Chronoverse supports periodic automatic scheduling and authenticated manual scheduling.
Automatic scheduling
The scheduling worker polls PostgreSQL for due workflows in bounded batches. It checks that each workflow is active, has a valid build, and still matches the generation being scheduled.
The resulting job state and Kafka publication intent use replay-safe persistence. Reprocessing a scheduling decision does not create unbounded duplicate jobs.
Manual scheduling
POST /workflows/{workflow_id}/jobs/schedule creates an immediate job with trigger MANUAL. The request requires session authentication, CSRF validation, and an Idempotency-Key header.
Timing model
Intervals are expressed in minutes. SCHEDULING_WORKER_POLL_INTERVAL determines scan frequency, while SCHEDULING_WORKER_BATCH_SIZE bounds work per pass.
Lower polling intervals improve responsiveness but increase database pressure. Horizontal scheduler scaling must still respect the repository's locking and replay-safety rules.
Stale work
Workflow generation checks prevent a delayed scheduling event from dispatching work for an older workflow definition.