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
10 changed files with 53 additions and 144 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 \
+5 -5
View File
@@ -197,7 +197,7 @@ export interface SyncResult {
/** Pages re-embedded during this sync's auto-embed step. 0 if --no-embed or skipped. */
embedded: number;
pagesAffected: string[];
failedFiles?: number; // count of per-file import/sync failures (Bug 9)
failedFiles?: number; // count of parse failures (Bug 9)
/**
* v0.41.13.0 partial-sync fields (only set when status === 'partial').
*
@@ -3183,7 +3183,7 @@ async function performSyncInner(engine: BrainEngine, opts: SyncOpts): Promise<Sy
await clearOpCheckpoint(engine, ckpt.target);
};
// issue #1939 adversarial finding #1: a file that failed to import (open ledger
// issue #1939 adversarial finding #1: a file that failed to parse (open ledger
// row) and is then deleted/renamed-away never re-enters failedFiles and never
// imports, so its row would never clear and would age doctor to a permanent
// FAIL. Treat removed paths as resolved so the ledger self-heals.
@@ -3215,9 +3215,9 @@ async function performSyncInner(engine: BrainEngine, opts: SyncOpts): Promise<Sy
} else {
const fileFailCount = failedFiles.filter(f => isSkippablePath(f.path)).length;
serr(
`\nSync blocked: ${fileFailCount} file(s) failed to import:\n` +
`\nSync blocked: ${fileFailCount} file(s) failed to parse:\n` +
`${codeBreakdown}\n\n` +
`Fix the listed file errors and re-run, or use 'gbrain sync --skip-failed' to ` +
`Fix the frontmatter and re-run, or use 'gbrain sync --skip-failed' to ` +
`acknowledge and move on. A file that keeps failing auto-skips after ` +
`${resolveAutoSkipThreshold()} consecutive syncs.`,
);
@@ -5355,7 +5355,7 @@ function printSyncResult(result: SyncResult, sink: NodeJS.WriteStream = process.
case 'dry_run':
break; // already printed in performSync
case 'blocked_by_failures':
write(`Sync BLOCKED at ${result.toCommit.slice(0, 8)}: ${result.failedFiles ?? 0} file(s) failed to import.`);
write(`Sync BLOCKED at ${result.toCommit.slice(0, 8)}: ${result.failedFiles ?? 0} file(s) failed to parse.`);
write(` See ~/.gbrain/sync-failures.jsonl for details, or run 'gbrain doctor'.`);
write(` Fix the files then re-run 'gbrain sync', or 'gbrain sync --skip-failed' to move on.`);
break;
-10
View File
@@ -1058,16 +1058,6 @@ export class PGLiteEngine implements BrainEngine {
RETURNING id, source_id, slug, type, title, compiled_truth, timeline, frontmatter, content_hash, created_at, updated_at, effective_date, effective_date_source, import_filename, source_kind, source_uri, ingested_via, ingested_at`,
[sourceId, slug, page.type, pageKind, page.title, page.compiled_truth, page.timeline || '', JSON.stringify(frontmatter), hash, effectiveDate, effectiveDateSource, importFilename, chunkerVersion, sourcePath, sourceKind, sourceUri, ingestedVia, ingestedAt]
);
// #2189: an INSERT … ON CONFLICT DO UPDATE … RETURNING that yields 0 rows
// (e.g. a BEFORE trigger suppressing the write) previously crashed in
// rowToPage with an opaque "undefined is not an object (row.deleted_at)".
// Throw a diagnosable error naming the row instead. Mirrors postgres-engine.ts.
if (!rows[0]) {
throw new Error(
`putPage: INSERT … RETURNING produced no row for slug='${slug}' source_id='${sourceId}'. ` +
`A trigger or rule on the pages table may be suppressing the write.`
);
}
return rowToPage(rows[0] as Record<string, unknown>);
}
-10
View File
@@ -1119,16 +1119,6 @@ export class PostgresEngine implements BrainEngine {
ingested_at = COALESCE(EXCLUDED.ingested_at, pages.ingested_at)
RETURNING id, source_id, slug, type, title, compiled_truth, timeline, frontmatter, content_hash, created_at, updated_at, effective_date, effective_date_source, import_filename, source_kind, source_uri, ingested_via, ingested_at
`;
// #2189: an INSERT … ON CONFLICT DO UPDATE … RETURNING that yields 0 rows
// (e.g. a BEFORE trigger suppressing the write) previously crashed in
// rowToPage with an opaque "undefined is not an object (row.deleted_at)".
// Throw a diagnosable error naming the row instead. Mirrors pglite-engine.ts.
if (!rows[0]) {
throw new Error(
`putPage: INSERT … RETURNING produced no row for slug='${slug}' source_id='${sourceId}'. ` +
`A trigger or rule on the pages table may be suppressing the write.`
);
}
return rowToPage(rows[0]);
}
+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;
}
/**
+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');
-74
View File
@@ -1,74 +0,0 @@
// #2189 regression guard: putPage's INSERT … ON CONFLICT DO UPDATE … RETURNING
// can yield 0 rows when brain-local DB state (e.g. a BEFORE INSERT trigger)
// suppresses the write. Pre-fix, rowToPage(rows[0]) crashed with the opaque
// "undefined is not an object (evaluating 'row.deleted_at')" that failed
// ~all files of a code sync. Post-fix, putPage throws a descriptive error
// naming the slug + source_id so the failure is diagnosable per-file.
//
// Same guard lands in postgres-engine.ts (engine-parity invariant); this test
// exercises the PGLite side, where the issue was reported.
import { describe, expect, test, beforeAll, afterAll } from 'bun:test';
import { PGLiteEngine } from '../src/core/pglite-engine.ts';
let engine: PGLiteEngine;
beforeAll(async () => {
engine = new PGLiteEngine();
await engine.connect({});
await engine.initSchema();
// Simulate the reporter's state-dependent failure: a trigger that
// suppresses inserts for one slug, making RETURNING produce no row.
await engine.executeRaw(`
CREATE OR REPLACE FUNCTION suppress_pages_insert() RETURNS trigger AS $$
BEGIN
IF NEW.slug = 'suppressed-page' THEN RETURN NULL; END IF;
RETURN NEW;
END;
$$ LANGUAGE plpgsql;
`);
await engine.executeRaw(`
CREATE TRIGGER suppress_pages_insert_trg
BEFORE INSERT ON pages
FOR EACH ROW EXECUTE FUNCTION suppress_pages_insert();
`);
});
afterAll(async () => {
await engine.executeRaw('DROP TRIGGER IF EXISTS suppress_pages_insert_trg ON pages');
await engine.executeRaw('DROP FUNCTION IF EXISTS suppress_pages_insert');
await engine.disconnect();
});
describe('putPage RETURNING guard (#2189)', () => {
test('0-row RETURNING throws a descriptive error, not row.deleted_at TypeError', async () => {
let err: Error | undefined;
try {
await engine.putPage('suppressed-page', {
type: 'code',
title: 'Suppressed',
compiled_truth: 'x',
timeline: '',
});
} catch (e) {
err = e as Error;
}
expect(err).toBeDefined();
expect(err!.message).toContain('putPage');
expect(err!.message).toContain("slug='suppressed-page'");
expect(err!.message).toContain("source_id='default'");
// The pre-fix crash signature must be gone.
expect(err!.message).not.toContain('deleted_at');
});
test('unsuppressed slugs still upsert normally with the trigger installed', async () => {
const page = await engine.putPage('normal-page', {
type: 'concept',
title: 'Normal',
compiled_truth: 'y',
timeline: '',
});
expect(page.slug).toBe('normal-page');
expect(page.source_id).toBe('default');
});
});
+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);
-25
View File
@@ -375,31 +375,6 @@ describe('performSync dry-run never writes', () => {
expect(messages.some(m => m.includes('git pull failed'))).toBe(false);
});
test('first PGLite code sync imports code files without runtime failures', async () => {
const { performSync } = await import('../src/commands/sync.ts');
mkdirSync(join(repoPath, 'src'), { recursive: true });
writeFileSync(
join(repoPath, 'src/example.ts'),
'export function add(left: number, right: number) { return left + right; }\n',
);
execSync('git add -A && git commit -m "add code file"', { cwd: repoPath, stdio: 'pipe' });
const result = await performSync(engine, {
repoPath,
noPull: true,
noEmbed: true,
noExtract: true,
strategy: 'code',
});
expect(result.status).toBe('first_sync');
expect(result.added).toBe(1);
expect(result.failedFiles ?? 0).toBe(0);
const page = await engine.getPage('src-example-ts');
expect(page?.type).toBe('code');
expect(page?.frontmatter).toMatchObject({ file: 'src/example.ts', language: 'typescript' });
});
test('incremental dry-run does NOT write to DB or advance the bookmark', async () => {
const { performSync } = await import('../src/commands/sync.ts');
// First do a real sync to seed the bookmark.