Developer docs
Integrate voice_orch into your product
Voice Orchestrator exposes every deployment as a real-time WebSocket. You can either drop our embeddable widget on any page, or speak the protocol directly from your own UI. Both routes use the same backend, the same events, and the same ephemeral-token security model.
Start here
Quickstart
Five-minute integration: stand up a token-mint endpoint on your server, drop one HTML tag on your page, and you have a working voice agent.
For most teams
Widget reference
Full attribute, event, and theming reference for the <voice-orch-mic> web component. Works in React, Vue, Svelte, or plain HTML.
For custom UIs
WebSocket protocol
Every event the runtime emits and consumes, when it fires, and what the payload looks like. Use this if you want to build your own mic + playback UI.
As a model
Daisy Voice model
A configurable open-source voice pipeline: one API key, one WebSocket, swappable speech-to-text / language / speech models, and client-executed tools for RAG and MCP.
On your hardware
Self-hosted container
Run the whole pipeline as one container in your own network. Same realtime WebSocket the hosted API serves, pulled with a key from the console.
Before you ship
Production checklist
Origin allowlist, key rotation, token TTLs, rate limits, and the security model behind the token-mint pattern.
The integration in one diagram
The deployment key never touches the browser. Your server holds it; the browser receives only short-lived per-session tokens.