API documentation

The Northwind AI API is organized around REST. All requests are made over HTTPS to /v1 and authenticated with a Bearer API key. JSON is returned for every response.

Authentication

Pass your key in the Authorization header: Authorization: Bearer NORTHWIND_API_KEY. Requests missing a valid key receive an HTTP 401 with a JSON error body. Keys are managed from the dashboard and can be rotated at any time.

Endpoints

• POST /v1/chat/completions — chat completions, optionally streamed.
• POST /v1/embeddings — vector embeddings.
• GET /v1/models — list available models.
• WSS /realtime/<session>/stream — bidirectional realtime token streaming.

Realtime streaming

Realtime sessions open a WebSocket to the streaming endpoint and exchange framed events for the lifetime of the session. The status board on the home page runs a read-only demo session so figures update without a page reload.

Rate limits

Limits are applied per key and per model tier. Exceeding a limit returns HTTP 429; back off and retry using the Retry-After header.