mirror of
https://github.com/garrytan/gbrain.git
synced 2026-07-27 22:15:33 +00:00
82cc7fff90429ef05585705b3d91a74afeffb4ce
5
Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
|
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>
|
||
|
|
cdba533a04 |
v0.36.2.0 feat: ZeroEntropy as default + zero-based README rewrite (#1136)
* feat(dims): OpenAI text-embedding-3 Matryoshka range validation (D13) dimsProviderOptions now fail-loud at the embed boundary when the configured embedding_dimensions is outside the model's native range (1..1536 for -small, 1..3072 for -large). Paste-ready fix hint in the AIConfigError.fix field. Closes the silent-HTTP-400 path that would have bit OpenAI-fallback users on v0.36.0.0 ZE-default installs. 16 new test cases in test/ai/dims-openai.test.ts pinning the contract across native-openai and openai-compatible adapter paths. * feat(ai): flip defaults to ZeroEntropy zembed-1 1280d + zerank-2 reranker Default embedding model is now zeroentropyai:zembed-1 at 1280d via Matryoshka. Real-corpus benchmark: 2.2x faster than OpenAI, 2.6x cheaper at regular pricing, wins 11/20 head-to-head queries. 1280 is the closest valid ZE Matryoshka step to the prior OpenAI 1536d default (valid set: 2560/1280/640/320/160/80/40). 1024 (Voyage's step) is NOT on ZE's list — pinned by AIConfigError fail-loud in dims.ts. balanced mode bundle now defaults reranker_enabled=true. zerank-2 reshuffles 60% of top-1 results in benchmarks. Missing-key fail-open contract in src/core/search/rerank.ts handles unauthenticated cases. Opt out with: gbrain config set search.reranker.enabled false Existing tests updated (gateway.test.ts, search-mode.test.ts) and a new test/balanced-reranker-default.test.ts (10 cases) pins the fail- open invariants. * feat(retrieval-upgrade): RetrievalUpgradePlanner + interactive prompt UX New src/core/retrieval-upgrade-planner.ts is the consolidated planner that computes the brain's pending retrieval-upgrade work (chunker bumps + ZE switch) in one pass and applies the schema transition + config updates atomically. Tagged-union ApplyResult enum (D15): 'applied' | 'skipped_already_ applied' | 'skipped_no_work' | 'declined' | 'planned' | 'failed'. No string-parsing reasons. Three config keys (D12): ze_switch_prompt_shown (UI state), ze_switch_requested (user intent), ze_switch_applied (work done). Plus ze_switch_previous_snapshot (JSON, full prior config for --undo per D16) and ze_switch_declined_at (90-day re-ask window). Schema transition (D18) is atomic: DROP indexes + ALTER COLUMN + CREATE INDEX inside a single engine.transaction(). HNSW recreation is part of the same transaction — no silent slow-search window. C3 eligibility logic: ze_switch_offered iff NOT on ZE + NOT declined recently + NOT applied + (legacy default OR >100 pages). C4 cost math: MAX(chunker_pending, dim_pending) not SUM — one re-embed pass invalidates both surfaces simultaneously. New src/core/retrieval-upgrade-prompt.ts wires the planner to a TTY-only interactive prompt with two-line cost split (D10) and privacy callout for the reranker flip. Tests: test/retrieval-upgrade-planner.test.ts (24 cases) pins the state machine. test/asymmetric-encoding-contract.test.ts (6 cases) pins D17: search read path uses gateway.embedQuery() not embed(), asserted via __setEmbedTransportForTests mock. * feat(cli): gbrain ze-switch — manual lever for the ZE switch New gbrain ze-switch CLI with --dry-run, --json, --resume, --force, --undo, --non-interactive, --confirm-reembed, --ignore-missing-key flags. Mirrors the upgrade prompt's UX symmetry: --undo presents a cost-warning before re-embedding back to the prior width. src/cli.ts: dispatch case + CLI_ONLY entry. ze-switch owns its own engine lifecycle (mirrors the doctor pattern). test/ze-switch-cli.test.ts (11 cases): --help, --dry-run, --json, --non-interactive, --ignore-missing-key, --resume, --undo, --confirm-reembed. Uses captureExit harness to test process.exit() paths without breaking the test process. * feat(doctor): ze_embedding_health + embedding_width_consistency checks Two new doctor checks (D-A5): ze_embedding_health: when embedding_model starts with zeroentropyai:, verify ZEROENTROPY_API_KEY is set (env or config). Paste-ready setup hint with the signup URL on failure. embedding_width_consistency: cross-check that the configured embedding_dimensions matches the actual vector(N) column width on content_chunks.embedding. Catches the half-applied switch state (schema migrated but config write crashed) with a paste-ready gbrain ze-switch --resume hint. Wired into runDoctor between reranker_health and the existing sync_freshness checks. Both checks gracefully no-op on non-ZE embedding configs. test/doctor-ze-checks.test.ts (8 cases) pins both checks across happy + missing-key + missing-config + drift paths. Uses withEnv() helper to clear ZEROENTROPY_API_KEY for the no-key path so tests are hermetic against contributor env state. test/e2e/v0_28_5-fix-wave.test.ts + test/openai-compat-multimodal.test.ts: updated to explicit-configure the gateway when the test depends on specific dims that diverge from the v0.36.0.0 default (1280d). * docs: README zero-based rewrite (884 -> 139 lines) + new docs files Strip 4 months of accreted "New in v0.X.Y" hero blocks and reorganize around what gbrain does today. 33 H2s -> 8. The Commands section (136 lines duplicating gbrain --help) moved out; the 6-table skills enumeration collapsed to a one-paragraph capability description with a link to skills/RESOLVER.md. Hero retains load-bearing facts: OpenClaw + Hermes credit, production numbers (17,888 pages / 4,383 people / 723 companies), BrainBench numbers (P@5 49.1% / R@5 97.9% / +31.4 lift), ZE comparison numbers, 30-min install claim. Adds one paragraph announcing the v0.36.0.0 ZE default with the explicit gbrain config set escape for OpenAI/Voyage users. New files: - docs/INSTALL.md: every install path consolidated (agent platform, CLI standalone, MCP server). Thin-client mode covered. - docs/architecture/RETRIEVAL.md: why the hybrid + graph stack works. BrainBench numbers, why each strategy alone fails, the source-aware ranking + intent classification + multi-query expansion story. - docs/ethos/ORIGIN.md: origin story lifted from the old README so the front door stays factual + concrete. test/readme-hero-anchors.test.ts (5 cases) is the D9 regression guard. Five load-bearing strings: OpenClaw, Hermes, ZE, production-numbers regex, P@5/R@5. Light anchors that let voice/ structure evolve but block accidental loss of headline facts. scripts/check-test-real-names.sh: allowlist entries for OpenClaw + Hermes literals in the anchor test (it explicitly asserts those strings appear in README). * chore: bump version and changelog (v0.36.0.0) ZeroEntropy as the new default for embedding (zembed-1 at 1280d via Matryoshka) and reranker (zerank-2 cross-encoder, on by default in balanced mode bundle). README zero-based rewrite (884 -> 139 lines). 3 new docs files. Two new doctor checks. New gbrain ze-switch CLI with --undo for symmetric reversibility. skills/migrations/v0.36.0.0.md tells the agent how to surface the retrieval-upgrade prompt post-upgrade. llms-full.txt regenerated via bun run build:llms. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * fix(docs): scrub Wintermute from RETRIEVAL.md per privacy rule * chore: rebump version 0.36.0.0 → 0.36.2.0 (queue collision) Three open PRs were claiming v0.36.0.0 (#1130 skillpack, #1139 hindsight, #1136 this PR). Ship-aware queue allocator says this branch lands at v0.36.2.0. Trio audit: VERSION 0.36.2.0 package.json 0.36.2.0 CHANGELOG ## [0.36.2.0] - 2026-05-17 Updates: VERSION, package.json, CHANGELOG header + body refs, README "New default in v0.36.2.0" announcement + credit line, skills/migrations/v0.36.0.0.md renamed to v0.36.2.0.md with frontmatter + body refs updated. llms-full.txt regenerated. * fix(test): pin gateway dim=1536 in cross-file-stateful PGLite tests CI shard 1 reported 10 failures across `query-cache.test.ts` (6) and `consolidate-valid-until.test.ts` (4). Both files hardcode 1536-dim vectors but rely on `PGLiteEngine.initSchema()` to size `vector(__EMBEDDING_DIMS__)` at the right width. Root cause: v0.36.2.0 flipped DEFAULT_EMBEDDING_DIMENSIONS from 1536 to 1280 (ZE Matryoshka step). The gateway module is process-singleton; when ANOTHER test file in the same shard's bun-test process configures the gateway before us, `pglite-engine.ts:216` reads `getEmbeddingDimensions() === 1280` and sizes the schema columns at vector(1280). The hardcoded 1536-dim INSERTs then fail with "expected 1280 dimensions, not 1536". Locally these tests pass in isolation because the gateway falls back through the try/catch at pglite-engine.ts:218 (1536 default). CI runs multiple test files in one process, so cross-file state poisons the schema width. Fix: explicit `resetGateway()` + `configureGateway({embedding_dimensions: 1536, ...})` at the top of `beforeAll`, plus `resetGateway()` in `afterAll`. Pins the schema width regardless of cross-file state. --------- Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com> |
||
|
|
03947665e4 |
v0.36.0.0 feat(skillpack): scaffold + reference + harvest (retire managed-block install) (#1130)
* feat(skillpack): extract copyArtifacts shared helper (T1)
Pure file-copy primitive for scaffold (gbrain→host) and harvest (host→gbrain).
Atomic-refusal contract: symlink-reject + canonical-path containment validate
every item before any write. Used by both directions of the v0.33 loop.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
* feat(skillpack): scaffold subcommand + SKILL.md frontmatter sources (T2)
New scaffold.ts replaces the managed-block installer. One-time additive copy
into the user's repo via copyArtifacts; refuses to overwrite existing files
(user owns them). Partial-state policy: copies missing paired sources even
when the skill dir already exists.
bundle.ts extended with loadSkillSources + enumerateScaffoldEntries — paired
source files declared in each SKILL.md's frontmatter sources: array, not in
openclaw.plugin.json. Single source of truth, co-located with the skill.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
* feat(skillpack): reference command + apply-clean-hunks (T4 + T15)
reference is the read-only diff lens with an agent-readable framing line. Pure-JS
unified-diff producer + parser + applier (no patch(1) dependency). Two-way merge
with documented limitation: without scaffold-time base tracking, applied hunks
align everything to gbrain. The agent dry-runs reference first, then decides.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
* feat(skillpack): migrate-fence + scrub-legacy-fence-rows (T5 + T16)
migrate-fence is the one-shot transition from the pre-v0.36 managed-block model.
Strips begin/end markers and the cumulative-slugs receipt comment; preserves
fence rows verbatim as user-owned routing during the transition to frontmatter
discovery. Receipt-then-row fallback (F-CDX-8) covers stale/missing receipts.
scrub-legacy-fence-rows is the opt-in cleanup after migrate-fence. Two-condition
gate: removes a row only when skills/<slug>/ exists AND that skill's frontmatter
declares non-empty triggers (proof frontmatter discovery covers it).
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
* feat(skillpack): harvest + privacy linter (T6 + T7)
The inverse loop: lift a proven skill from a host repo (~/git/wintermute, etc.)
back into gbrain so other clients can scaffold it. --from <host-repo-root> is
symmetric with scaffold's --workspace.
Security: symlink rejection + canonical-path containment (mirrors validateUploadPath).
Privacy: default-on linter scans harvested files against ~/.gbrain/harvest-private-patterns.txt
plus built-in defaults (Wintermute, email, Slack channel patterns). Any match
rolls back the copy and exits non-zero. --no-lint bypasses for the editorial
workflow after a manual scrub.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
* feat(repo-root): cwd_walk_up tier for non-OpenClaw hosts (T9 + D3)
autoDetectSkillsDir now walks up from cwd looking for any skills/ directory,
ahead of the implicit ~/.openclaw/workspace fallback. cd ~/git/wintermute &&
gbrain skillpack scaffold ... finds wintermute automatically without requiring
a RESOLVER.md/AGENTS.md to exist yet.
R5 regression preserved: $OPENCLAW_WORKSPACE still wins when explicitly set.
+5 test cases in test/repo-root.test.ts pin the new tier order and the R5 guard.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
* feat(skillpack): rewrite CLI dispatch, drop install + uninstall (T3 + T10)
skillpack.ts dispatcher rewritten for the v0.36 contract: scaffold, reference
(+ --apply-clean-hunks), migrate-fence, scrub-legacy-fence-rows, harvest, plus
the existing list / diff / check.
install and uninstall are gone — both exit non-zero with a hint pointing at
scaffold / migrate-fence. Clean break, no deprecated alias.
skillpack-check gains --strict for CI gating. When invoked as the subcommand
`gbrain skillpack check`, default is informational (exit 0 even with drift);
--strict opts back into the cron-friendly exit-1-on-issues behavior. Top-level
gbrain skillpack-check preserves its existing exit semantics for backwards compat.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
* feat(skills): skillpack-harvest editorial workflow + resolver wiring (T8)
The companion editorial skill for the gbrain skillpack harvest CLI. Walks the
genericization checklist (scrub fork names, generalize triggers, lift fork-
specific conventions to references) before the CLI runs. Routing-eval fixtures
use paraphrased intents to avoid the intent_copies_trigger lint.
Wires the new slug into openclaw.plugin.json#skills, skills/manifest.json, and
skills/RESOLVER.md.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
* test(skillpack): 9-case real-subprocess E2E flow (T11)
Spawns gbrain as a subprocess against tempdir workspaces. Covers: scaffold
first-run + re-run no-op, reference diff + --apply-clean-hunks, migrate-fence,
scrub-legacy-fence-rows, harvest privacy-lint catch + --no-lint bypass, and
the install removed-error path. No DATABASE_URL needed — skillpack is
filesystem-only.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
* chore: docs + VERSION + CHANGELOG for v0.36.0.0 (T13 + T14)
Skillpacks as scaffolding, not amber.
v0.36 retires the managed-block install model. Six new subcommands replace
install + uninstall: scaffold, reference (with --apply-clean-hunks), migrate-fence,
scrub-legacy-fence-rows, harvest, plus the existing list / diff / check
(check gains --strict for CI gating). Routing comes from each skill's
frontmatter triggers — gbrain does not touch your RESOLVER.md or AGENTS.md.
Companion editorial skill skillpack-harvest drives the genericization
checklist; default-on privacy linter catches Wintermute / email / Slack
references before they leak into gbrain core.
New docs guide at docs/guides/skillpacks-as-scaffolding.md walks the model
and the migration path for pre-v0.36 installs.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
* fix(ci): privacy checks — allow-list harvest-lint tests, scrub user-facing fork-name references
CI's check-privacy.sh and check-test-real-names.sh both flagged the literal
fork name across the v0.36 skillpack diff. Two failure modes, two fixes:
1. **Meta-rule-enforcement files** added to both allow-lists. The harvest
privacy linter's whole job is to catch the banned literal leaking into
gbrain; its source has the regex pattern, its tests verify the linter
fires by feeding it the banned string, and the skill markdown documents
the substitution policy. Same exception status as check-privacy.sh and
check-proposal-pii.sh themselves. Files allow-listed:
- src/core/skillpack/harvest-lint.ts
- test/skillpack-harvest-lint.test.ts
- test/skillpack-harvest.test.ts
- test/e2e/skillpack-flow.test.ts
- skills/skillpack-harvest/SKILL.md
2. **User-facing references** swapped for canonical phrasing per CLAUDE.md's
responsible-disclosure rule. README + new docs guide + 4 src docstrings
+ 1 test now say 'your OpenClaw' / 'host agent repo' / 'agentRepo' var
name. Behavior unchanged — only documentation strings touched.
Verify gate (the script CI runs) passes locally: EXIT=0.
Tests still pass: 60/60 across the affected files.
llms-full.txt regenerated.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
* fix(test): update check-resolvable-cli expectation for cwd_walk_up tier
Sister fix to the test/repo-root.test.ts update in commit
|
||
|
|
dd1cc121d8 |
v0.35.3.1 feat(eval): temporal-aware contradiction probe + verdict enum (#1052)
* rfc: temporal axis for contradiction probe Field report on residual HIGH findings from gbrain eval suspected-contradictions and proposal for a 4-phase fix (Phase 1 = judge prompt + verdict enum is the recommended starting point). Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * feat(eval): pass effective_date to judge prompt; bump PROMPT_VERSION Lane A1 of the temporal-contradiction-probe wave. Threads page-level effective_date through the search projection into the contradiction judge so the LLM can reason about supersession instead of treating every dated pair as a contradiction. Changes: - SearchResult interface adds optional effective_date + effective_date_source fields; rowToSearchResult populates them from the row data with date-only YYYY-MM-DD normalization (handles both postgres.js Date and PGLite string). - 8 SELECT projection sites (3 in postgres-engine, 5 in pglite-engine) now carry p.effective_date + p.effective_date_source through their inner CTEs and outer SELECTs so search results expose the field on both engines. - PairMember (eval-contradictions/types.ts) gets the two fields as required (string | null) so the type forces every constructor to think about temporal anchoring. Runner's searchResultToMember + takeToMember handle the normalization; takes inherit the chunk's page-level date. - buildJudgePrompt emits `Statement A (from: YYYY-MM-DD)` when effective_date is non-null, else `(date unknown)`. Prompt instructions explain the tag so the model knows what to do with it. - PROMPT_VERSION bumps '1' → '2'. Cache-key tuple shape unchanged; old rows miss naturally on first run against the new prompt. Test fixtures in 5 files updated to include the new required fields. All 205 eval-contradictions unit tests + 101 search-related tests pass. Typecheck clean. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * feat(eval): replace contradicts:boolean with verdict:enum (6 members) Lane A2 of the temporal-contradiction-probe wave. Expands the judge's classification vocabulary from a binary contradicts:bool to a six-member verdict enum so the probe can distinguish "this changed" from "this is wrong". Verdict taxonomy: no_contradiction — drop from findings contradiction — genuine conflict at same point in time temporal_supersession — newer claim updates/replaces older; not an error temporal_regression — metric/status went backwards over time (signal) temporal_evolution — legitimate change, neither supersession nor regression negation_artifact — judge misread an explicit negation Changes: - types.ts: Verdict union (6 members); Severity gains 'info'; ResolutionKind extended with temporal_supersede, flag_for_review, log_timeline_change; JudgeVerdict.contradicts → verdict; ContradictionFinding now carries verdict; ProbeReport adds queries_with_any_finding + verdict_breakdown (additive). - judge.ts: parseResolutionKind + parseVerdict guards; normalizeVerdict reads the new field and applies the C1 confidence floor only to verdict='contradiction' (the new verdicts are informational classifications, no floor). Prompt rubric rewritten to ask for verdict + extended severity scale. - severity-classify.ts: 'info' joins the rank with value 0; defaultSeverityForVerdict maps each verdict to its baseline severity (D7 — supersession=info, regression=high, etc.). parseSeverity gains a fallback param so consumers can override 'low' default. - auto-supersession.ts: classifyResolution + renderResolutionCommand handle the three new resolution kinds. Probe still NEVER auto-mutates — the new kinds render paste-ready commands or informational lines. - cache.ts: isJudgeVerdict shape check matches the new verdict field; old v1 rows fail the guard and treat as misses. - runner.ts: emit predicate at cache-hit and judge-success branches changes from `verdict.contradicts` to `verdict.verdict !== 'no_contradiction'`. Without this, the new verdicts vanish from the report. Added per-verdict tally + queriesWithAnyFinding alongside the strict queriesWithContradiction. - trends.ts: latest run verdict breakdown surfaces in the trend chart. Test fixtures updated across 8 test files. All 210 eval-contradictions unit tests pass. Typecheck clean. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * feat(eval): relax date-filter rule 3 when both sides dated Lane B of the temporal-contradiction-probe wave. The v1 date pre-filter skipped pairs whose chunk-text-extracted dates differed by >30 days as a cost-saving heuristic. That heuristic silently killed exactly the cases the new verdict taxonomy exists to surface — role transitions across years (e.g. a 2017 historical record vs. a 2025 current state), MRR claims years apart, status changes recorded over time. Lane A1+A2 made temporal supersession explicit and cheap to classify. The filter no longer needs to skip these pairs; the judge can label them. Changes: - date-filter.ts: shouldSkipForDateMismatch accepts optional effectiveDateA and effectiveDateB. When BOTH are non-null, returns skip=false with the new 'both_have_effective_date' reason — the judge will see the dates via the (from: YYYY-MM-DD) prompt tag from Lane A1. Other rules (same-paragraph dual-date override, missing-date fallback) preserved verbatim and still run first. - runner.ts: threads pair.{a,b}.effective_date into the date-filter call. Pairs that previously vanished into the skip bucket now reach the judge. Tests (R1 IRON RULE regression suite, 6 new cases): - both sides effective_date → not skipped - both sides effective_date overrides >30d chunk-text rule - rule 1 (same-paragraph dual-date) still wins over effective_date relaxation - rule 2 (missing chunk dates) still applies when effective_date partially present - undefined effective_dates fall through to v1 behavior (back-compat) - empty-string effective_date treated as missing (only real dates enable the relaxation) Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * feat(cli): cost-estimate prompt + --budget-usd + Haiku routing Lane C of the temporal-contradiction-probe wave. Three layers of cost guardrail, all stacked: (a) cost-estimate prompt at probe-run-time. Before the runner spends any tokens after a PROMPT_VERSION change, eval-suspected-contradictions reads the most recent persisted prompt_version from eval_contradictions_runs and compares. When they differ: - TTY: prints an upper-bound estimate + Ctrl-C window (default 10s, override via GBRAIN_PROBE_PROMPT_GRACE_SECONDS). - non-TTY: prints the estimate + auto-proceeds (autopilot path). - --yes override or GBRAIN_NO_PROBE_PROMPT=1: skip entirely. Mirrors the v0.32.7 runPostUpgradeReembedPrompt pattern. (b) --budget-usd N hard cap (pre-existing; PreFlightBudgetError surfaces when the estimate alone exceeds the cap, and CostTracker halts the run mid-flight when cumulative cost exceeds it). Documented in the help text alongside (a). (c) Judge model now routes through resolveModel() with configKey 'models.eval.contradictions_judge', tier 'utility' (Haiku-class default), and env var GBRAIN_CONTRADICTIONS_JUDGE_MODEL. The legacy --judge CLI flag still wins as the highest-precedence override. Doctor's model touchpoint registry (src/commands/models.ts:50) carries the new key so `gbrain models` and `gbrain models doctor` surface it. Also in this lane: - CLI: --severity accepts 'info' (the new Severity member from Lane A2). - CLI: --severity output shows [verdict] tag alongside slug pairs so operators distinguish genuine contradictions from temporal classifications. - Human summary: prints the new queries_with_any_finding metric and the per-verdict breakdown table. - Help text: explains the cost-prompt + budget-cap + model-routing interactions in one paragraph. New tests (9 cases on the cost-prompt helper): - --yes override skips - GBRAIN_NO_PROBE_PROMPT=1 skips - prompt_version unchanged → skips - non-TTY auto-proceeds with stderr note - TTY proceeds after grace - TTY aborts on Ctrl-C - fresh brain (no prior runs) fires the prompt - GBRAIN_PROBE_PROMPT_GRACE_SECONDS override honored - estimate banner contains query count + judge model + dollar amount All 225 eval-contradictions tests + 25 model-config tests pass. Typecheck clean. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * test(eval): R4/R5/R6 IRON-RULE regressions for the verdict-enum wave Lane D of the temporal-contradiction-probe wave. The Lanes A1/A2/B/C lanes landed the behavior; this lane pins the regressions that protect the wave against future drift. R4 (runner emit predicate): five new tests, one per non-no_contradiction verdict, prove the runner.ts emit rule surfaces each one as a finding with the correct verdict tag, and that: - queries_with_contradiction (Wilson-CI denominator) ONLY counts verdict ='contradiction' — the strict metric is preserved - queries_with_any_finding counts every non-no_contradiction verdict - verdict_breakdown tallies correctly Plus one negative case: verdict='no_contradiction' produces zero findings. Without R4, a future runner refactor could collapse the new verdicts back to /dev/null and the report would silently shrink. R5 (cache key shape): direct shape assertion on buildCacheKey output. The key tuple is exactly 5 fields (chunk_a_hash, chunk_b_hash, model_id, prompt_version, truncation_policy). Adding a 6th field would silently break every operator's brain (no migration path). R6 (contradiction severity unchanged): four tests on normalizeVerdict pin the legacy semantics — judge-supplied severity wins (whether 'high' or 'low'), and on garbage severity input the fallback is 'medium' (per defaultSeverityForVerdict('contradiction')) NOT 'low'. The contradiction verdict's severity must never default to 'low', which would silently mask genuine conflicts as cosmetic naming issues. The temporal_regression case is included for parity (garbage → 'high' since regressions are real investor red flags). 236 eval-contradictions tests pass (211 + 6 R4 + 1 R5 + 4 R6 + 9 cost-prompt from Lane C). Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * feat(ci): privacy lint for docs/proposals/*.md Captures the residual TODO from the temporal-contradiction-probe wave's plan: prevent the bug class where an RFC lands in docs/proposals/ with PII that should never appear in a public technical artifact. The original RFC had to be scrubbed at force-push time (Step 0); this lint catches the same patterns at CI time so the next one can't slip through. Sibling to scripts/check-privacy.sh: - check-privacy.sh: bans the literal "Wintermute" repo-wide. - check-proposal-pii.sh: focuses on docs/proposals/*.md and the OTHER PII classes — personal-relationship vocabulary, private repo refs. Design contract: the denylist names PATTERNS, not real people. Naming specific real names (deceased relatives, therapist first names, dealflow contacts) inside this script would leak PII into the repo just by appearing here. The structural patterns below catch the SURROUNDING vocabulary that always accompanies such content in personal RFC prose. Trade-off: a future RFC that names a real person without any contextual markers won't be caught — accepted as residual risk handled by human review. Patterns flagged in docs/proposals/*.md: - garrytan/brain (private repo reference) - trial separation, permanent separation - couples session, couples therapist - divorce attorney(s) - grandmother's funeral, aunt's funeral - wintermute (also caught by check-privacy.sh; listed here for proposal-scoped clarity) Bare common words (separation, funeral) are NOT banned — only the combined personal-context phrases. "Separation of concerns" and other software vocabulary survives. Wired into: - `bun run verify` (gates every push) - `bun run check:all` - `bun run check:proposal-pii` (standalone) Tests: 15 cases in test/scripts/check-proposal-pii.test.ts. - Each pattern flagged when present, plus exit-code + stderr signal. - Two negative cases (separation-of-concerns, funeral metaphor) prove the lint doesn't false-positive on legitimate software prose. - No-proposals-dir → exit 0 (not a failure). - Multi-hit case proves all patterns surface together with a summary count. - The two test fixtures that name "Wintermute" / "WINTERMUTE" as sentinel literals are allowlisted in check-test-real-names.sh per the same meta-rule-enforcement exception as check-privacy.sh itself. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * chore(privacy): allowlist new privacy-guard files in check-privacy.sh check-privacy.sh bans the literal Wintermute repo-wide. The two new files from the v0.34 privacy lint (scripts/check-proposal-pii.sh and its test) necessarily name the token to do their job. Same meta-rule-enforcement exception as scripts/check-privacy.sh itself, scripts/check-test-real-names.sh, test/recency-decay.test.ts, and the existing entries — describing what the rule forbids requires naming it. Without this allowlist, `bun run verify` fails on check:privacy. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * chore: bump version and changelog (v0.35.1.0) Temporal-contradiction-probe wave — Phase 1 of the RFC at docs/proposals/temporal-contradiction-probe.md. Headline: the contradiction probe now classifies pairs into a 6-member verdict enum (no_contradiction, contradiction, temporal_supersession, temporal_regression, temporal_evolution, negation_artifact) and sees the page-level effective_date for each chunk via a (from: YYYY-MM-DD) tag in the prompt. The pre-judge date filter no longer skips dated wide-gap pairs, so the role-transition class (e.g. a 2017 historical record vs. a 2025 current state) reaches the judge and gets classified as temporal_supersession instead of vanishing into the skip bucket. PROMPT_VERSION bumped 1 → 2 (cache fully invalidated). Three-layer cost guardrail: TTY-only cost-estimate prompt with Ctrl-C window, --budget-usd hard cap, Haiku-tier routing via new models.eval.contradictions_judge config key. Also adds a CI privacy lint (scripts/check-proposal-pii.sh) wired into bun run verify that catches PII patterns in docs/proposals/*.md so future RFCs can't ship with personal-context vocabulary the way this wave's source RFC did at draft time. Phases 2-4 deferred to follow-up RFCs per the plan. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> --------- Co-authored-by: garrytan-agents <garrytan-agents@users.noreply.github.com> Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com> |
||
|
|
bd2fe8a1fa |
v0.32.5 feat: gbrain-context OpenClaw context engine — deterministic temporal/spatial injection (#880)
* feat: gbrain-context OpenClaw context engine — deterministic temporal/spatial injection Adds a context engine plugin that runs on every assemble() call to inject structured live context into the system prompt: - Garry's current local time (computed from heartbeat-state.json timezone) - Current location (city + timezone from heartbeat or flight data) - Home time when traveling (e.g. 'Mon 7:58 AM PT') - Active travel status - Quiet hours detection - Airport→timezone mapping for 30+ airports This kills the 'time warp' bug class where compacted sessions lose track of time/location. The engine delegates compaction to the legacy runtime and only owns systemPromptAddition injection. Zero LLM calls, <5ms. Files: - src/core/context-engine.ts — engine implementation (SDK-free, testable) - src/openclaw-context-engine.ts — plugin entry point (requires SDK) - test/context-engine.test.ts — 9 tests, all passing Enable: plugins.slots.contextEngine = 'gbrain-context' * feat: add activity injection — calendar events + open tasks in context block Reads memory/calendar-cache.json and ops/tasks.md to inject: - **Right now:** current meeting (with attendees) from calendar - **Coming up:** next 3 events within 4-hour window - **Open tasks:** unchecked items from Today section - Stale calendar warning when cache is >6 hours old Skips all-day events and generic markers (Home, OOO, Out of Office). Caps upcoming events at 3 and tasks at 5 to keep prompt lean. 15 tests passing (was 9). * v0.32.5 feat: gbrain-context OpenClaw context engine — deterministic temporal/spatial injection Ships PR #873 by @garrytan-agents (two underlying commits preserved): - |