HTTP status
400 invalid_request_error
Typical cause
The model supports function tools only through/v1/responses. Sending tools via /v1/chat/completions is not supported for this model (e.g., gpt-6-astra).
Resolution
CallPOST /v1/responses with a Response API body:
tools_endpoints from each provider entry in the model directory. If the array is ["responses"], only /v1/responses accepts function tools for that model.