Scheduled jobs: cron and proactive runs

Maps to: Scheduled jobs: cron.

Scope

Proactive agent work on a schedule: sleep-time compute, monitoring loops, and batch jobs with the same durability and auth as interactive runs.

Design questions

• Stateful cron on a fixed thread versus stateless fresh thread per execution.
• Thread cleanup after batch runs and discoverability of historical cron output.
• Alerting when scheduled runs fail or silently skip.
• Time zone and daylight-saving behavior for operator-defined schedules.

Tradeoffs

• Stateful schedules accumulate context but grow checkpoints and privacy scope.
• Stateless batches simplify isolation but lose narrative continuity.
• Forgotten crons continue billing and load until explicitly removed.

Evaluation hooks

• Scheduled run survives transient model outage with retry consistent with interactive runs.
• Auth and middleware apply identically to cron-triggered runs.
• Operational runbook for disable and drain.

Reference notes

See LangChain runtime article (cron patterns figure).