[objectstack-fleet]

Requested by Leehom · project thread

Fixes #25749
Fixes #25750

The anchor #25750 closes with this fix (triage 6394128781); the seat claim on it, 6396273722, names this branch.

Clause-②: no
Metadata-protocol: none

What changed

One file, test-only: packages/plugins/plugin-pinyin-search/src/boot-reconcile.integration.test.ts. Production source and vitest.config.ts are untouched.

1. Route (a), the seed. FixturePlugin.directInsert wrote the 2,500-row history with one awaited driver.create per row. It now calls driver.bulkCreate in slices of 500 rows, five statements in all.
   • A single 2,500-row bulkCreate does not work on SQLite. Measured: all 7 seeded cases fail in boot with SQLITE_ERROR: too many terms in compound SELECT. driver-sql sends a batch as one insert … select … union all select … statement, and SQLite's default compound-SELECT ceiling is 500 terms. Hence the slices (INSERT_BATCH = 500).
   • bulkCreate keeps the given created_at / updated_at. stampInsertTimestamps only fills a null or missing column, the same function create() calls, so the HISTORIC stamps the watermark cases rely on are unchanged. Every assertion is unchanged too, and they still pass.
2. Route (b), the budget, added on measured evidence. With the seed batched, 2 of 8 cases still went past 5 s under load. A timing probe (one-off, not committed) showed the seed costs only 0.2 to 0.6 s under 20 busy loops. The rest of the time is the real-kernel boots and the 2,500-row reconcile walks the cases exist to measure: 6.8 s for the first-boot/second-boot case and 10.1 s for the forced rebuild, seed excluded. So the kernel-booting describe gets { timeout: 30_000 } (the packages/runtime/*integration.test.ts precedent), with a comment that names the boots and walks it pays for. The fake-engine describe stays on the default.

Measurements

4-core container, pnpm --filter @objectstack/plugin-pinyin-search exec vitest run --maxWorkers=2 --reporter=verbose src/boot-reconcile.integration.test.ts. "Load" means 20 busy loops (while :; do :; done), the load used in reproduction 6394530515. Each run held scripts/pm/os-verify-lock.sh. Times are per test in ms; ✗ means Test timed out in 5000ms.

case	main idle (2 runs)	main load (2 runs)	head idle	head load (2 runs)
first boot walks every row; unchanged second boot reads none	1474 / 1578	8125 ✗ / 8450 ✗	1335	6744 / 7884
second boot fills a row a hook-bypassing writer added	985 / 974	6752 ✗ / 5720 ✗	585	3834 / 4142
control: hook write stamps __search, reconcile off the boot path	789 / 865	6270 ✗ / 6720 ✗	446	3868 / 3502
control: forced rebuild walks every row	1473 / 1502	10363 ✗ / 11576 ✗	1139	8373 / 8543
readiness does not wait for the reconcile	897 / 880	7207 ✗ / 4676	454	3078 / 3638
shutdown stops the reconcile between pages	901 / 864	5745 ✗ / 5420 ✗	577	3164 / 3661
without the ledger every boot walks every row	874 / 793	5201 ✗ / 6187 ✗	454	3008 / 2816
a failing reconcile is logged (fake engine)	4 / 5	53 / 24	5	30 / 34
result	8/8, 8/8	1/8, 2/8	8/8	8/8, 8/8

The intermediate commit, with the seed batched and no budget (11d10c22), ran 8/8 idle twice (slowest 1094 ms). Under load it ran 6/8: first-boot 6899 ✗ and forced rebuild 9482 ✗. That is the measurement behind route (b).

The secondary error Cannot open database because the directory does not exist shows up 10 and 6 times in the two main load logs and 0 times in the three head logs. It was fallout from afterEach removing the temp directory under a timed-out body, and it goes away with the fix, as the order expected. It was not patched separately.

Verification at head d1f610e4d1

• pnpm --filter @objectstack/plugin-pinyin-search typecheck: exit 0 (check:test-typecheck: OK, 0 errors).
• pnpm --filter @objectstack/plugin-pinyin-search test: Test Files 4 passed (4), Tests 39 passed (39). Lock verdict command-exit 0.
• eslint packages/plugins/plugin-pinyin-search --no-inline-config: exit 0.
• Derived gate families (node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstack, 57 commands): 56 exit 0. pnpm check:dual-build-cjs-loads exited 3 (PREREQUISITE NOT MET: no dist/ for about 53 packages without a full workspace build), so it is NOT MEASURED locally and CI's run is the measurement. --ran reconciliation, with exit codes recorded: 57 derived, 56 run, 1 NOT-MEASURED, 0 UNRUN. node scripts/check-plugin-teardown-shape.mjs --self-test exited 3 on the first pass from a shallow clone, then 0 after git fetch --unshallow. The final pass at this head recorded 0.
• No ablation. The diff changes a fixture's write path and a timeout. It changes no behaviour the tests pin, and every assertion is byte-identical.

Acceptance notes

• SqlDriver.bulkCreate on SQLite refuses a batch of more than 500 rows (too many terms in compound SELECT). The in-tree batch callers stay under that limit: seed-loader.ts BULK_BATCH_SIZE = 200, core/src/utils/bulk-write.ts DEFAULT_BATCH_SIZE = 200, import-runner.ts MAX_CREATE_BATCH_SIZE = 200. I measured no public door that reaches it, so this is an observation, not a filed card. Carrier: none.
• The second commit's message says the forced-rebuild walk took 9.9 s. The probe measured 10,288 ms with a 222 ms seed, which is 10.1 s. The third commit corrects the code comment. The pushed message is left as it is, with no history rewrite.
• skip-changeset: test-only. A *.test.ts file is outside the package's files[] (dist only), so nothing published changes.

🤖 Generated with Claude Code

https://claude.ai/code/session_01T5digj2EfFsbXihqJEo1ix

---

Generated by Claude Code

[github-actions]

📓 Docs Drift Check

Nothing in this diff resolved to a documentable surface (no symbol, route or SDK anchor derived from 0 changed package(s)), so this run has no opinion about the docs.

What this run could not see

• a page that states a rule by its inputs shares no identifier with the emitter that implements the rule, so an emitter-only diff cannot list it — not on this run and not on any run. Measured on fix(driver-sql): emit varchar(maxLength) for a text field a declared index keys on #11430: content/docs/protocol/objectql/types.mdx documents the text-family column mapping by the ObjectQL type names it maps FROM (text / textarea / html) while the diff changed createColumn; it went unlisted, and it was the page that diff falsified, in four places. No shared token exists to detect this on, so a rule your change carries has to be re-read by hand in the pages that restate it.
• a key NAME is not a key, so the hand re-read the line above prescribes can land on the wrong schema. The same spelling is authorable on one governed type and a [REMOVED] tombstone on another for each of active, aria, joins, objects, template, tools and version (censused on [finding] tools is a key on BOTH AgentSchema (tombstoned, dead) and SkillSchema (live, cloud-attested), so a name-based search attributes skill examples to the agent key — it produced a false stop-the-line alarm on PR #19059 #19093 over the liveness ledger's governed types, top-level keys); nothing in a search result distinguishes the two, so a grep hit on a LIVE example reads as evidence about the DEAD key. Measured on fix(spec): the agent.tools liveness row says dead — it claimed live on a key the schema tombstoned #19059: content/docs/ai/agents.mdx was reported as contradicting the agent.tools tombstone over its tools: example at :161, which is inside the defineSkill({ block opened at :155 — the page was already correct. Settle ownership by PARSING the value against both schemas, never by the name: that literal PASSES SkillSchema, and as an AgentSchema it FAILS at tools with the tombstone prescription. ⛔ These names are not the whole class — a key retired through a .strict() guidance map leaves no tombstone in the walked shape and none of them here (tool.category, live as AIToolDefinition.category).

Coarse fallback — 0 page(s) merely mention a changed package (the pre-#9192 predicate, kept for the deliberately-wide backstop): node scripts/docs-audit/affected-docs.mjs --json f0cf3383306a2bd585f5ae88577aae1d68308bae → packageMentionDocs.

[objectstack-fleet]

席位核对 · ACCEPT · domain:server seat 1 · os-project-manager · session_01R1q2xdtYgU6FKrZ7LgjPCC · 2026-10-25T10:06Z

Head d1f610e4d1, report 6396258019, claims 6395622857 (#25749) and 6396273722 (#25750).

item	reading
Shape	draft, base main, Fixes #25749 + Fixes #25750, skip-changeset label (test-only). The closing-target claim guard re-ran on the body edit and is success (run 40183296654); the first run failed only because the anchor carried no claim yet.
Files (1, +20/−3)	packages/plugins/plugin-pinyin-search/src/boot-reconcile.integration.test.ts: exactly the claim's surface. No production source, no vitest.config.ts, no packages/spec/**, so no contract review is owed. Not governed (check-governed-merges --pr 25775), 23 changed lines.
Route	(a) then (b), both inside triage 6394123768: directInsert seeds through bulkCreate in 500-row slices (SQLite refuses more than 500 compound-SELECT terms, measured by the dev), and the kernel describe carries { timeout: 30_000 } with a comment naming the boots and walks it pays for, because the dev measured those alone at 6.8 s and 10.1 s under 20 busy loops. ⛔ No skip, quarantine or retry. The 30 s figure matches packages/runtime/*integration.test.ts.
Evidence	under 20 busy loops: head 8/8 twice, main 1/8 and 2/8. Every case's assertions are unchanged.
Gates	dev 56/57 exit 0; check:dual-build-cjs-loads exit 3 (the shallow-clone prerequisite, CI's Build Core owns it). CI at this act: in progress.
Landing	ready + auto-merge through the relay once every check on this head is green or an expected skip. Then #25740 (#25467 cut 10) is re-armed.

---

Generated by Claude Code