Skip to main content
Three changes: base URL, API key, model ID. Messages, streaming, tool calls, and the response shape stay the same.
The base URL must end in /v1; the SDK appends /chat/completions. Use catalog IDs: openai/gpt-5.5, not gpt-5.5. See model mapping. resp.choices[0].message.content is the text and resp.usage the tokens. Streaming works as in the OpenAI SDK: stream=True or stream: true.

Privacy

On models with an attested build, the :confidential variant runs inside sealed hardware and returns attestation evidence:
Find the :confidential model ID in GET /v1/models; request fresh evidence from Get attestation. See Privacy Levels.

Notes

  • Nito also serves a stateless POST /v1/responses shim. Chat Completions is the primary surface.
  • A 429 carries Retry-After and X-RateLimit-* headers. See Rate Limits.

Make Your First Request

Field by field.

Chat Completions

The endpoint.