Skip to main content

Guide 4: Run a Pipeline

This guide covers triggering a pipeline run against an issue and understanding what happens during execution.

Prerequisites​

  • Daemon is running (runs stay at "Pending" if it's not — see Daemon)
  • You have a default pipeline set on your project (Guide 2)
  • You have an issue in a non-terminal state (Guide 3)

Step 1: Trigger a run​

  1. Open the issue detail page
  2. In the Pipeline Stages card, find the stage that matches the issue's current state
  3. Click Run Stage

fluxaOS creates a pipeline_run and stage_run in pending status and returns immediately. The daemon picks up the run asynchronously.

Step 2: Watch execution​

The issue status changes to Queued, then Running as the daemon picks it up.

To see live output, click View Details on the run card (or open the Run Detail Modal). You'll see:

  • Status badge — updates in real time: Pending → Running → Completed / Failed
  • Stage timeline — each stage as a card, color-coded by status
  • Output tab — live transcript of everything the AI printed, including tool calls, tool results, and final cost
  • Gates tab — pass/fail for each gate rule (if configured)

The UI updates automatically via Supabase Realtime. If you don't see updates, check your network connection — see the Realtime dependency note in Signals.

Step 3: Handle a manual gate (hold)​

If a stage has gateMode: hold, execution pauses after the stage completes. You'll see three buttons:

  • Approve — proceed to the next stage
  • Rework — mark the stage as failed and route to rework
  • Abort — cancel the entire run

Review the stage output in the Output tab before deciding.

Step 4: Check the issue state after the run​

Once the run completes, the issue state advances automatically. If the research stage passed, the issue moves from Research to Implement state. The status returns to Open (ready for the next trigger) or Blocked (needs your attention).

To run the next stage, repeat from Step 1.

Next: Guide 5: Read the Results →