Google_GenerativeAI.Microsoft 4.0.1

Google_GenerativeAI.Microsoft

Google_GenerativeAI.Microsoft implements the Microsoft.Extensions.AI abstractions for Google Gemini, using the Google_GenerativeAI SDK.

Each type accepts either an API key (Gemini API) or an IPlatformAdapter, for example a VertextPlatformAdapter for Vertex AI.

Installation

dotnet add package Google_GenerativeAI.Microsoft

Microsoft.Extensions.AI.Abstractions 10.x is used on .NET 8 and .NET 9, and 11.x on .NET 10+, .NET Standard 2.0 and .NET Framework 4.6.2.

Chat

using GenerativeAI;
using GenerativeAI.Microsoft;
using Microsoft.Extensions.AI;

IChatClient chatClient = new GenerativeAIChatClient(
    System.Environment.GetEnvironmentVariable("GOOGLE_API_KEY")!, GoogleAIModels.Gemini38Flash);

// Single response
ChatResponse response = await chatClient.GetResponseAsync("What is the capital of France?");
Console.WriteLine(response.Text);

// Multi-turn conversation
var history = new List<ChatMessage>
{
    new(ChatRole.System, "You are a concise assistant."),
    new(ChatRole.User, "Suggest a name for a coffee shop.")
};
response = await chatClient.GetResponseAsync(history);
history.AddMessages(response);

// Streaming
await foreach (var update in chatClient.GetStreamingResponseAsync("Tell me a short story."))
    Console.Write(update.Text);

Function calling

Functions passed in ChatOptions.Tools are called automatically and their results are sent back to the model. Pass autoCallFunction: false to the constructor if you prefer ChatClientBuilder.UseFunctionInvocation().

[Description("Gets the current weather for a city")]
static string GetWeather(string city) => $"It is 22°C and sunny in {city}.";

var options = new ChatOptions { Tools = [AIFunctionFactory.Create(GetWeather)] };
var answer = await chatClient.GetResponseAsync("What's the weather in Paris?", options);

Dependency injection

builder.Services.AddSingleton<IChatClient>(_ =>
    new GenerativeAIChatClient(builder.Configuration["Gemini:ApiKey"]!, GoogleAIModels.Gemini38Flash));

Embeddings

IEmbeddingGenerator<string, Embedding<float>> embeddings =
    new GenerativeAIEmbeddingGenerator(System.Environment.GetEnvironmentVariable("GOOGLE_API_KEY")!);

var vectors = await embeddings.GenerateAsync(["first document", "second document"]);

Images

#pragma warning disable MEAI001
IImageGenerator images = new GenerativeAIImageGenerator(System.Environment.GetEnvironmentVariable("GOOGLE_API_KEY")!);

var generated = await images.GenerateAsync(new ImageGenerationRequest { Prompt = "A watercolor lighthouse at sunrise" });
var image = generated.Contents.OfType<DataContent>().First();
await File.WriteAllBytesAsync("lighthouse.png", image.Data.ToArray());

// Gemini image models can also edit and combine existing images.
var edited = await images.GenerateAsync(new ImageGenerationRequest
{
    Prompt = "Turn this into a night scene.",
    OriginalImages = [image]
});

Documentation

• Microsoft.Extensions.AI integration guide
• Image Generation
• Embeddings

License

MIT License. See LICENSE.