Troubleshooting

A run looks wrong

1. Open the run and check its evidence links — which data and steps produced the output.
2. Confirm the right data source and time range were used.
3. Reject or correct the output. Your correction becomes feedback for the next run.

A data source is stale or disconnected

1. Open the source in your workspace and check its connection status and last sync time.
2. Re-authorize access if credentials expired, then re-run the workflow.
3. Predictions made from stale data stay marked with the evidence they used — re-run after the sync to refresh them.

A proposal does not match what you asked

1. Say what is off, in plain language, on the proposal itself.
2. Review the regenerated proposal before committing.
3. Repeated mismatches are worth reporting — they point at how the request was interpreted, which helps improve future proposals.

Image: symptom-to-first-check flowchart: wrong output → evidence, stale data → sync state, bad proposal → scope.

Still stuck

• Work through Getting help with a run link and the evidence you already gathered.
• Reproduce the area's quickstart to isolate product from setup: Connect data, Run your first pipeline, Read your first run.