LiveKit Agents Google Plugin

This plugin provides Google Gemini LLM support for LiveKit Node Agents using the new unified @google/genai SDK.

Installation

npm install @livekit/agents-plugin-google
Copy

Usage

Basic Setup

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
});

// 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
});
Copy

To use the Gemini realtime model or TTS (TTS is Beta)

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(),
Copy

Live API model support

LiveKit supports the Gemini Live API through both the Gemini Developer API and Vertex AI. Model availability and behavior differ between APIs. Some models, such as gemini-3.8-live, are available on both.

For model/API pairs that appear incompatible, the plugin logs a warning and lets the API decide whether the model is available.

On the Gemini Developer API, gemini-3.8-live does not support thinkingConfig.thinkingLevel. Use gemini-3.8-live-extended-thinking for configurable thinking.

References:

• Gemini API Models
• Gemini Live API thinking
• Vertex Live API

Environment Variables

• GOOGLE_API_KEY or GOOGLE_GENAI_API_KEY: Your Google AI Studio API key
• GOOGLE_GENAI_USE_VERTEXAI: Set to true or 1 to enable Vertex AI
• GOOGLE_CLOUD_PROJECT: Your Google Cloud project ID (for Vertex AI)
• GOOGLE_CLOUD_LOCATION: Your preferred location (default: us-central1)

Supported Models

• gemini-1.5-pro - Most capable model
• gemini-1.5-flash - Fast and efficient
• gemini-1.5-flash-8b - Ultra-fast lightweight model
• gemini-2.0-flash-exp - Latest experimental model
• And more (see models.ts for full list)

Thinking Configuration

• thinkingConfig is model-specific.
• Gemini 3.1 live models use thinkingLevel such as minimal, low, medium, and high.
• Gemini 2.5 live models use thinkingBudget.

Advanced Configuration

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

Authentication

Google AI Studio

Set your API key via environment variable or constructor option:

export GOOGLE_API_KEY=your-api-key
Copy

Vertex AI

For Vertex AI, ensure you have:

1. Google Cloud CLI installed and authenticated
2. Vertex AI API enabled in your project
3. Proper authentication configured (Application Default Credentials)

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

License

Apache 2.0