Web SDK | Tried Docs

Web SDK

The browser client for web calls: methods, options and events of TriedWebClient.

title: Web SDK description: The browser client for web calls: methods, options and events of TriedWebClient. order: 2#

@tried/web-sdk connects a browser to a web call. It wraps livekit-client: you create the call on your server with Create web call, pass access_token and livekit_url to the browser, and call startCall(). The package lives in packages/web-sdk of the Tried repo.

The class is also exported as RetellWebClient, and its methods and event names follow the Retell browser SDK, so code written for it needs only the import changed.

Quick start#

import { TriedWebClient } from "@tried/web-sdk";

const client = new TriedWebClient();

client.on("call_started", () => console.log("connected"));
client.on("agent_start_talking", () => console.log("agent speaking"));
client.on("update", ({ transcript }) => console.log(transcript));
client.on("call_ended", () => console.log("ended"));
client.on("error", (err) => console.error(err));

// from your server: POST /v3/create-web-call
const { access_token, livekit_url } = await fetch("/api/web-call", { method: "POST" }).then((r) => r.json());
await client.startCall({ accessToken: access_token, livekitUrl: livekit_url });

Constructor options#

new TriedWebClient(options?)

Option	Type	Description
`url`	string	LiveKit URL used when `startCall` gets no `livekitUrl`. Default `wss://livekit.tried.com.au`.
`audioContainer`	`() => HTMLElement \| null`	Where agent audio elements are attached. Default `document.body`.
`roomFactory`	function	Supplies the room object instead of a new `livekit-client` `Room`. Used by tests.

Methods#

Method	Description
`startCall(config)`	Connects to the room, turns on the microphone and emits `call_started`. Rejects (and emits `error`) when it fails. Throws if `accessToken` is missing.
`stopCall()`	Leaves the room, removes the audio elements and emits `call_ended`.
`mute()`	Turns the microphone off.
`unmute()`	Turns the microphone on.
`startAudioPlayback()`	Starts audio after a user gesture. Call it from a click handler if the browser blocked autoplay.
`on(event, fn)`, `once(event, fn)`, `off(event, fn)`, `removeAllListeners(event?)`	Manage listeners.

startCall config#

Field	Type	Required	Description
`accessToken`	string	Yes	`access_token` from Create web call.
`livekitUrl`	string	No	`livekit_url` from Create web call.
`sampleRate`	number	No	Microphone sample rate.
`captureDeviceId`	string	No	Microphone device ID.
`playbackDeviceId`	string	No	Speaker device ID.
`emitRawAudioSamples`	boolean	No	Emit `audio` events with the agent's samples.

The microphone is opened with echo cancellation, noise suppression and auto gain control on.

Properties#

Property	Type	Description
`isAgentTalking`	boolean	Whether the agent is speaking now.
`room`	object or null	The underlying LiveKit room while connected.

Events#

Event	Payload	When
`call_started`	none	The browser joined the room and the microphone is on.
`call_ready`	none	The agent joined the room.
`agent_start_talking`	none	The agent's state changed to `speaking`.
`agent_stop_talking`	none	The agent stopped speaking.
`update`	`{ transcript: [{ role, content }] }`	The live transcript changed. `role` is `agent` or `user`. Each segment updates in place as it streams.
`metadata`	any	The agent sent a JSON data message (no topic, or topic `metadata`).
`audio`	`Float32Array`	Agent audio samples, only with `emitRawAudioSamples`.
`call_ended`	none	The call ended: `stopCall()`, the room disconnected, or the agent left. Emitted once.
`error`	unknown	Connection failure, a transcript stream error, or an error thrown by one of your listeners.

The agent's speaking state comes from the lk.agent.state participant attribute, and transcripts from the lk.transcription text stream that the agent worker publishes.