> ## Documentation Index
> Fetch the complete documentation index at: https://docs.overshoot.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Core flow

> The end-to-end lifecycle: create a stream, publish frames, query frames with chat completions.

```text theme={null}
client -> POST /streams -> Overshoot
client <- {id, publish:{url,token}} <- Overshoot

client -> publish media using publish.url + publish.token

client -> POST /chat/completions -> Overshoot
         messages include image_url/video_url references to ovs://streams/<id>?...
```

## Lifecycle

1. [`POST /streams`](/api-reference/create-stream) creates a stream and returns a publish token.
2. The client publishes video using `publish.url` and `publish.token` (see [LiveKit client flow](/api-reference/livekit-client-flow)).
3. Overshoot makes recent frames available for inference.
4. The client calls [`POST /chat/completions`](/api-reference/chat-completions) with text and optional `ovs://` media references.
5. The client calls [`POST /streams/{id}/keepalive`](/api-reference/keepalive-stream) before the 5-minute lease expires.
6. [`DELETE /streams/{id}`](/api-reference/delete-stream) ends the stream.

There is no public `POST /streams/{id}/infer` contract.