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¶

Type	Flag	Example	Description
`cron`	`--schedule`	`"0 9 * * *"`	Standard cron expression (minute, hour, day, month, weekday)
`every`	`--every`	`1h`	Interval-based repetition using Go duration syntax
`at`	`--at`	`2026-02-20T15:00:00Z`	One-time execution at a specific RFC3339 datetime

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"

List Jobs¶

lango cron list

Pause / Resume¶

lango cron pause --id <job-id>
lango cron resume --id <job-id>

Delete a Job¶

lango cron delete --id <job-id>

View History¶

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

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

Session Modes¶

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

Mode	Session Key Format	Behavior
`isolated` (default)	`cron:<name>:<timestamp>`	Fresh session every execution. No memory of previous runs.
`main`	`cron:<name>`	Shared session across all runs. Agent remembers previous outputs.

# Use shared session (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

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.

No Delivery Channel

If no delivery channel is configured (neither per-job nor default), job results are logged but not delivered to any channel. A warning is emitted in the logs.

Configuration¶

Settings: lango settings → Cron Scheduler

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

Key	Type	Default	Description
`cron.enabled`	`bool`	`false`	Enable the cron scheduling system
`cron.timezone`	`string`	`"UTC"`	Default timezone for cron expressions
`cron.maxConcurrentJobs`	`int`	`5`	Maximum concurrently executing jobs
`cron.defaultSessionMode`	`string`	`"isolated"`	Default session mode for new jobs
`cron.historyRetention`	`string`	-	Duration to retain execution history
`cron.defaultDeliverTo`	`[]string`	`[]`	Default delivery channels

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