Jobs¶

What it is¶

Jobs are durable tracked work records. They let humans and agents create work, inspect status, update progress, record results, cancel work, and run recurring worker templates.

When to use it¶

Use jobs when work needs lifecycle state that survives across turns: queued/running/succeeded/failed/canceled state, result summaries, cancellation, retry, dashboard inspection, or recurring path/backend workers.

Use ask for one question that needs a reply. Use Scheduling for future mesh messages that do not need durable work state.

Setup¶

Start the daemon. For executor-backed jobs, configure the backend command and allowed path through spawn configuration.

repowire serve

Worker folders can live under .repowire/agents/<name>:

repowire agents create daily-brief

Common workflows¶

Create a recurring worker job:

repowire jobs create "Daily brief" \
  --path .repowire/agents/daily-brief \
  --backend codex \
  --cron "@daily" \
  --prompt "Prepare the brief."

Inspect jobs:

repowire jobs list
repowire jobs show job-...

Update progress or record a result:

repowire jobs update job-... --state running --note "Started first pass"
repowire jobs update job-... --state succeeded --result-summary "Brief posted"
repowire jobs result job-...

Cancel work that should not continue:

repowire jobs cancel job-...

Unassigned path/backend jobs run with per-fire executors by default: Repowire spawns or backend-resumes an executor for the run, delivers the job ask, and releases the executor after a terminal job update. Recurring jobs can use continuity=resume to keep the backend-native runtime session id as the continuity handle for the next fire. Use --continuity fresh when a run should start without prior runtime context.

Commands and API¶

CLI: repowire jobs create, list, show, update, result, cancel.
MCP: job_create, job_list, job_status, job_show, job_update, job_result, job_cancel.
Dashboard: Jobs view for listing durable work and recurring templates, then running, retrying, or canceling eligible records.

schedule_cron is only for recurring mesh messages to an existing peer. Use job_create(..., cron=...) or repowire jobs create --cron ... for recurring durable executor work.

Limits¶

Jobs complement asks and schedules; they do not replace the ask/ack lifecycle for conversations.
Per-fire executor cleanup is tied to terminal job lifecycle state, not ack lifecycle.
Ack only confirms receipt. Terminal job updates and compatible cancellation paths release the executor.
Resume continuity depends on a backend that supports local resume and a captured runtime session id.

Troubleshooting¶

A recurring job did not create another run: inspect the job template with repowire jobs show and confirm it was created with --cron.
A worker did not resume prior context: confirm the job continuity is resume, the first run captured a runtime session id, and the backend supports resume.
A process stayed alive after work ended: confirm the job reached a terminal state with jobs update or jobs cancel.