Tutorial 4 — Create and Run a Task

A task combines everything: operator + connection + payload (+ config + actions).

Create it

  1. Navigate to your workflow folder.
  2. Click Create Task.
  3. Fill in:
    • Name — e.g. 01_insert_rows.sql (the demo names its steps after the SQL file)
    • Operator — e.g. PostgresqlExecuteSQL
    • Connection — e.g. postgresql
    • Frequency — ADHOC for one-off, or a cron expression (e.g. 0 2 * * *)
    • Payload — your SQL/JSON from the previous step
    • Config — optional: attach a config for retries, parameters, default actions
    • Start / End Dates — required for scheduled tasks; leave empty for ADHOC
  4. Click Create Task.

Or with the AI

The same task can be created and run entirely through the AI (Service AI chat or any MCP client):

"create a task 01_insert_rows.sql using operator PostgresqlExecuteSQL, connection postgresql, and this payload [...], schedule 0 2 * * *; then run it and check the status."

The agent calls create_catalog_item (item_type: task), then run_task and get_task_status. You can also deploy a whole usecase in one step — deploy usecase <name> — which creates all its tasks at once. See Usecases and MCP.

Run it

Adhoc — set frequency: ADHOC; the task runs immediately when triggered. No schedule, no generated instances.

Scheduled — set a cron expression + start_date/end_date. LeastAction generates one task instance per interval, each with its own logical_date:

frequency:  0 2 * * *
start_date: 2026-03-01T00:00:00Z
end_date:   2026-03-31T23:59:59Z

This creates 31 daily instances, each with the {{ds}} for that day.

View logs & status

Open the Logs section, or the task's Parent-Child / history views, to see timestamps, per-step messages, state, and action results in real time. Or ask the AI: "run task 01_insert_rows.sql and check its status."

See Monitoring & Logs and Task States.

Next

→ Tutorial 5 — Schedule & backfill