Skip to main content
The Overshoot REST API turns a live video feed into a conversation. Publish frames over WebRTC, then call an OpenAI-compatible chat completions endpoint to ask any vision-language model about your stream — over the last few seconds, the latest frame, or anywhere in the stream’s history.

​
Endpoints

  • List models — discover the vision-language models you can target in chat completions.
  • Create stream — open a live stream and get a LiveKit room URL + token to start publishing frames.
  • Get stream — inspect the state of a stream: frame counts, recent FPS, lease expiry.
  • Renew stream lease — keep a stream alive past the default 5-minute idle window.
  • Delete stream — end a stream and release its resources.
  • Chat completions — ask any vision-language model about a stream’s frames.

​
Base URL

Billing endpoints are mounted on the same host under /billing.

​
Authentication

Bearer tokens. Include your API key on every authenticated request:
Keys are prefixed with ovs- and managed in the Overshoot dashboard. List models is the only unauthenticated endpoint.
Calls using your API key incur cost. Treat keys as secrets — never commit them or expose them in client-side code.

​
Errors

Non-2xx responses use this shape:
  • The Stream — concept guide for the stream lifecycle.
  • Models — overview of available vision-language models.
  • Chat Completion — the URL grammar for referencing frames and segments.
  • Best practices — production tips for low latency and cost.