3tears-scheduled-jobs 0.18.0

Generic, payload-agnostic, multipod-safe scheduled-jobs core -- cross-pod-locked tick engine, reschedule math, store protocols, and a default kind+payload store

3tears-scheduled-jobs

A generic, payload-agnostic, multipod-safe scheduled-jobs core. Every agent/skill/webhook/conversation-specific concept is stripped out, leaving only the scheduling machinery.

What it gives you:

• scheduled_tick_job(...) -- one cross-pod-locked tick pump. Acquire the nats_distributed_lock at a caller-supplied key; on LockHeld skip silently; on KvError degrade open (the per-row optimistic-CAS is the real guard); enumerate due rows; per-row CAS-claim + reschedule; invoke an injected dispatch callback; drift / missed-fire accounting; per-row failure isolation. Takes the store(s), the dispatch callback, and the NATS client as parameters, with no domain knowledge.
• compute_next_fire_at(...) -- the pure reschedule math for every schedule type (daily_at, every_n_hours, random_within_window, one_shot_at, cron, relative_delay, interval) and both missed-fire policies (coalesce, catch_up). The cron branch imports APScheduler lazily, so non-cron consumers pay nothing.
• ScheduleStore / FireStore Protocols -- the exact surface the tick engine calls. The engine depends only on these, so a typed consumer collection can implement them.
• A default store -- ScheduledJobEntity / JobFireEntity + collections + scheduled_jobs / job_fires table factories + a v001 migration, keyed on an opaque kind (TEXT) + payload (JSONB). A simple consumer can use it as-is with no table of its own.
• Generic config / events / metrics -- a JobConfig protocol, the tick / fire / drift event-name constants, and cardinality-bounded Prometheus instruments.

The engine is pure-async, one tick per call with no internal polling. Drive cadence with whatever scheduler you like (an APScheduler IntervalTrigger, a while True: await asyncio.sleep(...), and so on). The engine does not own the scheduler.