The OpenAI Responses API: POST /v1/responses
Point the OpenAI SDK's responses.create, or Codex, at Inference Bureau. Requests pass through natively with the same Directives, holds, failover, idempotency and receipts as chat completions, including streaming, function tools and structured output.
It is stateless: nothing is stored, so each request carries the whole conversation in input, as with store: false. previous_response_id, store: true and conversation are refused with a clear invalid_request.
OpenAI's hosted tools (web search, file search, code interpreter, computer use) are refused, because their cost can't be reserved before the request runs. Function and custom tools, text.format structured output and reasoning settings work as in the OpenAI API.
More in Create a response.