Files
gbrain/src/core/audit-quality-probe.ts
T
94aaf7e396 v0.40.1.0 Track D — eval infrastructure (catch retrieval regressions, prove answer-quality wins) (#1298)
* feat(eval-longmemeval): --by-type flag + question field + resume-replace

Per-question JSONL row gains `question`, `question_type`, and (when
ground truth is available) `recall_hit` — additive fields that existing
consumers (LongMemEval's `evaluate_qa.py`) ignore. New `--by-type` flag
emits a `{kind:"by_type_summary", recall_by_type, aggregate}` line at
the end of the output, resume-safe: rebuilt from existing rows so the
final aggregate covers cumulative resumed questions, prior summary at
the tail replaced rather than appended. New `--by-type-floor F` exits
non-zero per breached question_type. Empty-bucket guard emits null rate
not NaN. Exports `buildByTypeSummary` + `emitByTypeSummary` +
`seedRecallByTypeFromFile` for unit testing.

* feat(eval-cross-modal): --batch flag + semaphore + DI seam

Adds `--batch <jsonl> [--limit N] [--concurrent N] [--max-usd FLOAT]
[--yes]` to the existing eval cross-modal command. Mutually exclusive
with --task. Reads LongMemEval-shape JSONL output, filters by_type_summary
rows automatically, fans out via a new `runWithLimit<T>` semaphore
primitive (default --concurrent 3 x 3 model slots = 9 simultaneous calls;
below tier-1 rate limits on all 3 providers). Pre-flight cost estimate
refuses past --max-usd (default $5) unless --yes. Per-question receipts
written to a per-batch tempdir + deleted at end of run so
~/.gbrain/eval-receipts/ stays clean; summary receipt inlines verdicts.

Exit precedence (new batch-level policy, not inherited from aggregate.ts):
ERROR > FAIL > INCONCLUSIVE > PASS — any per-question runtime error exits 2.

New `runEvalCrossModal(args, opts?: {runEval?})` DI seam mirrors the
existing eval-longmemeval pattern. Tests pass a stub runEval so unit tests
don't need API keys; gateway availability check is also skipped when
opts.runEval is provided. Pinned by 17 cases.

* test: hermetic qrels retrieval gate against synthetic basis-vector corpus

Adds test/eval-replay-gate.test.ts as a unit-shard test (NOT under
test/e2e/ — the unit-shard CI matrix runs every PR via bun test;
test/e2e/ is fixed-file). Seeds a PGLite engine with synthetic
placeholder-name pages whose embeddings are basis vectors (same pattern
as test/e2e/search-quality.test.ts:23-28) so retrieval is hermetic — no
API keys, no DATABASE_URL, fully deterministic.

The qrels fixture at test/fixtures/eval-baselines/qrels-search.json has
12 hand-curated queries; each maps to a ranked list of relevant slugs +
`first_relevant_slug` (expected top-1). For each query, the gate asserts
`top1_match_rate >= 0.80` AND `recall_at_10 >= 0.85`. Env-overridable
floors via GBRAIN_REPLAY_GATE_TOP1_FLOOR / GBRAIN_REPLAY_GATE_RECALL_FLOOR
through withEnv(). Gate-fire prints per-query HIT/miss + recall to stderr.

When ranking changes intentionally move expected slugs, edit
qrels-search.json directly with a 'Why:' line in the commit body —
documented in docs/eval-bench.md.

scripts/check-test-real-names.sh allowlist gains 6 entries for the
privacy-grep regression guard inside the test, which must literally
spell the names it forbids to assert they're NOT in the fixture (same
meta-rule exception as skillpack-harvest privacy tests).

* feat(autopilot): opt-in nightly cross-modal quality probe + doctor check

Composes `gbrain eval longmemeval --by-type` + `gbrain eval cross-modal
--batch` into a 24h-cadenced quality check. Default DISABLED — opt-in via
`gbrain config set autopilot.nightly_quality_probe.enabled true` so new
users don't discover background API spend.

src/core/cycle/nightly-quality-probe.ts ships the phase implementation
with a full NightlyProbeDeps DI surface (isEnabled, hasEmbeddingProvider,
resolveMaxUsd, resolveRepoRoot, runLongMemEval, runCrossModalBatch, now)
so tests stub every external effect — no PGLite, no real LLM calls.
Pure `shouldRunNightly(now, recentEvents, windowMs?)` rate-limit fn.

src/core/audit-quality-probe.ts is the ISO-week-rotated JSONL writer
(mirrors audit-slug-fallback.ts; honors GBRAIN_AUDIT_DIR). One event per
run: outcome (pass/fail/inconclusive/error/budget_exceeded/rate_limited/
no_embedding_key), exit code, pass/fail/error counts, est_cost_usd,
fixture_sha8.

src/commands/doctor.ts gains a `nightly_quality_probe_health` check:
SKIPPED with paste-ready enable command when disabled; OK with timestamp
when all PASS in last 7 days; WARN with per-outcome counts when any
FAIL/ERROR/BUDGET_EXCEEDED. Extracted as pure
`computeNightlyQualityProbeHealthCheck(probeEnabled, events)` for
unit testing.

test/fixtures/longmemeval-nightly.jsonl is a 10-question placeholder
dataset (synthetic names only) distinct from the existing 5-question
mini fixture so the probe has consistent regression signal.

Real expected cost: ~$0.35/night = ~$10.50/month. Worst-case at
default $5 cap: $150/month.

Pinned by 21 cases in test/nightly-quality-probe.test.ts covering the
rate-limit pure function, every outcome branch, and all 7 branches of
the doctor check.

Autopilot scheduler wiring deferred to v0.41+ — the phase is callable
in isolation today (via the DI surface); cycle-loop dispatcher
integration filed in TODOS.md as a follow-up.

* docs: document Track D eval surfaces + file v0.41+ follow-up TODOs

docs/eval-bench.md gains a 'v0.40.1.0 Track D — Eval infrastructure'
section covering: --by-type usage + resume-replace semantics, the
hermetic qrels gate workflow + 'Why:' commit-body refresh convention,
--batch end-to-end with cost-bound + concurrency knobs, and the opt-in
nightly probe enable workflow + cost ceiling.

TODOS.md files two follow-ups:
- v0.41+: contributor-mode CI capture for BrainBench-Real replay gate
  (the deferred original Task 2 design — replay against real captured
  queries is more valuable than synthetic qrels long-term, but needs CI
  secret + nightly capture pipeline + commit automation; deferred to a
  dedicated wave)
- v0.41+: wire the nightly quality probe into autopilot scheduling
  (phase callable in isolation today; cycle-loop dispatcher integration
  is a ~3-hour follow-up)

CLAUDE.md Key Files annotations extended for the four lanes:
eval-longmemeval gains the --by-type description, eval-cross-modal
gains the --batch + DI seam description, new entries for the qrels
gate test + the nightly probe + audit-quality-probe writer.

* chore: bump version and changelog (v0.40.1.0)

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(eval): close 4 codex-flagged eval-integrity bugs

Codex adversarial review on the Track D wave found 4 real ways the new
eval-gate code could silently bypass its gates. Each fix below either
counts what was previously dropped, fails fast on a parser edge case,
or enforces a gate that was previously skipped on an early-return path.

CDX-1: cross-modal --batch silently dropped failed/corrupt LongMemEval
rows. `gbrain eval longmemeval` emits {error:..., hypothesis:''} when
runOneQuestion throws; the batch reader's missing-field skip threw those
rows away, shrinking the denominator. A green eval on a subset is now
impossible:
  - eval-longmemeval.ts: error rows now carry `question` + `question_type`
    so the batch consumer can identify them as upstream failures, not
    skip them as malformed.
  - eval-cross-modal.ts: readBatchRows now returns {rows, upstream_errors,
    malformed_count}. Upstream errors fold into per_question with verdict
    'upstream_error'. BatchSummary gains `upstream_error_count` and
    `malformed_count`. ERROR exit precedence widens to include both, so
    any upstream failure exits 2.

CDX-2: --limit 0 was a direct CI bypass — zero-row check fired before
slicing, then the empty result fell through to verdict='pass'. Fixed
with a hard `limit >= 1` check.

CDX-3: --resume-from + --by-type-floor was a real gate skip. When a
prior run had every question answered, the early "nothing to do" return
fired BEFORE summary emission and floor enforcement. Now the no-op
resume path still seeds recallByType from the existing file, emits the
by_type_summary at the tail, and runs the floor gate.

CDX-5: doctor nightly_quality_probe_health only flagged fail / error /
budget_exceeded as warn. no_embedding_key / rate_limited / inconclusive
were silently reported as PASS — hiding misconfigurations and queue
backpressure. The bad-event filter is now `outcome !== 'pass'`, and the
counts string surfaces every bucket so the operator sees exactly what
went wrong.

scripts/check-privacy.sh: adds test/eval-replay-gate.test.ts to the
allowlist (the qrels test's privacy-grep regression guard literally
names what it forbids, same meta-rule exception as the existing
test/recency-decay.test.ts + skillpack-harvest allowlist entries).

Pinned by 8 new regression cases across eval-longmemeval (CDX-3),
eval-cross-modal-batch (CDX-1 + CDX-2), and nightly-quality-probe
(CDX-5). 76 Track D tests pass.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-05-22 22:16:25 -07:00

123 lines
4.4 KiB
TypeScript

/**
* v0.40.1.0 Track D / T6 — nightly quality probe audit trail.
*
* Writes one event per nightly cross-modal probe run to
* `~/.gbrain/audit/quality-probe-YYYY-Www.jsonl` (ISO-week rotation).
* Mirrors `audit-slug-fallback.ts` for filename + best-effort write
* semantics. Honors `GBRAIN_AUDIT_DIR` via the shared `resolveAuditDir`.
*
* Read by `gbrain doctor`'s `nightly_quality_probe_health` check to
* surface FAIL / ERROR / BUDGET_EXCEEDED runs from the last 7 days.
*/
import * as fs from 'node:fs';
import * as path from 'node:path';
import { resolveAuditDir } from './minions/handlers/shell-audit.ts';
export type QualityProbeOutcome =
| 'pass'
| 'fail'
| 'inconclusive'
| 'error'
| 'budget_exceeded'
| 'rate_limited'
| 'no_embedding_key';
export interface QualityProbeAuditEvent {
ts: string;
/** Verdict from the cross-modal batch summary (or short-circuit reason). */
outcome: QualityProbeOutcome;
/** Exit code of the underlying batch (0/1/2/-1 when short-circuited). */
exit_code: number;
pass_count: number;
fail_count: number;
inconclusive_count: number;
error_count: number;
/** Estimated cost in USD (0 when short-circuited before runs). */
est_cost_usd: number;
/** Sha-8 of the fixture file content for change detection. */
fixture_sha8?: string;
/** Optional human-readable detail (e.g. error message, "no chat provider configured"). */
detail?: string;
}
/** ISO-week-rotated filename: `quality-probe-YYYY-Www.jsonl`. Mirrors audit-slug-fallback. */
export function computeQualityProbeAuditFilename(now: Date = new Date()): string {
const d = new Date(Date.UTC(now.getUTCFullYear(), now.getUTCMonth(), now.getUTCDate()));
const dayNum = (d.getUTCDay() + 6) % 7;
d.setUTCDate(d.getUTCDate() - dayNum + 3);
const isoYear = d.getUTCFullYear();
const firstThursday = new Date(Date.UTC(isoYear, 0, 4));
const firstThursdayDayNum = (firstThursday.getUTCDay() + 6) % 7;
firstThursday.setUTCDate(firstThursday.getUTCDate() - firstThursdayDayNum + 3);
const weekNum = Math.round((d.getTime() - firstThursday.getTime()) / (7 * 86400000)) + 1;
const ww = String(weekNum).padStart(2, '0');
return `quality-probe-${isoYear}-W${ww}.jsonl`;
}
/**
* Append one quality-probe event. Best-effort: write failure logs to stderr
* but the probe phase continues.
*/
export function logQualityProbeEvent(event: Omit<QualityProbeAuditEvent, 'ts'> & { ts?: string }): void {
const stamped: QualityProbeAuditEvent = {
ts: event.ts ?? new Date().toISOString(),
outcome: event.outcome,
exit_code: event.exit_code,
pass_count: event.pass_count,
fail_count: event.fail_count,
inconclusive_count: event.inconclusive_count,
error_count: event.error_count,
est_cost_usd: event.est_cost_usd,
...(event.fixture_sha8 !== undefined ? { fixture_sha8: event.fixture_sha8 } : {}),
...(event.detail !== undefined ? { detail: event.detail } : {}),
};
const dir = resolveAuditDir();
const file = path.join(dir, computeQualityProbeAuditFilename());
try {
fs.mkdirSync(dir, { recursive: true });
fs.appendFileSync(file, JSON.stringify(stamped) + '\n', { encoding: 'utf8' });
} catch (err) {
const msg = err instanceof Error ? err.message : String(err);
process.stderr.write(`[gbrain] quality-probe audit write failed (${msg}); probe continues\n`);
}
}
/**
* Read recent quality-probe events from the current + prior ISO week files.
* Used by `gbrain doctor`'s nightly_quality_probe_health check. Missing
* files and corrupt rows are skipped silently.
*/
export function readRecentQualityProbeEvents(
days = 7,
now: Date = new Date(),
): QualityProbeAuditEvent[] {
const dir = resolveAuditDir();
const cutoff = now.getTime() - days * 86400000;
const out: QualityProbeAuditEvent[] = [];
const filenames = [
computeQualityProbeAuditFilename(now),
computeQualityProbeAuditFilename(new Date(now.getTime() - 7 * 86400000)),
];
for (const filename of filenames) {
const file = path.join(dir, filename);
let content: string;
try {
content = fs.readFileSync(file, 'utf8');
} catch {
continue;
}
for (const line of content.split('\n')) {
if (line.length === 0) continue;
try {
const ev = JSON.parse(line) as QualityProbeAuditEvent;
const ts = Date.parse(ev.ts);
if (Number.isFinite(ts) && ts >= cutoff) out.push(ev);
} catch {
// corrupt row — skip
}
}
}
return out;
}