Getting the Fastest Resolution

Describe the actual problem, not just the symptom

• When did it start? What changed around that time - a new model, a config update, a scaling event?
• What were you expecting to happen? What happened instead?
• Is it consistent or intermittent? Does it affect all models or a specific one?

For Inference API issues

• Include the model ID - use the correct model names e.g. mistralai/Mistral-Large-3-675B-Instruct-2512-NVFP4. Behaviour varies by model and we need to know exactly which one was in use.
• Include the endpoint path - e.g. /v1/chat/completions, /v1/embeddings, /v1/audio/transcriptions. A 429 from /v1/rerank is a different problem from a 429 on /v1/chat/completions.
• Paste the full error response body - the HTTP status code alone (401, 403, 429, 500) tells us very little without the message field.
• Include the X-Request-Id - every API response carries this header, and it lets us trace the exact request in our logs. See Request IDs.

For Private NeoCloud issues

• Kubernetes: Specify the cluster name, namespace, pod/deployment name, and attach relevant output from kubectl logs.
• Storage issues: Include the PVC name and the error observed.
• S3 storage: Provide the bucket name and the error or behavior observed.
• VPN access: Indicate whether the issue occurs with or without FortiGate VPN connected.
• Managed services: For PostgreSQL, Valkey, Qdrant, or Ferret issues: include the instance name and the error message.
• OpenWebUI: Include logs from the failing component (PostgreSQL, Qdrant, or the OpenWebUI deployment).
• Networking: For ingress or service issues: provide the service name, namespace, and any relevant error responses.

Attach supporting files

• Screenshots, screen recordings, kubectl output, OpenStack logs, or exported cluster configs
• File limit: 10 MB per file via portal, 25 MB via email attachment

---