@whereby.com/assistant-sdk

whereby.com/assistant-sdk

The @whereby.com/assistant-sdk lets you run headless participants in Node.js. Assistants can join rooms, combine audio from all participants, send messages, and perform in-room actions such as starting recordings. It powers use cases like transcriptions, captions or streaming to realtime AI agents. Use this SDK if you want to connect AI or backend services directly into Whereby rooms. For browser apps, start with the Browser SDK or React Native SDK.

Installation

npm install @whereby.com/assistant-sdk

or

yarn add @whereby.com/assistant-sdk

or

pnpm add @whereby.com/assistant-sdk

[!IMPORTANT] Assistants require FFmpeg to be installed and available in your system PATH if you want to use combined audio streams.

Usage

In order to use assistants, you must first create an assistant in your Whereby dashboard. This will give you an API key which you can then pass into your Assistant to allow it to join rooms. See here for more details - Assistants Documentation

Getting started

import "@whereby.com/assistant-sdk/polyfills"; // Required to run in Node.js
import { Assistant, AUDIO_STREAM_READY } from "@whereby.com/assistant-sdk";

async function main() {
    // Create an assistant instance
    const assistant = new Assistant({
        roomUrl: "https://your-subdwhereby.com/your-room", // Room URL to join
    });

    // Join the room
    try {
        await assistant.joinRoom(roomUrl);
    } catch (error) {
        console.error("Failed to join room:", error);
        return;
    }

    // To receive all audio data from the call:
    // You can now use the audio data in a transcription service, save to file, etc.
    const combinedAudioSink = assistant.getCombinedAudioSink();
}

main();

Core Concepts

Combined Audio

Assistants can output a single MediaStream that mixes all participant audio. Ideal for transcription, audio only recording or passing to realtime AI services. FFmpeg is required for this feature.

In-room Actions

Assistants can perform common room operations:

• Send and receive chat messages
• Start and stop cloud recordings
• Spotlight participants
• Request mic/camera changes

Trigger API

The Trigger API allows you to listen for specific webhooks, and create your assistant when those webhooks are received. This is useful for creating assistants on-demand, for example when a meeting starts. See the Trigger API docs for more details.

The Trigger API:

• Runs a lightweight server to listen for webhooks
• You define webhookTrigger - functions that decide whether to start an assistant based on the webhook payload.
• When the trigger condition is met, a TRIGGER_EVENT_SUCCESS event is emitted with the webhook payload, and you can create your assistant.

Typical usage:

import "@whereby.com/assistant-sdk/polyfills"; // Required to run in Node.js
import { Assistant, Trigger, TRIGGER_EVENT_SUCCESS, AUDIO_STREAM_READY } from "@whereby.com/assistant-sdk";

let hasJoinedRoom = false;

const trigger = new Trigger({
    webhookTriggers: {
        "room.client.joined": () => !hasJoinedRoom, // Start an assistant when first client joins
    },
    port: 3000, // Port to listen on
});

trigger.on(TRIGGER_EVENT_SuCCESS, async ({ roomUrl }) => {
    // Create and start your assistant when the trigger condition is met
    const assistant = new Assistant({
        roomUrl,
        startCombinedAudioStream: true,
        assistantKey: "your-assistant-key",
    });

    try {
        await assistant.joinRoom(roomUrl);
        hasJoinedRoom = true;
    } catch (error) {
        console.error("Failed to join room:", error);
        return;
    }

});

trigger.start();

Learn more

• Assistant SDK API reference - API Reference
• Trigger API docs - API Reference Assistant example app