Three CI-only check failures, two root causes:
1. windowTurnCount ignored GBRAIN_RETRIEVAL_REFLEX_WINDOW_TURNS when
loadConfig() returned null (no config file AND no DATABASE_URL — a clean
CI shard with no brain). loadConfig drops its env→config mapping in that
case, so the documented escape hatch silently died and the window fell
back to 4 → windowed extraction widened when the test set window=1 →
prior-turn entity leaked. Fixed: read the env var directly in
windowTurnCount, mirroring reflexEnabled's direct process.env read. This
is a real product bug, not just a test artifact — any config-less host
using the env hatch was affected. Regression test pins it.
2. sync-cost-preview + doctor-federation-health failed only IN-SHARD: a
sibling test configured a non-legacy (ZeroEntropy 1280-d / $0.05) gateway
and never reset it. The legacy-embedding preload only restores the
OpenAI/1536 default when the gateway slot is EMPTY, so a non-empty foreign
config survives into the next file — and a file's beforeAll runs BEFORE
the preload's restoring beforeEach, so federation-health built a
vector(1280) column and its 1536-d fixture hit CheckExpectedDim. My new
test files reshuffled the deterministic file→shard assignment, exposing
this latent ordering bug. Hardened both victims to establish the gateway
state they assert (sync-cost-preview resets to the unconfigured fallback;
federation-health pins legacy 1536 before initSchema) so they're
order-independent. Verified against a simulated leaker run before them.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Four red-team findings on the #2095 push-context surface:
- RT1 starvation: watch's session-dedupe Set filtered AFTER volunteerContext's
cap, so a recurring already-pushed entity burned cap slots every turn and
starved fresh pages behind it. VolunteerOpts.excludeSlugs now skips inside
the pointer loop BEFORE the confidence gate and the cap.
- RT3 honest stats: reflex-channel event logging moved from inside the
resolver to the DELIVERY point — serve's resolve-IPC onDelivered hook fires
only after the response write succeeds, and buildReflexAddition logs only
after the per-turn timeout admits the block. A block the client's 250ms
budget abandoned was never injected and no longer counts as volunteered.
(logChannel resolver opt removed; logDeliveredReflexPointers is the seam.)
- RT5 unbounded window: --window-turns is clamped to [1, 64] so a config typo
can't reintroduce the re-scan-everything-per-turn cost class.
- RT2/RT4 documented + filed: PGLite watch connection monopoly (WATCH_HELP,
push-context guide, TODO to route watch via serve IPC); host-resolver
suppression contract at ResolveEntitiesFn (TODO for a capability gate).
Tests: starvation guard (watch + volunteerContext unit), window clamp floor +
ceiling, delivery-side logging (helper writes channel=reflex through the
drained sink; bare resolver writes nothing; empty list no-op), IPC wiring test
rewired to onDelivered. KEY_FILES.md + push-context.md updated; build:llms run.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Five specialist reviewers (testing/maintainability/security/performance/
data-migration) on the reconciled diff; every finding applied:
Performance: the alias arm now resolves all granted sources CONCURRENTLY
(a federated caller paid M sequential RTTs per turn — ~355ms at 5 sources
cross-region, inside the reflex's 1.5s budget); watch's session dedupe is
O(1) Set membership instead of a monotonically growing priorContext string
(O(T²) over a long-lived session); getWindowTurns iterates from the tail
(per-turn cost no longer grows with session length); the resolver's
provenance maps fold into the existing candidate pass.
Security: volunteer_context clamps caller-supplied attribution at the trust
boundary — session_id capped at 256 chars (a read-scoped token could bank
~1MiB TEXT per request, retained 90 days), turn logged only when a safe
integer (a non-integer threw inside the batched INSERT and silently dropped
the whole batch). The privacy comments now state precisely what rationale
may contain (the matched entity's surface form — which by construction
resolved to an existing alias/title/slug — never free conversation text).
Maintainability: TURN_PREFIX_RE + formatVolunteeredPage exported from
volunteer.ts and shared by watch/cli (the two surfaces can no longer
drift); volunteerEventRowsFrom is the single VolunteerEventRow assembly
site for all three channels; watch's window default now honors the same
retrieval_reflex_window_turns config knob the reflex reads; the stale
pre-v116 comments swept to pre-v117.
Testing: the two flake-class CRITICALs fixed (pipe test asserts the
backstop banner instead of a cold-CI-hostile 9s wall bound; the SIGINT test
waits on watch's new machine-readable ready line instead of a fixed 15s
sleep — 2.5s and deterministic now); new coverage for the sink's timeout
branch + ghost-reference drop, watch per-turn fail-open, untrusted knob
clamps (min_confidence/max_pages/days), window-cap ordering (newest user
mention survives), serve-IPC suppression passthrough + channel=reflex
logging, windowTurnCount edge semantics, and structural pins for the sink
registration + cycle purge wiring. The exit-verdict pin's grep is now
operator/whitespace-tolerant.
Deferred with TODOs: resolver index shapes for the per-turn query;
batched first-prune after a long dream-cycle gap.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Ship coverage audit (85%, gate PASS) named five gaps; the two substantive
cheap ones close here: the codex-D11 logChannel='reflex' path now has a
behavioral pin (events land on channel 'reflex' through the drained sink;
no logChannel → no events), and gbrain watch's --window-turns / --max-pages
flags are exercised (turn-1 attribution under window=1; cap to one page).
Remaining flagged-not-blocking: the wallclock-timeout branch (untestable
without >10s real-clock flake — same rationale as the arming pin),
formatResult's volunteer case (module-private), and the cycle purge wiring.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The default-on retrieval reflex now extracts entities from the last N turns
(retrieval_reflex_window_turns, default 4; env
GBRAIN_RETRIEVAL_REFLEX_WINDOW_TURNS; window=1 reproduces the legacy
current-turn-only behavior exactly). assemble() passes the recent
user/assistant turns (hard cap 12); the reflex slices to the configured
window. Assistant-introduced entities and "what did she invest in?"
follow-ups whose antecedent was NAMED in the window now surface pointers —
the issue's "zero agent-initiated queries" success criterion on the
ambient path.
Under windowing, suppression switches to slug-only (codex D7): the legacy
title-whole-word rule would suppress every entity merely MENTIONED in a
prior window turn, breaking the feature by construction. Slugs only enter
prior context when a pointer/page was actually surfaced, so
already-surfaced pages still suppress. The suppression mode flows through
all three resolver rungs (host opts, serve IPC request, direct Postgres).
Ambient-channel feedback (codex D11): the server-side resolver paths
(serve IPC + direct Postgres) log volunteered pointers with
channel: 'reflex' through the drained volunteer-events sink, so
`gbrain volunteer-context --stats` measures the default-on path where most
volunteering happens. Host-injected resolvers (no gbrain engine) can't
log — documented gap. Precision gates, 1.5s ceiling, fail-open, and the
pointer cap are unchanged.
Tests: prev-assistant-turn entity fires; window=1 legacy parity; slug-only
vs already-surfaced suppression; throwing resolver stays fail-open
(16 green).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* fix(extract,ingest): well-form lone UTF-16 surrogates before jsonb (#2011)
excerpt() in link-extraction.ts sliced the link-context window by raw UTF-16
index, so a boundary landing inside a non-BMP char (emoji, math, CJK) left an
unpaired surrogate half in `context`. Serialized to JSONB for the
jsonb_to_recordset batch insert, Postgres rejects it at the ::jsonb cast and
aborts the whole batch — wedging `extract --stale` because the staleness
bookmark only advances on a clean finish.
- text-safe.ts: new ensureWellFormed() (Bun isWellFormed/toWellFormed) — one
shared surrogate-cleaning primitive.
- link-extraction.ts: excerpt() well-forms the slice (root-cause fix).
- batch-rows.ts: new sanitizeForJsonb() = ensureWellFormed(stripNul(s)) applied
to every free-text body field (link context; timeline summary/detail/source;
take claim/source). Identity/security fields stay un-sanitized and fail closed.
- postgres-engine.ts + pglite-engine.ts: scalar addLink + addTimelineEntry use
sanitizeForJsonb too, matching the batch path on both engines.
- brainstorm/orchestrator.ts: consolidate hand-rolled sanitizeUnicode onto
ensureWellFormed (also fixes consecutive-lone-surrogate mishandling).
Tests: ensureWellFormed unit cases (incl. consecutive lone surrogates), an
excerpt window-split regression, PGLite + Postgres-e2e surrogate cases across
all free-text fields and scalar paths, and fail-closed identity-field tests
proving sanitization was NOT extended to slugs/holders.
* v0.42.39.0 chore: bump version and changelog (#2011)
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* docs: update project documentation for v0.42.39.0
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* v0.42.40.0 chore: re-slot release version (was 0.42.39.0)
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* test: fix env-mutation isolation violations in retrieval-reflex tests
check:test-isolation (rule R1) flags direct process.env mutation in
non-serial test files — bun's parallel runner loads multiple files into
one process, so a leaked GBRAIN_RETRIEVAL_REFLEX flips reflex behavior
in unrelated tests. Both files landed via the #2019 merge; convert the
beforeEach/afterEach env juggling to the canonical withEnv() wrapper,
which restores the prior value via try/finally even on throw.
Fixes the failing `verify` CI check on #2031 (and the `test-status`
aggregate that inherits it). All 30 verify checks green locally.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(integrations): parameterize resolver-row fence by recipe id
The install fence was hardcoded gbrain:agent-voice:resolver-rows, so any
second copy-into-host-repo recipe wrote a mislabeled block (and refresh/
uninstall keyed on recipe id would miss it). Derive it from manifest.recipe.
* feat(context): Retrieval Reflex — teach the agent when/what to retrieve (#1981)
Deterministic per-turn pointer layer in the context engine: a zero-LLM,
precision-biased scan resolves salient entities (names, @handles) to existing
brain pages and injects compact pointers (name → slug → safe synopsis). Detect
+ point, never auto-dump. Fail-open, capped, suppression on prior context only.
Engine-aware resolver ladder (no second DB connection): host ctx.brainQuery →
PGLite serve resolve IPC (unix socket) → Postgres cached direct → disabled.
Synopsis runs through get_page's privacy strip. Plus the retrieval-reflex recipe
+ policy skill, the retrieval_reflex_health doctor check, config gate, and the
init next-step hint.
* chore: bump version and changelog (v0.42.38.0)
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* docs(architecture): KEY_FILES entries for Retrieval Reflex surface (#1981)
Document the new src/core/context/ modules, the context-engine resolver
ladder, the serve resolve IPC, the retrieval_reflex_health doctor check,
and the recipe-id-keyed install fence. Current-state only.
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
