Inspect Failed Steps
Failed rows are the starting point for triage. A failure may mean the application has a bug, the test needs an update, the target environment is not ready, or a shared dependency such as login or test data failed.
Open the failed run, filter the Tests tab to its failed rows, and read the failure panel from top to bottom: the error, the location, the final screenshot, and the video. Match what you see to a likely cause, go deeper with the trace and Debug with Assistant, and diagnose the same run from the command line.
Open failure details
- Open Runs.
- Select the failed run.
- Open the Tests tab.
- Find the failed scenario rows when the run has many.
- Select the failed row to open its details.
The details view can show the scenario location, the error, the final screenshot, the video, and a Debug with Assistant button that opens the Playwright trace workspace with the Assistant.
Read the failure panel
Use the panel from top to bottom:
| Area | What to look for |
|---|---|
| Error | The Playwright error text, assertion output, and the failure source. |
| Location | The spec file and configuration that identify where the failure occurred. |
| Final screenshot | The browser state at the end of the failed scenario. |
| Video | The browser session leading up to the failure. |
| Debug with Assistant | Opens the full action-by-action Playwright trace with an Assistant docked beside it that diagnoses the failure automatically. |
Decide what kind of failure it is
| Signal | Likely cause |
|---|---|
| Screenshot shows the wrong page | Navigation, login, or environment setup may be wrong. |
| Screenshot shows the expected page but the assertion failed | The test expectation may be too strict or out of date. |
| Error mentions missing selector or timeout | The UI changed, the app was slow, or the selector needs an update. |
| Error appears before the test reaches the app | Check session setup, dependencies, and secrets. |
| Many rows fail at the same prerequisite | Fix the prerequisite before editing individual tests. |
Use the next evidence layer
If the failure panel does not explain the problem:
| Need | Open |
|---|---|
| See the exact browser sequence | Watch videos |
| Inspect page actions, DOM snapshots, and request timing | Open traces |
| Check browser console errors or network requests | Open traces |
Debug with Assistant investigates the evidence and prepares a concrete fix for review when a code change is supported. Review and apply the proposed change yourself. If the problem is environmental or the evidence is incomplete, it should explain the next useful action instead of inventing a patch.
After you make a change, re-run the smallest useful scope first. For re-runs, see Cancel or re-run tests.
From the command line
With the CLI configured, list failed runs and inspect one:
testvibe runs list --status failed
testvibe runs diagnose <runId>
The diagnosis reports available failing-scenario errors, stack details, log output, and artifact paths. Use --json for the structured response; signed artifact URLs are omitted in both formats unless you add --links. Signed URLs expire, so treat them as private links and request fresh ones when needed.