Runs, approvals, and troubleshooting
Inspect a paused or failed workflow and decide whether it is safe to continue.
Open the affected run
In Workflows → Runs & Approvals, open Run Center, select the run for the project and time, and inspect its current step or first failed node. The run details show the saved definition used for this execution and the available step events, errors, approval records, and artifacts.
| What you see | What to check |
|---|---|
| Pending or running | Look for a queued or active step and check its latest events. If a provider returned a task ID, keep it with the run record. |
| Waiting for approval | Open the pending request in run details or Approval Center. Verify the step, contents, destination, account, and run context before deciding. |
| Completed | Open the expected artifact or project delivery. For an external action, compare the receipt and the target system’s state. |
| Failed or partial success | Find the first failing step and inspect earlier outputs. Some earlier steps may already have created files, consumed provider credits, or written externally. |
| Cancelled | Check the last completed step and verify side effects already dispatched to a provider or destination. |
Approve the specific action
Put a gate before the first high-impact action, including uploads or autosaving edits when they write externally. Approving its request lets the run resume under its saved definition. A later step may have a separate approval. A template input such as reviewApproved selects a branch when the workflow uses it. Use the pending approval record to record the human decision.
If a request is missing the destination or cannot be tied to the intended account and content, reject it. Correct the flow or run input, save and activate any definition changes, and test a new run.
Choose between retry and a new run
- For an input or expression error, check the run input and the step’s actual upstream output. Changes to either call for a new run with the corrected input or saved active definition.
- For a connection or permission error, inspect the route, connection instance, resource, and grant. Repair the binding before retrying.
- When an upload, payment-related generation, send, or submit may have happened, reconcile the provider or target system first. Check for the original task, record, or receipt before starting anything that could repeat it.
- Retry Node, where allowed, creates a linked run using the original definition and input, with the failed node and affected downstream scope reset. The runtime restricts this to explicitly retry-safe reads, supported media recovery with the original receipt, and terminal project delivery, subject to its checks.
See common workflow errors for more diagnosis paths and node setup for mapping and readiness checks.