Skip to main content
This page lists changes to the Overshoot API, the hosted models, the documentation and the terms. Customers receive prior notice of scheduled changes to the Services they use.

Terms of Service updated

September 28, 2026The Terms of Service were published with a last updated date of September 28, 2026. Read them at platform.overshoot.ai/terms.html.

Camera sources

August 5, 2026Streams can be created from a remote camera source. Pass a source object with an RTSP URL or an HLS playlist URL when you create the stream. The server pulls the video, so no client publisher is needed. See Camera sources.

Viewer token endpoint

July 21, 2026POST /streams/{stream_id}/viewer-token returns a subscribe-only token. Use it to watch the video that the model is analyzing. The token cannot publish media or data. See Create viewer token.

LiveKit integration guide

July 20, 2026New guide on building an agent that joins a LiveKit room, analyzes frames with Overshoot and publishes structured results to the room. See LiveKit.

API reference rewritten for v1beta

June 16, 2026The API reference was rewritten for the v1beta API. All guides use the base URL https://api.overshoot.ai/v1beta. The version picker label was set to v1beta on June 17. See the API introduction.

Billing endpoints

June 16, 2026The API reference documents four billing endpoints: list pricing, get model pricing, get prepaid balance and create checkout session.

Model list refreshed

June 16, 2026The Models page and the code examples use the model IDs returned by the live /models endpoint. The endpoint remains the source of truth.

Quickstart snippets fill in from page inputs

May 16, 2026The Quickstart has inputs for the API key, the stream ID and the model. The code snippets on the page update as you type.

Multimodal processor rewrite

May 13, 2026Engineering post on the rewrite of the multimodal processor that prepares video input for the Qwen3 and Gemma 4 model families. Model outputs are unchanged. Read the post.

Gemma 4 vision encoder batching

May 12, 2026Engineering post on a change to the Gemma 4 vision encoder. Video frames are processed as a batch instead of one by one. Model outputs are unchanged. Read the post.

Models page reorganized

May 1, 2026The Models page was refreshed to match the model registry. It is split into Overshoot-hosted models and proprietary passthrough models.

Stream URL scheme

April 30, 2026Stream references in chat completion requests use the ovs://streams/<id> URL scheme, for single frames and for video segments. See Chat Completion.

v0.2 API deprecated

April 30, 2026The v0.2 API is deprecated. Its reference stays online for existing integrations, with a deprecation notice on each page. New integrations use the current API. See the v0.2 reference.

New documentation site

April 29, 2026The documentation site launched with the API reference for the current API: streams, chat completions and models.

Gemma 4 models available

April 2, 2026Gemma 4 models were made available on Overshoot on the day of their release. Read the post.