The "INTERNAL_ERROR: AI service did not respond with a valid answer" in Databricks AI/BI Genie typically means the Genie service failed to process your query, often due to one of a few common issues. This can include problems with the table existence, schema mismatches, malformed function arguments, or AI service configuration errors. Sometimes, this error arises if Genie can't access or understand your data sources, or if there are permission or connectivity issues within the Databricks workspace.โ
Troubleshooting Steps
-
Confirm that the data table referenced in your prompt actually exists and is accessible by Genie. Try running a simple SQL query in a Databricks notebook (e.g., SELECT COUNT(*) FROM Opportunity) to test connectivity.โ
-
Simplify your initial prompt to something less complex, such as asking Genie to list available tables or describe the schema, which can reveal if the tool can see your data at all.โ
-
Check Genie configuration to ensure it has access to the specific database catalog or schema required. Pay particular attention to Unity Catalog permissions; Genie must have SELECT privileges on the target tables.โ
-
Validate your cluster setup and permissions, including the bearer token used for authentication. The error may also be caused by expired or insufficient API permissions.โ
-
If issues persist, try restarting your Databricks cluster or connecting to a different one, as temporary service disruptions or backend outages can cause such errors.โ
Additional Considerations
-
Examine audit logs and Genie logs for more detailed error messages, especially if data source or API connectivity is questionable.โ
-
If you recently changed data schemas, re-ingest or recalculate Genie embeddings to keep the metadata in sync.โ
-
Some errors might be temporaryโDatabricks service downtime has occasionally impacted Genie responses, so retrying after some time may also work.โ
If the problem continues after these steps, reach out to Databricks support or check recent Genie release notes for bug fixes related to API errors or dashboard access
