OpenAI Responses API
The Responses mirror is POST /proxy/openai/v1/responses. Reuse the service/provider credentials and OpenAI client from the complete quickstart. This is a fragment, to replace its Chat Completions call; Python requires the same pinned openai==3.41.0 dependency.
raw = client.responses.with_raw_response.create(
model="gpt-4.1-mini", input="Say hello", max_output_tokens=64)
print(raw.parse().output_text)
print(raw.headers.get("Makas-Was-Routed"))
Responses uses input and max_output_tokens, not the Chat Completions request schema. Provider feature/model access still applies. Buffered metadata is conditional; streamed outcomes must be checked after termination in Request Logs.