AI ServicesgeminiIndex

Google Gemini Overview

You can use the Gemini provider to integrate with Google’s Generative AI API for various AI capabilities such as text generation, chat completions, and more.

The Semantic Kernel Gemini Provider (@semantic-kernel/gemini) is a package that allows you to interact with Google’s Gemini AI models seamlessly within the Semantic Kernel framework.

Installation

npm install @semantic-kernel/gemini

Usage

import { GeminiChatClient } from '@semantic-kernel/gemini';
import { ChatMessage } from '@semantic-kernel/ai';
 
// Create a Gemini chat client
const client = new GeminiChatClient({
  apiKey: 'your-google-ai-api-key',
  modelId: 'gemini-1.5-flash',
});
 
// Send a chat message
const messages = [
  new ChatMessage({ role: 'user', content: 'Hello, how are you?' })
];
 
const response = await client.getResponse(messages);
console.log(response.message.text);

Streaming

// Get streaming response
const streamingResponse = client.getStreamingResponse(messages);
 
for await (const update of streamingResponse) {
  if (update.contents.length > 0) {
    console.log(update.contents[0]);
  }
}

Configuration Options

The GeminiChatClient constructor accepts the following options:

  • apiKey (string): Your Google AI API key
  • modelId (string): The Gemini model to use (e.g., ‘gemini-1.5-flash’, ‘gemini-1.5-pro’)
  • generativeModel (optional): Pre-configured GenerativeModel instance

Supported Models

  • gemini-1.5-flash - Fast and versatile performance
  • gemini-1.5-pro - Complex reasoning tasks
  • gemini-1.0-pro - Natural language tasks
  • gemini-1.0-pro-vision - Multimodal reasoning

Function Calling

The Gemini provider supports function calling (tools):

import { AIFunction, ChatOptions } from '@semantic-kernel/ai';
 
const weatherFunction = new AIFunction({
  name: 'get_weather',
  description: 'Get the current weather for a location',
  schema: {
    type: 'object',
    properties: {
      location: { type: 'string', description: 'The city name' }
    },
    required: ['location']
  }
});
 
const options = new ChatOptions();
options.tools = [weatherFunction];
 
const response = await client.getResponse(messages, options);
Chat CompletionSamples