FAQ

Why do requests return 401?

This usually means the API key does not exist, has expired, or the request is missing the correct Bearer token header.

Why does the same code report “model not found” after switching gateways?

The model name used by the client is not enabled in OpenFrog, or the exposed model name does not match the client configuration.

Why do some requests work while others time out?

Check these items first:

• Upstream route health
• Gateway timeout settings
• Whether the request hit rate limits or concurrency controls

How do I roll this out more safely?

Start with curl and one model, validate the smallest successful path, and then move real workflows over incrementally.