Gateway operations and troubleshooting
Use the status code to narrow down a failed call, then look up its request ID in enterprise logs. Find readiness checks, documentation changes and how to contact your contracted support channel.
Troubleshoot by symptom
401 / 403- Check the environment, key expiry, credential header, source IP and model access. Try the model list with the same key.
400 / 404- Check for a duplicated /v1, then verify the endpoint, model ID, request body and capability.
402- Review wallet balance, remaining allowance and balance admission rules in the enterprise console.
429- Honor Retry-After when present. Bound retry attempts and use exponential backoff with jitter.
502 / 503 / 504- Save the request ID and check readiness and the affected model. For partially streamed or submitted async requests, check the original result before retrying.
SSE- Compare direct gateway access with access through your proxy. Check SSE delivery, proxy buffering and read timeouts.
Check instance version and readiness
GET /api/public/status- Read the instance version and site information. The status field is a fixed marker in the public snapshot.
GET /healthz- Operators can check whether the gateway process is alive in the deployment environment.
GET /readyz- Operators can check whether the instance is ready to accept traffic. Verify model and channel availability with the relevant API call.
Contact enterprise support
For contracted enterprise support, contact your enterprise administrator or use the support channel specified in your service agreement. Include the environment, instance version, time and time zone, endpoint, model ID, HTTP status, error code and x-request-id if present. For streaming issues, indicate whether the first SSE event arrived.
Include a minimal redacted request if useful. Remove API keys, sign-in JWTs, customer content and personal information. For billing issues, attach the affected time range and relevant usage records.
Documentation updates
- Added integration guides for coding tools, agents and clients, plus guidance on enterprise migration, model protocols and troubleshooting. Explore tools and workflows
Before switching models, confirm the target ID with your current key, review its protocol and effective rate, and test the change in your agreed rollout window.
