Scheduled tasks

Background worker schedules, their run history, and running one now.

Administrators only. Every periodic job the platform runs for itself: dequeuing, diagnosis, the watchdog, storage scans, registry reconciliation, the daily report.

Background task schedules and their run history.

What you see

Column	What it shows
Schedule	The interval, or that the task is off
Enabled	Whether it runs at all
Last run	When, how long it took, and whether it succeeded
Trigger	Whether that run was scheduled or started by hand

What you can do

• Change an interval for tasks with an adjustable cadence. Some are fixed, and say so.
• Turn a task off.
• Run now — useful after changing a setting the task reads, rather than waiting out its interval.
• Read the history — recent runs with outcome and duration.

Two banners worth understanding

Every background task is stopped

No Redis is configured. Background roles need a shared lock so that several console replicas do not all run the same task at once, and without one the platform stops them rather than running them unsafely.

On a single instance this is a warning you can accept. On more than one replica it means storage accounting, diagnosis and the daily report are simply not happening.

Workers run in a separate process

FORGE_INPROCESS_WORKERS is off, so these tasks are executed by python -m server.worker rather than inside the web process. That is the recommended shape at scale — the slow LLM-calling ticks stop inflating API tail latency — and this page still shows their schedules and history.

If a task is not running, check that the worker process for its role is actually up.