Troubleshooting Strive
Find the failed layer and recover without repeating completed work.
Start with the exact object and action that failed. An app, its deployment, a workflow run, a connection, a file transfer, a document extraction, and a notification each have different saved states.
Identify the layer
| Symptom | First place to inspect |
|---|---|
| Cannot sign in or a link is denied | Access and sessions |
| App does not open or start | App troubleshooting |
| Processing failed or outputs are wrong | Workflow troubleshooting |
| Service access expired | The connection's status, test, and reconnect flow |
| Transfer stopped with some files copied | Its storage operation and per-item receipts |
| Extraction or a citation is missing | The document, extraction result, and original source |
| Save reports a newer revision | Concurrent edits |
| Request ended without a clear result | Uncertain outcomes |
| Data appears empty | Missing results |
Preserve useful context
Record the item name or identifier, time, action, visible error, and operation or run identifier. Include whether another tab or person was editing the same item. Avoid sending secrets, public share-link tokens, or unnecessary private document contents.
Retry after understanding
Fix the cause before rerunning. A bad credential needs access repaired; a revision conflict needs a fresh read; a budget limit needs a policy or workload decision. A repeat button cannot resolve every class of failure.
The Assistant can inspect failed runs and propose a correction. Review changes and verify their result before resuming a schedule or sharing the app again.