Cron Scheduling¶

Persistent cron scheduling system powered by robfig/cron/v3 with Ent ORM storage. Jobs survive application restarts -- on startup, all enabled jobs are loaded from the database and re-registered with the scheduler.

Schedule Types¶

One-time (at) jobs are automatically disabled after execution.

CLI Commands¶

Add a Cron Job¶

# Daily summary at 9 AM
lango cron add --name "daily-summary" \
  --schedule "0 9 * * *" \
  --prompt "Summarize yesterday's activity" \
  --deliver-to telegram

# Every 2 hours
lango cron add --name "health-check" \
  --every 2h \
  --prompt "Check all systems status" \
  --deliver-to slack

# One-time execution
lango cron add --name "deploy-reminder" \
  --at "2026-02-20T15:00:00Z" \
  --prompt "Remind team about the deployment window"

--deliver and --deliver-to are equivalent. Both accept one or more delivery targets such as slack, telegram, or channel-specific identifiers.

Per-Job Timeout¶

Each job can specify a per-job timeout that overrides the global cron.defaultJobTimeout. If no per-job timeout is set, the global default (30 minutes) applies.

# 5-minute timeout for a quick check
lango cron add --name "quick-check" \
  --every 30m \
  --prompt "Check API latency" \
  --timeout 5m

# 2-hour timeout for a long-running report
lango cron add --name "monthly-report" \
  --schedule "0 2 1 * *" \
  --prompt "Generate full monthly report" \
  --timeout 2h

The timeout parameter accepts Go duration syntax (e.g., 5m, 1h30m, 2h). Precedence:

1. Per-job --timeout — if set on the job, this value is used
2. Global cron.defaultJobTimeout — fallback when the job has no explicit timeout (default: 30m)

List Jobs¶

lango cron list

Pause / Resume¶

lango cron pause <id-or-name>
lango cron resume <id-or-name>

# The documented --id form is also accepted.
lango cron pause --id <id-or-name>
lango cron resume --id <id-or-name>

Delete a Job¶

lango cron delete <id-or-name>

# The documented --id form is also accepted.
lango cron delete --id <id-or-name>

View History¶

# History for a specific job
lango cron history <id-or-name> --limit 10

# The documented --id form is also accepted.
lango cron history --id <id-or-name> --limit 10

# History across all jobs
lango cron history --limit 20

Session Modes¶

Each cron job runs in its own agent session by default. The session mode controls whether conversations persist across runs:

# Use shared session for this job (agent remembers previous runs)
lango cron add --name "weekly-report" \
  --schedule "0 9 * * 1" \
  --prompt "Write this week's report, building on previous ones" \
  --isolated=false

lango cron add uses cron.defaultSessionMode when no explicit --isolated flag is supplied. The default configuration is isolated.

Result Delivery¶

Job results are delivered to configured communication channels after execution. If no deliver_to is specified per-job, the system falls back to cron.defaultDeliverTo from the configuration.

Configuration¶

Settings: lango settings → Cron Scheduler

{
  "cron": {
    "enabled": true,
    "timezone": "Asia/Seoul",
    "maxConcurrentJobs": 5,
    "defaultSessionMode": "isolated",
    "defaultJobTimeout": "30m",
    "historyRetention": "30d",
    "defaultDeliverTo": ["telegram"]
  }
}

Architecture¶

The cron system consists of three main components:

• Scheduler (internal/cron/scheduler.go) -- manages job registration, lifecycle, and the concurrency semaphore
• Executor (internal/cron/executor.go) -- runs individual jobs via AgentRunner, persists history, and delivers results
• Store (internal/cron/store.go) -- Ent ORM persistence layer for jobs and execution history