Running Workflows

Starting an Execution

There are two ways to run a workflow:

1. From the workflow list — click the Play button on any workflow card
2. From the builder — click the Run button in the top bar (save first)

When a workflow starts, the engine:

1. Validates the DAG structure
2. Creates a workflow execution record
3. Creates step execution records for every step (all start as "pending")
4. Finds root steps (no dependencies) and starts them
5. For each started step, creates a real MC task

Monitoring Progress

Navigate to the execution view (/workflows/:id/executions/:execId) to see:

• Visual DAG — same layout as the builder, but read-only with status colors:

  • ⬜ Gray = Pending
  • ⚡ Blue (pulsing) = Running
  • ⏳ Amber = Waiting for approval
  • ✅ Green = Completed
  • ❌ Red = Failed
  • ⏭️ Gray strikethrough = Skipped

• Progress bar — X/Y steps completed

• Timeline — chronological list of step events with timestamps

• Task links — click to view the actual MC task for any step

The view auto-refreshes every 3 seconds while the execution is running.

How Steps Advance

When any MC task's status changes to "done":

1. The workflow engine checks if the task belongs to a workflow
2. If yes, it marks the step as completed
3. It checks if any downstream steps now have all dependencies completed
4. Those steps are started (creating new MC tasks)
5. When all steps are terminal (completed/failed/skipped), the execution is marked complete

Cancelling

Click Cancel on a running execution. All pending and running steps are marked as "skipped" and the execution status changes to "cancelled".

Retrying

Failed executions can be retried. Click Retry to create a brand new execution with the same input data. The original execution is preserved for history.

Parallel Execution

Steps that don't depend on each other run in parallel. For example, if steps B and C both depend only on step A, they'll both start as soon as A completes.