One base URL, standard OpenAI request shapes, streaming with usage counts.
https://inference.chessurge.org/v1
Authorization: Bearer sk-...
curl https://inference.chessurge.org/v1/models
curl https://inference.chessurge.org/v1/chat/completions \
-H "Authorization: Bearer $INFERENCE_KEY" \
-H "Content-Type: application/json" \
-d '{"model":"chessurge/deepseek-v4.1-flash","messages":[{"role":"user","content":"hello"}],"stream":true}'
Server-sent events, terminated by data: [DONE]. Usage is returned once in the final chunk.
Standard HTTP status codes with an OpenAI-shaped error object: 400 invalid request, 401 bad key, 402 budget exhausted, 404 unknown model, 429 rate limited, 5xx upstream.
Provider metadata is public at /v1/models in schema 2.4, and at /v1/models?format=legacy for integrations built against the legacy format.