Daemon Process (double-fork)

SynlynkDaemon (subclasses WatchDaemon)

Inherited from WatchDaemon

🔵 Poll Thread

Every 3s: check mtime on project-docs/, regenerate context.md on change, run _reconcile_jobs(), dispatch queued + ready jobs

New in SynlynkDaemon

🟣 HTTP Thread

http.server on localhost:27471 — serves context, job queue, stories, capability, sentinel over REST. Runs serve_forever() in a daemon thread.

Pidfile: .synlynk/daemon.pid · Log: .synlynk/daemon.log

Persistence

Job Queue in state.db

New daemon_jobs table — survives restarts. Fields: job_id, agent, task, story_id, status, priority, depends_on, pid, enqueued_at, started_at, completed_at, exit_code

Service Registration

--install-service

macOS: launchd plist → ~/Library/LaunchAgents/com.synlynk.daemon.plist
Linux: systemd user unit → ~/.config/systemd/user/synlynk-daemon.service
Fallback: @reboot crontab entry

HTTP API — localhost:27471

GET /context

Current context.md (plain text or JSON)

GET /status

Daemon health, queue depth, uptime

GET /jobs

All jobs — queued, running, done

GET /jobs/:id

Single job + log tail

POST /dispatch

Enqueue job — agent, task, priority, depends_on

GET /stories

Story list from state.db

GET /stories/:id

Single story detail

GET /capability

Capability ratings by agent × domain

GET /sentinel

Active sentinel alerts

POST /checkpoint

Trigger context regeneration on demand

CLI Surface

synlynk daemon <action>

start · stop · status · restart
--install-service · --uninstall-service

Job Queue Rules

Dispatch logic (poll thread)

Concurrency cap: max_parallel in config (default 4)
Priority: integer 1–10, lower = higher priority
Dependencies: depends_on = JSON array of job_ids; job stays queued until all deps are done