- Create packages/store/ with log-store, log-archive, blob-store (~900 LOC)
- daemon depends on @uncaged/nerve-store (workspace:*)
- CLI depends on @uncaged/nerve-store, delete daemon-types.ts
- Move store-related tests to packages/store/src/__tests__/
- All store tests pass (73/73), no new regressions
- Filter __start__ messages in getThreadRoundCount SQL to fix round offset
- Remove duplicate parseWorkflowField, use parseSenseWorkflowDirective
- Remove unnecessary double casts in workflow CLI
- Add runtime validation for Role meta in workflow-worker
- Export DEFAULT_ENGINE_MAX_ROUNDS from types.ts
小橘 🍊（NEKO Team）
- Replace loose payload types with WorkflowLaunchParams { prompt, maxRounds }
- Add SenseResult.workflow field with pipe-separated format (name|rounds|prompt)
- Add parseWorkflowField utility and routeSenseComputeOutput in @nerve/core
- Integrate sense→workflow routing in kernel
- Remove deprecated workflow reflex kind from ReflexScheduler
- Update all test files to use new type-safe interfaces
小橘 🍊（NEKO Team）
Implements the workflow thread CLI command that retrieves
workflow execution context (logs, events, state) for a given run.
- Add 'nerve workflow thread <runId>' subcommand
- Add log-store query API in daemon
- Add tests for CLI and log-store
- Export new daemon types for thread data
小橘 <xiaoju@shazhou.work>
The original code only accepted object-type triggerPayload, silently
discarding string values by replacing them with {}. This meant
`nerve workflow trigger <name> --payload '"some string"'` would
lose the payload entirely.
Changed to `triggerPayload ?? {}` so strings (and other non-null
values) pass through correctly.
小橘 🍊（NEKO Team）
Replace tsup (esbuild-based) with rslib (rspack-based) across all packages.
tsup's built-in nodeProtocolPlugin strips the 'node:' prefix from all
Node.js builtins. Unlike node:fs etc., node:sqlite has no unprefixed
form, causing ERR_MODULE_NOT_FOUND at runtime. rslib handles node:
imports correctly without any workarounds.
Changes:
- Replace tsup.config.ts with rslib.config.ts in core, daemon, cli
- Swap tsup → @rslib/core in devDependencies
- Fix log-store.ts params type (Record<string, unknown> → Record<string, string | number>)
- Fix logStream.fd type cast in start.ts
- Exclude __tests__ from CLI tsconfig to avoid DTS errors
- All 356 tests pass, nerve init works correctly
Closes#70
小橘 🍊（NEKO Team）
tsup's built-in node-protocol-plugin strips the 'node:' prefix from
all builtins. Unlike node:fs etc., node:sqlite has no unprefixed form,
causing ERR_MODULE_NOT_FOUND at runtime.
- Add onSuccess hook to both cli and daemon tsup configs to restore
'node:sqlite' imports in bundled output
- Fix log-store params type to Record<string, string | number>
小橘 🍊（NEKO Team）
- Upgrade drizzle-orm from 0.43.1 to 1.0.0-beta.23
- Replace better-sqlite3 with node:sqlite (DatabaseSync) in:
- sense-runtime.ts (Drizzle driver)
- log-store.ts (raw SQL)
- all test files
- Replace sqlite.pragma() with sqlite.exec('PRAGMA ...')
- Replace sqlite.transaction() with manual BEGIN/COMMIT/ROLLBACK
- Update CLI init command to verify node:sqlite instead of better-sqlite3
- Remove better-sqlite3 and @types/better-sqlite3 from dependencies
- Zero native addons remaining in the monorepo 🎉Closes#67
小橘 <xiaoju@shazhou.work>
- Add engines >= 22.5.0 to root and cli package.json (node:sqlite requirement)
- Remove unused @types/better-sqlite3 from cli devDeps (leftover from sql.js migration)
- Add files/publishConfig to core package.json (parity with other packages)
- Fix daemon test type errors: add getAllWorkflowRuns to mock LogStore,
fix array destructuring on mock.calls, fix sense-runtime callback signatures
All 356 tests pass across all packages.
小橘 🍊（NEKO Team）
- Add .cursor/rules/no-dynamic-import.mdc: ban dynamic import() in
production code with documented exceptions
- Add .cursor/rules/gitea-access.mdc: tea CLI usage guide
- Add explanatory comments on the 2 legitimate dynamic imports in
sense-runtime.ts and workflow-worker.ts
Ignore SIGINT/SIGTERM only when fork IPC is active (process.send) so terminal signals do not race the kernel shutdown in nerve dev, without breaking standalone worker CLIs (fixes#55).
Pipe worker stderr through the parent with a rolling capture buffer; log exit signal name and stderr tail on worker exit (fixes#56). Apply the same exit logging to workflow workers.
Made-with: Cursor
- createBlobStore(root) with write/read/exists API
- sha256 hex, first 2 chars as shard directory
- Atomic writes via temp file + rename
- CAS mismatch detection on read and write
- Inject blobStore into sense compute via options.blobs
- Export createBlobStore, normalizeBlobHash, BlobStore type
- Add meta table with archived_up_to watermark in logs.db
- Archive logs older than 30 days to data/archive/logs/YYYY-MM-DD.jsonl
- Idempotent: same-day re-export overwrites file
- Single transaction: DELETE + UPDATE meta
- Optional VACUUM after archive loop
- CLI: nerve store archive [--vacuum]
- 15+ new tests for archive logic
- daemon-ipc: add list-senses request type returning SenseInfo[]
- kernel: implement listSenses querying logStore for last signal time
- CLI: nerve sense list with table output, fallback to nerve.yaml when daemon is down
- 25 new tests across daemon-ipc and CLI
- Remove private:true from cli and daemon package.json
- Add files and publishConfig fields
- Add shebang banner via tsup for CLI entry
- Add trigger-sense IPC support in daemon and client
Closes#40
小橘 <xiaoju@shazhou.work>
- daemon-ipc: wrap startWorkflow() in try/catch so errors are sent back
as {ok:false, error:msg} instead of silently dropping the socket
- init-workflow.test: import buildWorkflowTemplate from init.ts instead
of maintaining an inline copy
Addresses review follow-up suggestions from PR #31.
小橘 <xiaoju@shazhou.work>
Critical:
1. trigger: use Unix socket IPC to daemon instead of direct DB write
- new daemon-ipc.ts (server) + daemon-client.ts (client)
- kernel accepts ipcSocketPath, auto-starts IPC server
2. init workflow: validate name (lowercase alphanumeric + hyphens only)
Should fix:
3. getAllWorkflowRuns: SQL query on workflow_runs table instead of O(n) scan
4. limit/offset: robust parseIntArg() helper with NaN handling
5. statusIcon: exhaustive switch with never type check
6. trigger: end-to-end Unix socket tests added
12 new tests. All 224 tests pass.
小橘 🍊（NEKO Team）
Critical:
1. replayAndResume: remove double moderate() call, reuse loop result
2. drainAndRespawn: check workflow still in config before respawn
3. drain: mark in-flight runs as 'interrupted' in DB before clearing
Should fix:
4. crash recovery: dedup runId before re-queuing/re-activating
5. drain timeout: DEFAULT_DRAIN_TIMEOUT_MS > WORKER_SHUTDOWN_TIMEOUT_MS
6. crash-loop protection: max 5 crashes in 60s window, then stop respawn
5 new tests added. All 173 tests pass.
小橘 🍊（NEKO Team）
- Add mkdirSync for data/senses/ in init command (#23)
- Add defensive mkdirSync in sense-runtime before DB open (#23)
- Change init template output from index.ts to index.js (#24)
- Remove TypeScript type annotations from CPU usage template (#25)
Closes#23, closes#24, closes#25
- kernel.ts: create WorkflowManager, expose on Kernel, wire into
ReflexScheduler via workflowTriggerFn, add activeWorkflows to
KernelHealth, graceful shutdown ordering
- reflex-scheduler.ts: handle kind:'workflow' reflexes — subscribe to
bus signals and delegate to workflowTriggerFn
- workflow-manager.ts: add totalActiveCount(), updateConfig() for hot
reload support
All 140 tests pass.
小橘 🍊（NEKO Team）
Critical fixes:
1. triggerPayload spread → safe field assignment with validation
2. Graceful shutdown stopping flag → no false crash warnings
3. Shutdown awaits in-flight work (10s timeout) before exit
4. Thread loop safety valve (MAX_STEPS=1000)
5. active counter: bare number → Set<string> by runId
Should-fix:
6. ThreadEventMessage.eventType → union type with validation
7. SQLite status cast → runtime validateWorkflowRunStatus()
8. getActiveWorkflowRuns → pre-compiled prepared statements
小橘 <xiaoju@shazhou.work>
- reloadConfig: restart existing groups when new senses added
- sense-worker: per-sense timeout/gracePeriod lookup at compute time (RFC §5.3)
- restartGroup: await worker ready promise before returning
- Comments: scheduler in-flight state loss, fs.watch Linux caveats, grace period blast radius
小橘 🍊（NEKO Team）
- file-watcher.ts: watch nerveRoot for .ts and nerve.yaml changes
- kernel.ts: restartGroup(), reloadConfig(), getHealth(), auto-respawn on crash
- sense-worker.ts: compute try/catch error isolation, grace_period hard kill
- ipc.ts: new message types for health, restart, reload
- examples/senses/nerve-health.ts: built-in daemon health sense
- Integration tests for hot reload, error isolation, grace period
小橘 🍊（NEKO Team）
xingyue
xiaoju
xiaomo