The workflow runs quality checks on each recording: pre-checks on the raw files, then checks on the standardized recording, such as timestamp alignment, action smoothness, stream gaps, and feature-schema drift. Open Quality in the project sidebar for recent quality jobs and failed versions.
Status values
| Status | Meaning |
|---|---|
| pass | The check ran and passed. |
| warn | The check ran and found something to review. |
| fail | The check ran and failed. |
| error | The check could not complete. |
| not_evaluated | The check did not run. |
| not_applicable | The check does not apply to this data. |
Only pass is a pass. error, not_evaluated, and not_applicable are never shown as green, and a missing extractor or input is reported with a named code instead of a silent fallback.
Where you see results
- The Recordings list shows quality and a suggested next step for each recording.
- A recording that a blocking check stops shows the check, the cause, and the next step.
- In the Episodes page you can filter by failing check.
- On a dataset version, the verdict summary counts Pass, Warning, Fail, Incomplete, and Not checked episodes.
Failed or flagged items become repair tasks you can assign and resolve.