Runs and results
A flow definition describes what to do; a run executes it once. Each run has its own ID, start and end times, status, events, and outputs.
Run lifecycle
A submitted task may start as pending, then become running. Terminal states include succeeded, failed, cancelled, timedOut, and interrupted.
A 202 response from POST .../runs means the task was accepted. Query GET /api/runs/{runId} for its final result.
Events, outputs, and artifacts
| Question | Endpoint |
|---|---|
| Which step is executing? | /api/runs/{runId}/events |
| What values did nodes expose? | /api/runs/{runId}/outputs |
| Were files generated? | /api/runs/{runId}/workpieces |
Events are ordered by sequence. Pass afterSequence to fetch only newer events. Use /events/stream for a continuous feed and Last-Event-ID to resume after disconnecting.
When to debug
If a result is unexpected, start a debug session, set a breakpoint, and step through execution. It owns a separate debug run. See Debug sessions API and Runs API.