### Install LiveAvatar Plugin

Source: https://github.com/livekit/agents-js/blob/main/plugins/liveavatar/README.md

Install the LiveAvatar plugin using npm or pnpm.

```bash
npm install @livekit/agents-plugin-liveavatar
```

```bash
pnpm add @livekit/agents-plugin-liveavatar
```

--------------------------------

### Install MiniMax Plugin

Source: https://github.com/livekit/agents-js/blob/main/plugins/minimax/README.md

Install the MiniMax plugin for LiveKit Agents using pnpm. Ensure you have Node.js and pnpm installed.

```bash
pnpm add @livekit/agents-plugin-minimax
```

--------------------------------

### Install LiveKit Agents Core Library

Source: https://github.com/livekit/agents-js/blob/main/README.md

Install the core Agents library along with any desired plugins using pnpm. This command installs the main package for agent functionality.

```bash
pnpm install @livekit/agents
```

--------------------------------

### Install Runway Plugin

Source: https://github.com/livekit/agents-js/blob/main/plugins/runway/README.md

Install the Runway plugin for LiveKit Agents using npm.

```bash
npm install @livekit/agents-plugin-runway
```

--------------------------------

### Migrate Installation

Source: https://github.com/livekit/agents-js/blob/main/plugins/mistral/README.md

Install the new package instead of the deprecated one to migrate your project.

```diff
- npm install @livekit/agents-plugin-mistral
+ npm install @livekit/agents-plugin-mistralai
```

--------------------------------

### Running an Example Agent

Source: https://github.com/livekit/agents-js/blob/main/CLAUDE.md

Command to build and run a basic example agent. Ensure all environment variables for LiveKit and any required providers are set before execution.

```bash
pnpm build && node ./examples/src/basic_agent.ts dev --log-level=debug
```

--------------------------------

### Multi-Agent Handoff Example

Source: https://github.com/livekit/agents-js/blob/main/README.md

Demonstrates a multi-agent system where an introductory agent gathers user information and then hands off to a story agent. This setup is useful for complex conversational flows requiring different agent specializations.

```typescript
type StoryData = {
  name?: string;
  location?: string;
};

class IntroAgent extends voice.Agent<StoryData> {
  constructor() {
    super({
      instructions: `You are a story teller. Your goal is to gather a few pieces of information from the user to make the story personalized and engaging. Ask the user for their name and where they are from.`,
      tools: {
        informationGathered: llm.tool({
          description:
            'Called when the user has provided the information needed to make the story personalized and engaging.',
          parameters: z.object({
            name: z.string().describe('The name of the user'),
            location: z.string().describe('The location of the user'),
          }),
          execute: async ({ name, location }, { ctx }) => {
            ctx.userData.name = name;
            ctx.userData.location = location;

            return llm.handoff({
              agent: new StoryAgent(name, location),
              returns: "Let's start the story!",
            });
          },
        }),
      },
    });
  }

  // Use inheritance to create agent with custom hooks
  async onEnter() {
    this.session.generateReply({
      instructions: '"greet the user and gather information"',
    });
  }
}

class StoryAgent extends voice.Agent<StoryData> {
  constructor(name: string, location: string) {
    super({
      instructions: `You are a storyteller. Use the user's information in order to make the story personalized.
        The user's name is ${name}, from ${location}`,
    });
  }

  async onEnter() {
    this.session.generateReply();
  }
}

export default defineAgent({
  prewarm: async (proc: JobProcess) => {
    proc.userData.vad = await silero.VAD.load();
  },
  entry: async (ctx: JobContext) => {
    await ctx.connect();
    const participant = await ctx.waitForParticipant();
    console.log('participant joined: ', participant.identity);

    const userdata: StoryData = {};

    const session = new voice.AgentSession({
      vad: ctx.proc.userData.vad! as silero.VAD,
      stt: new inference.STT({ model: 'deepgram/nova-3', language: 'en' }),
      llm: new inference.LLM({ model: 'openai/gpt-4.1-mini' }),
      tts: new inference.TTS({ model: 'cartesia/sonic-3', voice: '9626c31c-bec5-4cca-baa8-f8ba9e84c8bc' }),
      userData: userdata,
    });

    await session.start({
      agent: new IntroAgent(),
      room: ctx.room,
    });
  },
});

```

--------------------------------

### Install Perplexity Plugin

Source: https://github.com/livekit/agents-js/blob/main/plugins/perplexity/README.md

Install the Perplexity plugin for LiveKit Agents using pnpm.

```bash
pnpm add @livekit/agents-plugin-perplexity
```

--------------------------------

### Install Mistral AI Plugin

Source: https://github.com/livekit/agents-js/blob/main/plugins/mistralai/README.md

Install the Mistral AI plugin for LiveKit Agents using npm.

```bash
npm install @livekit/agents-plugin-mistralai
```

--------------------------------

### Install pnpm

Source: https://github.com/livekit/agents-js/blob/main/README.md

Install pnpm globally using npm if it is not already installed. This is a prerequisite for installing the LiveKit Agents library.

```bash
npm install -g pnpm
```

--------------------------------

### AvatarSession.start Method

Source: https://github.com/livekit/agents-js/blob/main/plugins/bey/README.md

Starts the avatar session and connects it to a LiveKit room.

```APIDOC
## AvatarSession.start

Starts the avatar session and connects it to the room.

### Method

`start(agentSession: AgentSession, room: Room, options?: StartOptions): Promise<void>`

### Parameters

#### Path Parameters

None

#### Query Parameters

None

#### Request Body

None

### Options for `start` method

- **livekitUrl** (string) - Optional - The LiveKit server URL. Can also be set via `LIVEKIT_URL` environment variable.
- **livekitApiKey** (string) - Optional - Your LiveKit API key. Can also be set via `LIVEKIT_API_KEY` environment variable.
- **livekitApiSecret** (string) - Optional - Your LiveKit API secret. Can also be set via `LIVEKIT_API_SECRET` environment variable.

### Request Example

```typescript
import { AvatarSession } from '@livekit/agents-plugin-bey';
import { AgentSession } from '@livekit/agents';

const avatarSession = new AvatarSession({
  avatarId: 'your-avatar-id',
  apiKey: process.env.BEY_API_KEY,
});

await avatarSession.start(agentSession, room, {
  livekitUrl: process.env.LIVEKIT_URL,
  livekitApiKey: process.env.LIVEKIT_API_KEY,
  livekitApiSecret: process.env.LIVEKIT_API_SECRET,
});
```

### Response

#### Success Response (200)

- **void** - Indicates the avatar session has started successfully.

#### Response Example

(No specific response body for success, the promise resolves to void.)
```

--------------------------------

### Initialize and Start Avatar Session

Source: https://github.com/livekit/agents-js/blob/main/plugins/lemonslice/README.md

Demonstrates how to instantiate an AvatarSession with an image URL and start the session within a LiveKit room.

```typescript
import { AvatarSession } from '@livekit/agents-plugin-lemonslice';

const avatar = new AvatarSession({
  agentImageUrl: 'publicly-accessible-image-url',
  apiKey: 'your-lemonslice-api-key',
  extraPayload: {
    aspect_ratio: '9x16',
  },
});

await avatar.start(agentSession, room);
```

--------------------------------

### Install D-ID Plugin

Source: https://github.com/livekit/agents-js/blob/main/plugins/did/README.md

Install the D-ID plugin for LiveKit Agents using npm.

```bash
npm install @livekit/agents-plugin-did
```

--------------------------------

### Start Beyond Presence Avatar Session in LiveKit Agent

Source: https://github.com/livekit/agents-js/blob/main/plugins/bey/README.md

Demonstrates how to create and start an AvatarSession using the @livekit/agents-plugin-bey. This involves initializing the session with avatar details and LiveKit connection options, then starting it within an AgentSession.

```typescript
import { AvatarSession } from '@livekit/agents-plugin-bey';
import { AgentSession } from '@livekit/agents';

// Create an avatar session
const avatarSession = new AvatarSession({
  avatarId: 'your-avatar-id', // optional, defaults to stock avatar
  apiKey: process.env.BEY_API_KEY, // optional if set via env var
});

// Start the avatar in your agent
await avatarSession.start(agentSession, room, {
  livekitUrl: process.env.LIVEKIT_URL,
  livekitApiKey: process.env.LIVEKIT_API_KEY,
  livekitApiSecret: process.env.LIVEKIT_API_SECRET,
});
```

--------------------------------

### Basic LLM Setup with Google AI Studio

Source: https://github.com/livekit/agents-js/blob/main/plugins/google/README.md

Initialize the LLM client using an API key from Google AI Studio. Alternatively, set the GOOGLE_API_KEY environment variable.

```typescript
import { LLM } from '@livekit/agents-plugin-google';

// Using Google AI Studio (API Key)
const llm = new LLM({
  model: 'gemini-1.5-flash',
  apiKey: 'your-api-key', // or set GOOGLE_API_KEY env var
});
```

--------------------------------

### Install LemonSlice Plugin

Source: https://github.com/livekit/agents-js/blob/main/plugins/lemonslice/README.md

Commands to install the LemonSlice plugin package using common Node.js package managers.

```bash
npm install @livekit/agents-plugin-lemonslice
pnpm add @livekit/agents-plugin-lemonslice
yarn add @livekit/agents-plugin-lemonslice
```

--------------------------------

### Run Agent Production Mode

Source: https://github.com/livekit/agents-js/blob/main/README.md

Starts the agent with production-ready optimizations. This command is used for deploying the agent to a production environment.

```bash
pnpm run build && node ./examples/src/restaurant_agent.ts start

```

--------------------------------

### Build Project with pnpm

Source: https://github.com/livekit/agents-js/blob/main/README.md

Build the project using the pnpm package manager. Ensure pnpm is installed and configured for the project.

```bash
pnpm build
```

--------------------------------

### Install Tavus Plugin

Source: https://github.com/livekit/agents-js/blob/main/plugins/tavus/README.md

Install the Tavus plugin using npm. This command adds the necessary package to your project.

```bash
npm install @livekit/agents-plugin-tavus
```

--------------------------------

### Basic LLM Setup with Vertex AI

Source: https://github.com/livekit/agents-js/blob/main/plugins/google/README.md

Initialize the LLM client for Vertex AI, specifying the model, project ID, and location. Alternatively, set GOOGLE_GENAI_USE_VERTEXAI, GOOGLE_CLOUD_PROJECT, and GOOGLE_CLOUD_LOCATION environment variables.

```typescript
import { LLM } from '@livekit/agents-plugin-google';

// Using Vertex AI
const llm = new LLM({
  model: 'gemini-1.5-pro',
  vertexai: true,
  project: 'your-project-id', // or set GOOGLE_CLOUD_PROJECT env var
  location: 'us-central1', // or set GOOGLE_CLOUD_LOCATION env var
});
```

--------------------------------

### Initialize and Start Avatar Session

Source: https://github.com/livekit/agents-js/blob/main/plugins/liveavatar/README.md

Initialize an AvatarSession with configuration options and start it with an agent session and room details. Environment variables can be used for fallback configuration.

```typescript
import { AvatarSession } from '@livekit/agents-plugin-liveavatar';
import { AgentSession } from '@livekit/agents';

const avatarSession = new AvatarSession({
  avatarId: 'your-avatar-id', // or via LIVEAVATAR_AVATAR_ID
  apiKey: process.env.LIVEAVATAR_API_KEY,
  videoQuality: 'high', // optional: 'very_high' | 'high' | 'medium' | 'low'
});

await avatarSession.start(agentSession, room, {
  livekitUrl: process.env.LIVEKIT_URL,
  livekitApiKey: process.env.LIVEKIT_API_KEY,
  livekitApiSecret: process.env.LIVEKIT_API_SECRET,
});
```

--------------------------------

### AvatarSession.start

Source: https://github.com/livekit/agents-js/blob/main/plugins/liveavatar/README.md

Starts the avatar session, establishing a LiveAvatar streaming session and routing audio.

```APIDOC
## AvatarSession.start(agentSession, room, options?)

Starts the avatar session, brings up a LiveAvatar streaming session, opens the realtime websocket, and routes the agent's audio output through to the avatar.

### Parameters

- **agentSession** - The AgentSession object.
- **room** - The LiveKit Room object.
- **options** (StartOptions) - Optional configuration for starting the session.

#### StartOptions

- **livekitUrl** (string) - Optional - Falls back to `LIVEKIT_URL`.
- **livekitApiKey** (string) - Optional - Falls back to `LIVEKIT_API_KEY`.
- **livekitApiSecret** (string) - Optional - Falls back to `LIVEKIT_API_SECRET`.
```

--------------------------------

### Install @livekit/agents-plugin-bey

Source: https://github.com/livekit/agents-js/blob/main/plugins/bey/README.md

Installs the Beyond Presence plugin for LiveKit Agents using npm or pnpm. This package provides the necessary tools to integrate virtual avatars.

```bash
npm install @livekit/agents-plugin-bey
```

```bash
pnpm add @livekit/agents-plugin-bey
```

--------------------------------

### Dockerfile for Pre-downloading Plugin Assets

Source: https://github.com/livekit/agents-js/blob/main/CLAUDE.md

Example Dockerfile snippet demonstrating how to pre-download plugin assets. This approach caches the download layer separately from application code.

```dockerfile
COPY package.json pnpm-lock.yaml ./ 
RUN pnpm install --frozen-lockfile
RUN npx livekit-agents download-files
COPY src/ ./src
RUN pnpm build
```

--------------------------------

### Define a Voice Agent with Phonic Integration

Source: https://github.com/livekit/agents-js/blob/main/plugins/phonic/README.md

This example demonstrates how to define a voice agent using the Phonic plugin. It includes setting up LLM tools and configuring the Phonic realtime model with specific voice and audio speed settings. The agent is configured to greet the user and can toggle lights using a defined tool.

```typescript
// SPDX-FileCopyrightText: 2026 LiveKit, Inc.
//
// SPDX-License-Identifier: Apache-2.0
import { type JobContext, ServerOptions, cli, defineAgent, llm, voice } from '@livekit/agents';
import * as phonic from '@livekit/agents-plugin-phonic';
import { fileURLToPath } from 'node:url';
import { z } from 'zod';

const toggleLight = llm.tool({
  description: 'Toggle a light on or off. Available lights are A05, A06, A07, and A08.',
  parameters: z.object({
    light_id: z.string().describe('The ID of the light to toggle'),
    state: z.enum(['on', 'off']).describe('Whether to turn the light on or off'),
  }),
  execute: async ({ light_id, state }) => {
    console.log(`Turning ${state} light ${light_id}`);
    await new Promise((resolve) => setTimeout(resolve, 1_000));
    return `Light ${light_id} turned ${state}`;
  },
});

export default defineAgent({
  entry: async (ctx: JobContext) => {
    const agent = new voice.Agent({
      instructions: 'You are a helpful voice AI assistant named Alex.',
      tools: {
        toggle_light: toggleLight,
      },
    });

    const session = new voice.AgentSession({
      // Uses PHONIC_API_KEY environment variable when apiKey is not provided
      llm: new phonic.realtime.RealtimeModel({
        voice: 'sabrina',
        audioSpeed: 1.2,
      }),
    });

    await session.start({
      agent,
      room: ctx.room,
    });

    await ctx.connect();

    await session.generateReply({
      instructions: 'Greet the user, asking about their day.',
    });
  },
});

cli.runApp(new ServerOptions({ agent: fileURLToPath(import.meta.url) }));

```

--------------------------------

### Initialize and Start Trugen Avatar Session in TypeScript

Source: https://github.com/livekit/agents-js/blob/main/plugins/trugen/README.md

Demonstrates how to initialize and start a Trugen avatar session within a LiveKit agent. This involves creating an AvatarSession instance and calling its start method before the agent session begins. It requires LiveKit connection details and optionally a Trugen avatar ID.

```typescript
import { AvatarSession } from '@livekit/agents-plugin-trugen';

// inside your existing defineAgent entry function:

const avatarSession = new AvatarSession({
  avatarId: process.env.TRUGEN_AVATAR_ID, // optional
  apiKey: process.env.TRUGEN_API_KEY,
});

// must be called BEFORE agentSession.start()
await avatarSession.start(agentSession, room, {
  livekitUrl: process.env.LIVEKIT_URL,
  livekitApiKey: process.env.LIVEKIT_API_KEY,
  livekitApiSecret: process.env.LIVEKIT_API_SECRET,
});
```

--------------------------------

### Run Agent Development Mode

Source: https://github.com/livekit/agents-js/blob/main/README.md

Starts the agent worker in development mode, ready to accept connections. Ensure all required environment variables are set before running.

```bash
pnpm run build && node ./examples/src/restaurant_agent.ts dev

```

--------------------------------

### Install Sarvam AI Plugin for LiveKit Agents

Source: https://github.com/livekit/agents-js/blob/main/plugins/sarvam/README.md

Installs the Sarvam AI plugin for LiveKit Agents using pnpm. This package provides Text-to-Speech (TTS) and Speech-to-Text (STT) capabilities.

```bash
pnpm add @livekit/agents-plugin-sarvam
```

--------------------------------

### Create Expressive D-ID Agent

Source: https://github.com/livekit/agents-js/blob/main/plugins/did/README.md

Example of creating an expressive D-ID agent using the D-ID API. Ensure you replace <YOUR_API_KEY> with your actual D-ID API key.

```bash
curl -X POST https://api.d-id.com/agents \
  -H "Authorization: Basic <YOUR_API_KEY>" \
  -H "Content-Type: application/json" \
  -d '{ \
    "presenter": { \
      "type": "expressive", \
      "presenter_id": "public_mia_elegant@avt_TJ0Tq5" \
    }, \
    "preview_name": "My Expressive Agent" \
  }'
```

--------------------------------

### Install Trugen Plugin for LiveKit Agents

Source: https://github.com/livekit/agents-js/blob/main/plugins/trugen/README.md

Installs the Trugen plugin for LiveKit Agents using npm or pnpm. This package provides the necessary components to integrate Trugen's realtime avatars.

```bash
npm install @livekit/agents-plugin-trugen
```

```bash
pnpm add @livekit/agents-plugin-trugen
```

--------------------------------

### Simple Voice Agent with LiveKit Agents JS

Source: https://github.com/livekit/agents-js/blob/main/README.md

Defines a simple voice agent using LiveKit Agents JS. This example demonstrates setting up Speech-to-Text (STT), Large Language Model (LLM), Text-to-Speech (TTS), and Voice Activity Detection (VAD). It utilizes the LiveKit Inference Gateway, so no third-party API keys are required.

```typescript
import {
  type JobContext,
  type JobProcess,
  WorkerOptions,
  cli,
  defineAgent,
  llm,
  voice,
  inference,
} from '@livekit/agents';
import * as silero from '@livekit/agents-plugin-silero';
import { fileURLToPath } from 'node:url';
import { z } from 'zod';

const lookupWeather = llm.tool({
  description: 'Used to look up weather information.',
  parameters: z.object({
    location: z.string().describe('The location to look up weather information for'),
  }),
  execute: async ({ location }, { ctx }) => {
    return { weather: 'sunny', temperature: 70 };
  },
});

export default defineAgent({
  prewarm: async (proc: JobProcess) => {
    proc.userData.vad = await silero.VAD.load();
  },
  entry: async (ctx: JobContext) => {
    const agent = new voice.Agent({
      instructions: 'You are a friendly voice assistant built by LiveKit.',
      tools: { lookupWeather },
    });

    const session = new voice.AgentSession({
      // Speech-to-text (STT) is your agent's ears, turning the user's speech into text that the LLM can understand
      // See all available models at https://docs.livekit.io/agents/models/stt/
      stt: new inference.STT({ model: 'deepgram/nova-3', language: 'en' }),
      // A Large Language Model (LLM) is your agent's brain, processing user input and generating a response
      // See all available models at https://docs.livekit.io/agents/models/llm/
      llm: new inference.LLM({ model: 'openai/gpt-4.1-mini' }),
      // Text-to-speech (TTS) is your agent's voice, turning the LLM's text into speech that the user can hear
      // See all available models as well as voice selections at https://docs.livekit.io/agents/models/tts/
      tts: new inference.TTS({ model: 'cartesia/sonic-3', voice: '9626c31c-bec5-4cca-baa8-f8ba9e84c8bc' }),
      // VAD and turn detection are used to determine when the user is speaking and when the agent should respond
      // See more at https://docs.livekit.io/agents/build/turns
      vad: ctx.proc.userData.vad! as silero.VAD,
      turnDetection: new livekit.turnDetector.MultilingualModel(),
      // to use realtime model, replace the stt, llm, tts and vad with the following
      // llm: new openai.realtime.RealtimeModel(),
    });

    await session.start({
      agent,
      room: ctx.room,
    });

    await session.generateReply({
      instructions: 'greet the user and ask about their day',
    });
  },
});

cli.runApp(new WorkerOptions({ agent: fileURLToPath(import.meta.url) }));

```

--------------------------------

### Install Cerebras Plugin for LiveKit Agents

Source: https://github.com/livekit/agents-js/blob/main/plugins/cerebras/README.md

Use this command to add the Cerebras plugin to your LiveKit agents project.

```bash
pnpm add @livekit/agents-plugin-cerebras
```

--------------------------------

### AvatarSession Class

Source: https://github.com/livekit/agents-js/blob/main/plugins/trugen/README.md

The AvatarSession class is the primary interface for initializing and starting a Trugen avatar session within a LiveKit agent.

```APIDOC
## AvatarSession

### Description
Initializes a new Trugen avatar session to be used within a LiveKit agent.

### Constructor Options
- **avatarId** (string) - Optional - The specific avatar ID to use.
- **apiUrl** (string) - Optional - The Trugen API URL (defaults to https://api.trugen.ai).
- **apiKey** (string) - Optional - The Trugen API key.
- **avatarParticipantIdentity** (string) - Optional - Identity for the avatar participant (defaults to 'trugen-avatar').
- **avatarParticipantName** (string) - Optional - Display name for the avatar participant (defaults to 'Trugen Avatar').

### Methods
#### start(agentSession, room, options)
Starts the avatar session and connects it to the specified LiveKit room.

- **agentSession** (Object) - Required - The current agent session instance.
- **room** (Object) - Required - The LiveKit room instance.
- **options** (Object) - Optional - Connection options including livekitUrl, livekitApiKey, and livekitApiSecret.

### Usage Example
```typescript
const avatarSession = new AvatarSession({
  avatarId: process.env.TRUGEN_AVATAR_ID,
  apiKey: process.env.TRUGEN_API_KEY,
});

await avatarSession.start(agentSession, room, {
  livekitUrl: process.env.LIVEKIT_URL,
  livekitApiKey: process.env.LIVEKIT_API_KEY,
  livekitApiSecret: process.env.LIVEKIT_API_SECRET,
});
```
```

--------------------------------

### Initialize Sarvam AI TTS

Source: https://github.com/livekit/agents-js/blob/main/plugins/sarvam/README.md

Initializes the Text-to-Speech (TTS) service using Sarvam AI's Bulbul models. Supports specifying speaker, model, and target language. This example uses the default streaming WebSocket path.

```typescript
import * as sarvam from '@livekit/agents-plugin-sarvam';

const tts = new sarvam.TTS({
  speaker: 'anushka',
  model: 'bulbul:v2',
  targetLanguageCode: 'en-IN',
});
```

--------------------------------

### Initialize Sarvam AI STT for Transcription

Source: https://github.com/livekit/agents-js/blob/main/plugins/sarvam/README.md

Initializes the Speech-to-Text (STT) service using Sarvam AI's Saaras models for transcription. Supports specifying the model and language code. This example uses the default streaming WebSocket path.

```typescript
import * as sarvam from '@livekit/agents-plugin-sarvam';

const stt = new sarvam.STT({
  model: 'saaras:v3',
  languageCode: 'en-IN',
  mode: 'transcribe',
});
```

--------------------------------

### Initialize and Stream Baseten STT

Source: https://github.com/livekit/agents-js/blob/main/plugins/baseten/README.md

Shows how to configure the STT service and iterate over a stream of speech events.

```typescript
import { STT } from 'livekit-plugin-baseten';

const stt = new STT({
    apiKey: process.env.BASETEN_API_KEY,
    modelId: process.env.BASETEN_STT_MODEL_ID
});

const stream = stt.stream();
for await (const event of stream) {
    // Handle speech events
}
```

--------------------------------

### Initialize Baseten LLM Service

Source: https://github.com/livekit/agents-js/blob/main/plugins/baseten/README.md

Demonstrates how to instantiate the LLM class from the Baseten plugin using an API key and model identifier.

```typescript
import { LLM } from 'livekit-plugin-baseten';

const llm = new LLM({
    model: 'openai/gpt-4o-mini',
    apiKey: process.env.BASETEN_API_KEY
});
```

--------------------------------

### AnamAPI Class Definition

Source: https://github.com/livekit/agents-js/blob/main/plugins/anam/etc/agents-plugin-anam.api.md

Defines the AnamAPI class for interacting with Anam services. It includes methods for creating session tokens and starting engine sessions.

```typescript
import type { Room } from '@livekit/rtc-node';
import { voice } from '@livekit/agents';

// @public (undocumented)
export class AnamAPI {
    constructor(apiKey: string, apiUrl?: string, conn?: APIConnectOptions);
    // (undocumented)
    createSessionToken(params: {
        personaConfig: PersonaConfig;
        livekitUrl?: string;
        livekitToken?: string;
        sessionOptions?: SessionOptions;
    }): Promise<{ 
        sessionToken: string;
    }>;
    // (undocumented)
    startEngineSession(params: {
        sessionToken: string;
    }): Promise<{ 
        sessionId: string;
    }>;
}
```

--------------------------------

### Realtime Model and TTS Initialization

Source: https://github.com/livekit/agents-js/blob/main/plugins/google/README.md

Instantiate the realtime model and Text-to-Speech (TTS) client for Gemini. Note that TTS is currently in Beta.

```typescript
import * as google from '@livekit/agents-plugin-google';

const realtimeModel = new google.realtime.RealtimeModel({
  model: 'gemini-2.5-flash-native-audio-preview-12-2025',
})
const geminiTTS = new google.beta.TTS(),
```

--------------------------------

### Initialize Perplexity LLM Client

Source: https://github.com/livekit/agents-js/blob/main/plugins/perplexity/README.md

Initialize the Perplexity LLM client. The API key is automatically picked up from the PERPLEXITY_API_KEY environment variable if not provided.

```typescript
import { LLM } from '@livekit/agents-plugin-perplexity';

const llm = new LLM({
  model: 'sonar-pro',
  // apiKey is picked up from PERPLEXITY_API_KEY if omitted
});
```

--------------------------------

### Common Commands for LiveKit Agents.js

Source: https://github.com/livekit/agents-js/blob/main/CLAUDE.md

A collection of pnpm commands for building, testing, linting, and formatting the LiveKit Agents.js monorepo. Use these for managing the project's development lifecycle.

```bash
pnpm build                  # Build all packages (turbo)
pnpm build:agents           # Build only @livekit/agents and its deps
pnpm build:plugins          # Build only plugins and their deps
pnpm clean:build            # Clean dist/ dirs then rebuild
pnpm test                   # Run all tests (vitest)
pnpm test -- --testPathPattern=agents/src/llm  # Run tests by path
pnpm test -- --testNamePattern="chat context"  # Run tests by name
pnpm test:watch             # Watch mode
pnpm lint                   # ESLint all packages
pnpm lint:fix               # ESLint with auto-fix
pnpm format:check           # Prettier check
pnpm format:write           # Prettier format
pnpm api:check              # API Extractor validation
pnpm api:update             # Update API declarations
```

--------------------------------

### Configure Vertex AI Authentication

Source: https://github.com/livekit/agents-js/blob/main/plugins/google/README.md

Set up authentication for Vertex AI by logging in with gcloud, setting the project ID, and enabling Vertex AI usage via an environment variable.

```bash
gcloud auth application-default login
export GOOGLE_CLOUD_PROJECT=your-project-id
export GOOGLE_GENAI_USE_VERTEXAI=true
```

--------------------------------

### Initialize LLM with Mistral AI

Source: https://github.com/livekit/agents-js/blob/main/plugins/mistralai/README.md

Initialize the LLM component using a Mistral AI model. The API key defaults to process.env.MISTRAL_API_KEY.

```typescript
import * as mistral from '@livekit/agents-plugin-mistralai';

const llm = new mistral.LLM({
  model: 'mistral-small-latest',
  // apiKey defaults to process.env.MISTRAL_API_KEY
});
```

--------------------------------

### Initialize and Synthesize Baseten TTS

Source: https://github.com/livekit/agents-js/blob/main/plugins/baseten/README.md

Configures the TTS service and demonstrates how to synthesize text into an audio stream.

```typescript
import { TTS } from 'livekit-plugin-baseten';

const tts = new TTS({
    apiKey: process.env.BASETEN_API_KEY,
    modelEndpoint: 'your-model-endpoint-url'
});

const stream = tts.synthesize('Hello world');
for await (const frame of stream) {
    // Process audio frames
}
```

--------------------------------

### Debugging Individual Plugins

Source: https://github.com/livekit/agents-js/blob/main/CLAUDE.md

Command to build and run a test file for debugging a specific plugin. This method avoids the need for the `defineAgent` wrapper.

```bash
pnpm build && node ./examples/src/test_my_plugin.ts
```

--------------------------------

### AvatarSession Constructor

Source: https://github.com/livekit/agents-js/blob/main/plugins/liveavatar/README.md

Initializes a new AvatarSession with various configuration options.

```APIDOC
## AvatarSession Constructor Options

- **avatarId** (string) - Optional - The LiveAvatar avatar id. Falls back to `LIVEAVATAR_AVATAR_ID`.
- **apiUrl** (string) - Optional - Override the LiveAvatar API base URL.
- **apiKey** (string) - Optional - Your LiveAvatar API key. Falls back to `LIVEAVATAR_API_KEY`.
- **isSandbox** (boolean) - Optional - Use the LiveAvatar sandbox (1 minute connection limit). Defaults to `false`.
- **videoQuality** ('very_high' | 'high' | 'medium' | 'low') - Optional - Avatar video quality requested from the service. When omitted, the LiveAvatar service decides.
- **avatarParticipantIdentity** (string) - Optional - Identity for the avatar participant. Defaults to `'liveavatar-avatar-agent'`.
- **avatarParticipantName** (string) - Optional - Display name for the avatar participant. Defaults to `'liveavatar-avatar-agent'`.
- **connOptions** (APIConnectOptions) - Optional - API retry/timeout options.
```

--------------------------------

### Run Basic Agent with Debug Logging

Source: https://github.com/livekit/agents-js/blob/main/README.md

Execute the basic agent script with debug logging enabled. This is useful for testing changes and plugins.

```bash
node ./examples/src/basic_agent.ts dev --log-level=debug
```

--------------------------------

### License Header for New Files

Source: https://github.com/livekit/agents-js/blob/main/CLAUDE.md

All new files must include this SPDX license header.

```typescript
// SPDX-FileCopyrightText: 2026 LiveKit, Inc.
//
// SPDX-License-Identifier: Apache-2.0

```

--------------------------------

### STTOptions Interface

Source: https://github.com/livekit/agents-js/blob/main/plugins/soniox/etc/agents-plugin-soniox.api.md

Defines the configuration options available for the STT service.

```APIDOC
## Interface STTOptions

### Description
Configuration options for initializing the STT service.

### Properties
- `apiKey` (string) - Optional API key for authentication.
- `baseUrl` (string) - The base URL for the STT service.
- `clientReferenceId` (string) - Optional client reference ID for tracking.
- `context` (ContextObject | string) - Optional context information to improve recognition accuracy.
- `enableLanguageIdentification` (boolean) - Enables automatic language detection.
- `enableSpeakerDiarization` (boolean) - Enables speaker diarization to distinguish between speakers.
- `languageHints` (string[]) - An array of language codes to hint at possible languages.
- `languageHintsStrict` (boolean) - If true, only languages specified in `languageHints` will be considered.
- `maxEndpointDelayMs` (number) - Maximum delay in milliseconds before an endpoint is considered finished.
- `model` (string) - The STT model to use.
- `numChannels` (number) - The number of audio channels.
- `sampleRate` (number) - The audio sample rate.
- `translation` (TranslationConfig) - Configuration for translation services.
```

--------------------------------

### Set Google AI Studio API Key via Environment Variable

Source: https://github.com/livekit/agents-js/blob/main/plugins/google/README.md

Export your Google AI Studio API key as an environment variable for authentication.

```bash
export GOOGLE_API_KEY=your-api-key
```

--------------------------------

### Initialize STT for Realtime Streaming

Source: https://github.com/livekit/agents-js/blob/main/plugins/mistralai/README.md

Initialize the STT component for realtime streaming. This requires a VAD and uses a specific realtime model. If no VAD is provided, Silero VAD is loaded automatically.

```typescript
// Realtime streaming (requires a VAD)
const stt = new mistral.STT({
  model: 'voxtral-mini-transcribe-realtime-2602',
});
```

--------------------------------

### Advanced LLM Configuration

Source: https://github.com/livekit/agents-js/blob/main/plugins/google/README.md

Configure advanced LLM parameters such as temperature, max output tokens, topP, topK, and tool choice for more fine-tuned model behavior.

```typescript
const llm = new LLM({
  model: 'gemini-1.5-pro',
  temperature: 0.7,
  maxOutputTokens: 2048,
  topP: 0.8,
  topK: 40,
  toolChoice: 'auto',
});
```

--------------------------------

### Initialize Perplexity Agent API LLM Client

Source: https://github.com/livekit/agents-js/blob/main/plugins/perplexity/README.md

Initialize the Perplexity LLM client for the Agent API. This client uses a specific base URL and disables WebSocket transport.

```typescript
import { responses } from '@livekit/agents-plugin-perplexity';

const llm = new responses.LLM({
  model: 'sonar-pro',
  // apiKey is picked up from PERPLEXITY_API_KEY if omitted
});
```

--------------------------------

### SessionOptions Type

Source: https://github.com/livekit/agents-js/blob/main/plugins/anam/etc/agents-plugin-anam.api.md

Options for configuring a session.

```APIDOC
## Type: SessionOptions

### Description
Defines options that can be applied to a session, such as video dimensions.

### Properties

- **videoWidth** (number) - Optional - The desired width of the video stream.
- **videoHeight** (number) - Optional - The desired height of the video stream.
```

--------------------------------

### AvatarSession Constructor

Source: https://github.com/livekit/agents-js/blob/main/plugins/bey/README.md

Initializes a new AvatarSession for managing Beyond Presence avatars.

```APIDOC
## AvatarSession

Creates a new Beyond Presence avatar session.

### Constructor Options

- **avatarId** (string) - Optional - The avatar ID to use. Defaults to stock avatar.
- **apiUrl** (string) - Optional - The Bey API URL. Defaults to `https://api.bey.dev`.
- **apiKey** (string) - Optional - Your Bey API key. Can also be set via `BEY_API_API_KEY` environment variable.
- **avatarParticipantIdentity** (string) - Optional - The identity for the avatar participant. Defaults to `'bey-avatar-agent'`.
- **avatarParticipantName** (string) - Optional - The name for the avatar participant. Defaults to `'bey-avatar-agent'`.
- **connOptions** (APIConnectOptions) - Optional - Connection options for retry logic.
```

--------------------------------

### Initialize TTS with Mistral AI

Source: https://github.com/livekit/agents-js/blob/main/plugins/mistralai/README.md

Initialize the TTS component using a Mistral AI model and a specific voice preset.

```typescript
const tts = new mistral.TTS({
  model: 'voxtral-mini-tts-latest',
  voice: 'en_paul_neutral',
});
```

--------------------------------

### LLMOptions Interface

Source: https://github.com/livekit/agents-js/blob/main/plugins/perplexity/etc/agents-plugin-perplexity.api.md

Configuration options for the Perplexity LLM client.

```APIDOC
## LLMOptions Interface

### Description
Configuration options for the Perplexity LLM client.

### Fields

- **apiKey** (string) - Optional - The API key for Perplexity.
- **baseURL** (string) - Optional - The base URL for the Perplexity API.
- **client** (OpenAI) - Optional - An existing OpenAI client instance.
- **model** (string | PerplexityChatModels) - Required - The Perplexity model to use (e.g., 'sonar-pro').
- **parallelToolCalls** (boolean) - Optional - Whether to enable parallel tool calls.
- **temperature** (number) - Optional - The sampling temperature.
- **toolChoice** (llm.ToolChoice) - Optional - Configuration for tool choice.
- **topP** (number) - Optional - The nucleus sampling probability.
```

--------------------------------

### Migrate Import Path

Source: https://github.com/livekit/agents-js/blob/main/plugins/mistral/README.md

Update your import statements to use the new package name when migrating from the deprecated plugin.

```diff
- import * as mistral from '@livekit/agents-plugin-mistral';
+ import * as mistral from '@livekit/agents-plugin-mistralai';
```

--------------------------------

### Set Bey API Key Environment Variable

Source: https://github.com/livekit/agents-js/blob/main/plugins/bey/README.md

Sets the BEY_API_KEY environment variable with your developer API key obtained from the Bey creator dashboard. This key is required for authenticating with the Bey service.

```bash
export BEY_API_KEY=<your-bey-api-key>
```

--------------------------------

### Initialize Sarvam AI STT for Translation

Source: https://github.com/livekit/agents-js/blob/main/plugins/sarvam/README.md

Initializes the Speech-to-Text (STT) service using Sarvam AI's Saaras models for translation from an Indic language to English. Requires specifying a prompt for context. Authentication can be handled via the SARVAM_API_KEY environment variable or by passing the apiKey directly.

```typescript
import * as sarvam from '@livekit/agents-plugin-sarvam';

const stt = new sarvam.STT({
  model: 'saaras:v2.5',
  prompt: 'Technical interview discussion',
});
```

--------------------------------

### Configure Trugen API Key

Source: https://github.com/livekit/agents-js/blob/main/plugins/trugen/README.md

Sets the TRUGEN_API_KEY environment variable with your developer API key obtained from the Trugen dashboard. This key is required for authenticating with the Trugen service.

```bash
export TRUGEN_API_KEY=<your-trugen-api-key>
```

--------------------------------

### Initialize Sarvam AI STT (Non-Streaming REST)

Source: https://github.com/livekit/agents-js/blob/main/plugins/sarvam/README.md

Initializes the Speech-to-Text (STT) service using Sarvam AI's Saaras models in a non-streaming mode, preferring REST over WebSocket. This configuration will result in LiveKit using Voice Activity Detection (VAD) and non-streaming recognize() calls within an AgentSession.

```typescript
import * as sarvam from '@livekit/agents-plugin-sarvam';

const stt = new sarvam.STT({
  model: 'saaras:v3',
  streaming: false,
});
```

--------------------------------

### Initialize STT for Batch Transcription

Source: https://github.com/livekit/agents-js/blob/main/plugins/mistralai/README.md

Initialize the STT component for batch transcription using a specified Mistral AI model.

```typescript
// Batch transcription
const stt = new mistral.STT({ model: 'voxtral-mini-latest' });
```

--------------------------------

### STTOptions Interface

Source: https://github.com/livekit/agents-js/blob/main/plugins/soniox/etc/agents-plugin-soniox.api.md

Defines the configuration options for STT services, including API keys, base URLs, language identification, speaker diarization, and model settings.

```typescript
export interface STTOptions {
    // (undocumented)
    apiKey?: string;
    // (undocumented)
    baseUrl: string;
    // (undocumented)
    clientReferenceId?: string;
    // (undocumented)
    context?: ContextObject | string;
    // (undocumented)
    enableLanguageIdentification: boolean;
    // (undocumented)
    enableSpeakerDiarization: boolean;
    // (undocumented)
    languageHints?: string[];
    // (undocumented)
    languageHintsStrict: boolean;
    maxEndpointDelayMs: number;
    // (undocumented)
    model: string;
    // (undocumented)
    numChannels: number;
    // (undocumented)
    sampleRate: number;
    // (undocumented)
    translation?: TranslationConfig;
}
```

--------------------------------

### Set LiveAvatar API Key

Source: https://github.com/livekit/agents-js/blob/main/plugins/liveavatar/README.md

Set the LIVEAVATAR_API_KEY environment variable with your developer API key from the LiveAvatar dashboard.

```bash
export LIVEAVATAR_API_KEY=<your-liveavatar-api-key>
```

--------------------------------

### Initialize Sarvam AI TTS (Non-Streaming REST)

Source: https://github.com/livekit/agents-js/blob/main/plugins/sarvam/README.md

Initializes the Text-to-Speech (TTS) service using Sarvam AI's Bulbul models in a non-streaming mode, preferring REST over WebSocket. This configuration will result in LiveKit using sentence chunking and non-streaming synthesize() calls within an AgentSession.

```typescript
import * as sarvam from '@livekit/agents-plugin-sarvam';

const tts = new sarvam.TTS({
  model: 'bulbul:v2',
  streaming: false,
});
```

--------------------------------

### LLM Class

Source: https://github.com/livekit/agents-js/blob/main/plugins/perplexity/etc/agents-plugin-perplexity.api.md

Represents a Language Model client for Perplexity. It extends the base LLM class from `@livekit/agents-plugin-openai` and provides Perplexity-specific configurations.

```APIDOC
## LLM Class

### Description
Represents a Language Model client for Perplexity. It extends the base LLM class from `@livekit/agents-plugin-openai` and provides Perplexity-specific configurations.

### Methods

- **constructor(opts?: Partial<LLMOptions>)**: Initializes a new instance of the LLM class.
- **label()**: Returns the label for this LLM instance.
- **provider**: Returns the provider name for this LLM instance.

### Properties

- **provider**: string - The name of the LLM provider.
```

--------------------------------

### ContextObject Interface

Source: https://github.com/livekit/agents-js/blob/main/plugins/soniox/etc/agents-plugin-soniox.api.md

Provides structured context for speech recognition, including general terms, specific terms, and translation configurations.

```APIDOC
## Interface ContextObject

### Description
Represents contextual information to enhance speech recognition accuracy.

### Properties
- `general` (ContextGeneralItem[]) - An array of general key-value pairs for context.
- `terms` (string[]) - An array of specific terms to recognize.
- `text` (string) - A block of text for context.
- `translationTerms` (ContextTranslationTerm[]) - Configuration for translation terms.
```

--------------------------------

### APIConnectOptions Type

Source: https://github.com/livekit/agents-js/blob/main/plugins/anam/etc/agents-plugin-anam.api.md

Configuration options for API connections.

```APIDOC
## Type: APIConnectOptions

### Description
Defines options for configuring API connections, such as retry mechanisms and timeouts.

### Properties

- **maxRetry** (number) - Optional - The maximum number of retries for a connection attempt.
- **retryInterval** (number) - Optional - The interval in milliseconds between retry attempts.
- **timeout** (number) - Optional - The timeout in milliseconds for connection attempts.
```

--------------------------------

### Initialize Avatar Session

Source: https://github.com/livekit/agents-js/blob/main/plugins/runway/README.md

Initialize an AvatarSession with a preset ID. The API key defaults to the RUNWAYML_API_SECRET environment variable.

```typescript
import { AvatarSession } from '@livekit/agents-plugin-runway';

const avatarSession = new AvatarSession({
  presetId: 'your-avatar-preset-id',
  // apiKey defaults to RUNWAYML_API_SECRET env var
});

await avatarSession.start(agentSession, room);
```