Garry's Opinionated OpenClaw/Hermes Agent Brain
TypeScript
chore: bump version and changelog (v0.60.466.0)
garrytanf34f576496e9
diff --git a/CHANGELOG.md b/CHANGELOG.mdindex aa5bd03509..52b4651a2e 100644--- a/CHANGELOG.md+++ b/CHANGELOG.md@@ -10,6 +10,37 @@ credits are retained; no result has been reassigned to another provider. Originaidentifiers and attribution are available in the pre-removal Git revision`6040075c6cb95be5881cc2e1b76ef7d71f4e5d29` (retained on 2026-09-23).+## [0.60.466.0] - 2026-10-26++**A brain whose database is at schema v265 now upgrades instead of stopping with `column "pending" does not exist`, and on a brain with a large write journal each write claim is a quick index lookup again instead of a scan of the whole journal after every restart.**++Two upgrade problems are fixed. First, a brain at schema v265 could not reach any later release: the upgrade replays the current schema before running migrations, and one index in it filters on a column that only the next migration adds, so the replay failed before any migration ran and every retry failed the same way. Second, after a restart the write queue checked "has a foreground write committed on this worktree since I last looked?" with no index to answer it, so on a brain with hundreds of thousands of past write receipts each claim walked every receipt. Claims could overrun their claim phase, the queue grew, and sync looked stuck.++### What changed++| Situation | Before | Now |+|---|---|---|+| Upgrading a brain at schema v265 (Postgres or PGLite) | every upgrade attempt failed with `column "pending" does not exist` | the upgrade adds the missing column first (exactly as migration v266 defines it, no data changes), then migrates to the latest schema |+| A write claim right after a restart, on a large journal | each claim scanned the journal (about 175 ms per claim on a test brain with ~590k receipts) | migration v276 adds the partial index `persistence_requests_foreground_committed`; the same claim takes under 1 ms |+| The new index is missing or INVALID on a brain already at the latest schema | — | `gbrain doctor` (`persistence_request_indexes`) names it and says write claims can time out; `gbrain repair request-indexes` previews the fix and `--apply` rebuilds it |++### What to watch for after you upgrade++- On Postgres, migration v276 builds the index with `CREATE INDEX CONCURRENTLY`, so writes keep flowing; on a very large journal the build can take a while. If it is interrupted, `gbrain doctor` reports the INVALID index and `gbrain repair request-indexes --apply` rebuilds it. PGLite builds it inline.+- The index also covers compacted receipts of every kind, so on a long-lived brain it holds most committed receipts.+- Nothing rewrites your data. The v265 fix only adds a column with its default.++### Itemized changes++- `src/core/engine-sql/bootstrap.ts`: the forward-reference bootstrap probes `derived_generation_items` and its `pending` column and adds the column on a v265 brain before the schema replay (#7645).+- `src/core/persistence/schema.ts`, `src/core/schema-migrations/v276-persistence-foreground-committed-index.ts`: `persistence_requests_foreground_committed` on `(worktree_id, sequence)`, with the write claims' committed-foreground predicate written the way the claim probe writes it; built online by v276 (transaction-free, `CONCURRENTLY` on Postgres, an INVALID leftover rebuilt) and listed among the request-table online indexes (#7649).+- `src/core/repair/request-indexes.ts`, `src/commands/doctor/checks/persistence-requests.ts`: repair and doctor cover the new index; the doctor message names write claims.+- `test/fixtures/schema-history/`, `scripts/capture-schema-history-fixture.ts`: a fresh brain as each release from schema v260 to v275 created it, for both engines, captured from that release's own code with a round-trip fidelity check. `test/schema-history-upgrade.slow.test.ts` (PGLite) and `test/e2e/schema-history-upgrade.test.ts` (Postgres) upgrade every fixture to the latest schema and compare the result with a fresh install, so the next missing forward reference fails a test instead of a user's upgrade.+- `test/schema-bootstrap-coverage.test.ts`: the bootstrap coverage check now reads columns in partial-index `WHERE` predicates and `INCLUDE` lists.+- Tests for the index: `test/migration-persistence-foreground-committed-index.test.ts`, `test/e2e/persistence-claim-committed-probe-7649-postgres.test.ts` (both claim paths use the index on a 160k-receipt Postgres journal).++Reported by @glen-carlson (#7645) and @andreineacsu (#7649). Contributed by @Masashi-Ono0611.+## [0.60.465.0] - 2026-10-26**On a managed brain, releasing a held fact (`gbrain trust release h<id>`) now writes it into its entity page's `## Facts` table through the page-write door, so the released fact lives in Git and a re-import keeps it, instead of a database-only row the file never had.**diff --git a/VERSION b/VERSIONindex 20926ae007..58fc6572a2 100644--- a/VERSION+++ b/VERSION@@ -1 +1 @@-0.60.465.0+0.60.466.0diff --git a/package.json b/package.jsonindex 8b8c249403..a2d97e440a 100644--- a/package.json+++ b/package.json@@ -246,7 +246,7 @@"bun": ">=1.4.0"},"license": "MIT",- "version": "0.60.465.0",+ "version": "0.60.466","overrides": {"@ai-sdk/provider-utils": "4.0.56","@hono/node-server": "^2.0.5",