AI Provider Diagnostics & Troubleshooting

This guide assists Tenant Administrators in resolving AI connectivity and configuration issues.

---

1. "AI Provider Unavailable" or "Key Validation Failed"

• Symptom: Assistant or extraction tasks report that the AI provider is unavailable.
• Causes:
  • Invalid API key or revoked token.
  • Organization ID mismatch in Azure or OpenAI BYO settings.
  • Network firewall blocking outgoing API calls.
• Resolution:
  1. Navigate to Administration > AI Settings (/settings).
  2. Re-enter the API key and click Verify & Save.
  3. Check provider status dashboards (e.g., status.openai.com, status.anthropic.com).

2. Rate Limits & Quota Exhaustion

• Symptom: Requests return HTTP 429 (Too Many Requests).
• Resolution: Review your external provider account usage limits. Consider switching to a higher tier or rotating to another configured model provider.

3. Fallback Embeddings (Hash Fallback)

• Symptom: Evidence search or semantic retrieval returns degraded results.
• Cause: The local fastembed model failed to load in the backend container.
• Resolution: Check backend container memory allocations and restart the extraction service.