Diagnostics and troubleshooting
A practical path for investigating missing models, incomplete runs, low confidence, webhook failures and unavailable Enterprise controls.
How to interpret this document
This content describes technical and methodological behavior that is implemented or explicitly planned in the product. When a control depends on configuration, a provider, a secret, a contract or legal approval, that dependency must remain visible.
Models do not appear
Check ModelConfig enabled state, methodology support, provider availability and plan limits.
Run does not complete
Inspect MeasurementJobs, queue state, heartbeats, provider errors and entitlement limits. Do not manually mark a run complete just to unlock analysis.
Confidence is Insufficient
Check unique prompts, model count, effective sample size and execution completeness. Improve design and execution rather than hiding the warning.
Source Intelligence is empty
Confirm the methodology produces web evidence, structured sources were captured and valid Grounded responses exist.
Integration does not connect
Distinguish connector availability from OAuth authorization. Check tenant safety, the integration_hub entitlement, provider authorization, granted scopes and connector-specific configuration such as site_url or property_id.
Integration Hub blocked by tenant safety
In app_shared_single_tenant mode, more than one Organization blocks app-scoped connector use. Move to tenant_dedicated_external or equivalent isolation before synchronizing data.
Go-live blocked
Open the GLR-1.0 gate and resolve critical checks before expanding rollout. Warnings should be understood and documented rather than ignored.