From chat to programmatic access

Everything the chat UI does is available via API — programmatically. You send an HTTP request with messages and parameters, and get a structured response (usually JSON). This lets you embed the model into your product.

What a request contains

• messages — a list of turns with roles: system, user, assistant.
• model — which model to use.
• parameters — temperature, max_tokens and others.
• API key — in the auth header; keep it secret.

POST /v1/messages
Authorization: Bearer $KEY
{
  "model": "some-model",
  "max_tokens": 500,
  "messages": [{"role":"user","content":"Hi!"}]
}

Key principles

• Stateless. The API doesn't remember past calls — you pass the dialogue history yourself in messages.
• Token-based billing. You pay for input and output volume (next lessons).
• Non-determinism. The same request can yield different answers.

An API is the same AI, but under your control: you decide what goes in and what to do with the response.