Gemini-livejs NPM

Gemini Live WebSocket Client

A TypeScript client for real-time communication with Google's Gemini API via WebSocket, supporting text, audio, function calls, and executable code responses.

Features

Feature	Support	Description
Text to Text	✅	Standard text conversation
Text to Audio	✅	Convert text input to audio response
Audio to Text	✅	Convert audio input to text response
Audio to Audio	✅	Process audio input and respond with audio
Function Calls	✅	Handle function calling capabilities
Executable Code	✅	Receive executable code responses
Real-time Streaming	✅	Stream responses in real-time
Custom Configuration	✅	Configurable model parameters
Safety Settings	✅	Customizable content safety filters
Connection Events	✅	Handlers for connection lifecycle

Limitations

Feature	Status	Description
Code Execution	❌	tools.codeExecution not implemented yet
Multi-part Response	❌	Only supports text and inlineData parts
Multiple Function Calls	❌	Limited to single function call output
Multiple Code Blocks	❌	Limited to single executable code output

Installation

npm install gemini-livejs

Basic Usage

import { GeminiLive } from "gemini-livejs";

// Initialize client
const client = new GeminiLive("YOUR_API_KEY", {
  generationConfig: {
    temperature: 0.7,
    maxOutputTokens: 2000,
  },
});

// Basic text conversation
async function chat() {
  const response = await client.send({ prompt: "Hello, how are you?" });
  console.log(response.text);
}

// Real-time audio streaming
client.realtime((response) => {
  if (response.type === "audio") {
    // Handle audio response
    console.log(response.audio.data);
  } else if (response.type === "text") {
    // Handle text response
    console.log(response.text);
  }
});

Note: send and realtime wont work if the client is not connected or handshaked.

Advanced Features

Function Calling

const client = new GeminiLive("YOUR_API_KEY", {
  tools: {
    functionDeclarations: [
      {
        name: "get_weather",
        description: "Get current weather",
        parameters: {
          type: ParameterType.OBJECT,
          properties: {
            location: {
              type: ParameterType.STRING,
              description: "City name",
            },
          },
        },
      },
    ],
  },
});

// Handle function calls
client.send({ prompt: "What's the weather in Tokyo?" }).then((response) => {
  if (response.type === "function" && response.functionCall) {
    console.log(response.functionCall);
  }
});

Event Handling

client
  .on_open(() => console.log("Connected!"))
  .on_handshake(() => console.log("Handshake complete"))
  .on_close((reason) => console.log("Closed:", reason));

Configuration Options

type GeminiConfig = {
  generationConfig?: {
    maxOutputTokens?: number;
    temperature?: number;
    topP?: number;
    topK?: number;
    responseType?: "TEXT" | "AUDIO";
    voiceName?: "Aoede" | "Charon" | "Fenrir" | "Kore" | "Puck";
  };
  systemInstruction?: string;
  tools?: ToolsConfig;
  safetySettings?: SafetySettings[];
};

Examples

For more examples, see the examples directory.

License

MIT

Contributing

Contributions welcome! Please feel free to submit a Pull Request.

google-gemini realtime-api gemini-realtime

node-wav ws

1.0.1

6 months ago

1.0.0

6 months ago