Requests & responses
Request shape, content types, and the response format.
Every endpoint speaks JSON over HTTP. Shape, encoding, and response format are the same everywhere.
Base URL
Every path hangs off api_prefix — /api/v1 by default. In the examples the backend runs locally at http://localhost:8000.
bash
http://localhost:8000/api/v1Chat body limits
messagesarray · 1–100role is system | user | assistant, content 1–32000 characters.
temperaturenumber · 0–2null = default from configuration.
max_tokensinteger · 1–32000Counts thinking tokens on reasoning models.
top_pnumber · >0–1null = default.
Stateless
The backend remembers nothing between requests — send the full history with every chat call.