codai Chat & Completions API
OpenAI-compatible inference surface. `POST /v1/chat/completions` is the hot path every SDK, IDE plugin and CLI speaks; `POST /v1/responses` is a thin translation layer for clients that speak the OpenAI Responses wire format and re-enters the chat-completions route in-process, so auth, rate limits, routing, caching and usage recording are identical. The only publicly addressable model is `codai` (the smart-router alias); unknown model names are coerced to `codai` instead of being rejected.