zoom-rtms

The zoom-rtms skill is a reference for Zoom Realtime Media Streams (RTMS), the platform capability that gives an application access to live audio, video, transcript, chat, and screen-share data from Zoom meetings, webinars, Video SDK sessions, and Zoom Contact Center Voice. It solves the problem of building real-time media pipelines without running a meeting bot in the media plane, using a WebSocket-based protocol built on open standards.

The skill explains that RTMS is primarily a backend media ingestion service that is event-triggered: the backend waits for an RTMS start webhook before it begins handling streams, and it can optionally pair with a Zoom Apps SDK frontend for in-client UI. It documents a two-phase WebSocket design that separates the control plane (a signaling socket handling authentication, media-server discovery, start/stop, and heartbeats) from the data plane (a media socket carrying binary audio, video, screen share, transcript, and chat data). It covers signature generation using HMAC-SHA256 over client ID, meeting or session ID, and stream ID; the heartbeat protocol with its timeouts; the fact that RTMS does not auto-reconnect and requires exponential-backoff reconnection logic; split versus unified connection modes; and geo-routed server URLs. Reference and example files cover the SDK quickstart, manual WebSocket control, media types, an end-to-end bot, AI integration, webhooks, data types, and environment variables.

Intended users are backend developers building live-media applications such as real-time transcription, meeting bots, AI companions, or contact-center voice analytics. The credential handling described is standard, deriving the HMAC signature from OAuth client secret or SDK secret held server-side. It is a technically deep, security-neutral reference within the Zoom partner plugin.