'Testing',
compiled_truth: 'Testing keeps JSON contracts honest.',
timeline: '',
});
await engine.upsertChunks('concepts/testing', [
{ chunk_index: 0, chunk_text: 'Testing keeps JSON contracts honest.', chunk_source: 'compiled_truth' },
]);
const lines: string[] = [];
const logSpy = spyOn(console, 'log').mockImplementation((msg: string) => { lines.push(String(msg)); });
await runDream(engine, ['--dir', repo, '--phase', 'embed', '--dry-run', '--json']);
logSpy.mockRestore();
const output = lines.join('\n');
expect(output.trimStart().startsWith('{')).toBe(true);
const parsed = JSON.parse(output);
expect(parsed.schema_version).toBe('1');
expect(parsed.phases[0].phase).toBe('embed');
// The stale chunk was still counted in the structured report.
expect(parsed.phases[0].details.would_embed).toBe(1);
});
test('human output for clean status mentions "Brain is healthy"', async () => {
const lines: string[] = [];
const logSpy = spyOn(console, 'log').mockImplementation((msg: string) => { lines.push(String(msg)); });
+4 -4
View File
@@ -39,6 +39,7 @@ 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';
@@ -741,7 +742,7 @@ describe('skillopt T3 — F11 held-out gate, ablation opts, no-DB-pollution', ()
} finally { fixture.cleanup(); }
});
test('--no-mutate writes proposed.md (best.md), leaves SKILL.md untouched', async () => {
test('--no-mutate writes proposed.md and best.md, leaves SKILL.md untouched', async () => {
const fixture = setupFixture(SKILL_PEOPLE_ONLY, CITATIONS_BENCHMARK);
try {
installStub({
@@ -753,10 +754,9 @@ 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).toBeDefined();
// proposed.md (best.md) exists and carries the improvement.
expect(fs.existsSync(result.proposedPath!)).toBe(true);
expect(result.proposedPath).toBe(proposedPath(fixture.skillsDir, SKILL));
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');
+15
View File
@@ -12,9 +12,11 @@ import {
bestPath,
historyPath,
loadHistory,
proposedPath,
revertAllPending,
skillPath,
versionsDir,
writeProposed,
} from '../../src/core/skillopt/version-store.ts';
let tmpDir: string;
@@ -79,6 +81,19 @@ 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);