Mscc.GenerativeAI.Microsoft

MSCC GenerativeAI for .NET integrating with Microsoft.Extensions.AI and Microsoft Semantic Kernel.

Integrate Gemini API into your .NET applications using Microsoft.Extensions.AI. This package implements the IChatClient and IEmbeddingGenerator interfaces offering a unified experience for .NET applications and libraries.

Benefits of Microsoft.Extensions.AI

Microsoft.Extensions.AI offers a unified API abstraction for AI services, similar to our successful logging and dependency injection (DI) abstractions. Our goal is to provide standard implementations for caching, telemetry, tool calling, and other common tasks that work with any provider.

Read more Introducing Microsoft.Extensions.AI Preview – Unified AI Building Blocks for .NET

Getting started

To get started, you can create a console application and install the Mscc.GenerativeAI.Microsoft package. Have your API key from Google AI Studio and you're ready to go.

Chat

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

// Chat with Gemini API.
var apiKey = Environment.GetEnvironmentVariable("GOOGLE_API_KEY");
var model = "gemini-1.5-pro-latest";
var prompt = "What is AI?";

IChatClient chatClient = new GeminiChatClient(apiKey, model);

var response = await chatClient.CompleteAsync(prompt);
Console.WriteLine(response.Message);

response = await chatClient.CompleteAsync( 
    "Translate the following text into Pig Latin: I love .NET and AI"); 
Console.WriteLine(response.Message);

Embeddings

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

// Create embeddings using the appropriate model.
var apiKey = Environment.GetEnvironmentVariable("GOOGLE_API_KEY");
var model = "text-embedding-004";
var prompt = "What is AI?";

IEmbeddingGenerator<string,Embedding<float>> generator = 
    new GeminiEmbeddingGenerator(apiKey, model);

var embeddings = await generator.GenerateAsync([prompt]);
Console.WriteLine(string.Join(", ", embeddings[0].Vector.ToArray()));

More examples

The folders samples and tests contain more examples.

Troubleshooting

tba

Feedback

You can create issues at the https://github.com/mscraftsman/generative-ai repository.