In this blog post, we’ll learn how to use a Microsoft.Extensions.AI in our .NET applications. In my earlier post I was using Microsoft.Extensions.AI to connect to my local SLM. We can use Microsoft.Extensions.AI so that we can write code using AI abstractions rather than a specific SDK. AI abstractions enable us to change the underlying AI model with minimal code changes. For this demo I am using Open AI API Key. We can use Azure Open AI or GitHub Models or Ollama. We can get the Open AI API Key from here. Once created, use dotnet user secret tool to store the key as secret or set the key as environment variable.

First I am creating a console application using dotnet new console command. Once the console app created, add the reference of Microsoft.Extensions.AI package. We also require Microsoft.Extensions.AI.OpenAI package. Since it is in preview, while adding the reference we need to use the --prerelease flag.

Once we added the two nuget packages, we can write the following code - which will connect and authenticate to the AI model.

var key = Environment.GetEnvironmentVariable("OPENAI_APIKEY");
if (string.IsNullOrEmpty(key))
{
    throw new InvalidOperationException("OpenAI API key is not set in the environment variables.");
}

var model = Environment.GetEnvironmentVariable(":OPENAI_MODEL") ?? "gpt-3.5-turbo";

var chatClient = new OpenAIClient(key).GetChatClient(model).AsIChatClient();

Next we will create a system prompt to provide the AI model with initial role context and instructions. For this demo I am setting a very simple system prompt.

var chatHistory = new List<ChatMessage>
{
    new(ChatRole.System, "You are a helpful assistant.")
};

Next we will be creating a loop that accepts an input prompt from the user, sends the prompt to the model, and prints the response completion. If the user input is empty, I am existing from the loop.

while (true)
{
    Console.Write("You: ");
    var userInput = Console.ReadLine();
    if (string.IsNullOrEmpty(userInput))
    {
        break;
    }

    chatHistory.Add(new ChatMessage(ChatRole.User, userInput));
    var response = string.Empty;
    Console.WriteLine("Assistant: ");
    await foreach (var item in chatClient.GetStreamingResponseAsync(chatHistory))
    {
        Console.Write(item.Text);
        response += item.Text;
    }
    chatHistory.Add(new ChatMessage(ChatRole.Assistant, response));
    Console.WriteLine();
}

Now we can execute the dotnet run command to run the application and interact with it.

This way we can create a conversational .NET console chat app using an OpenAI. The app uses the Microsoft.Extensions.AI library so you can write code using AI abstractions rather than a specific SDK. AI abstractions enable you to change the underlying AI model with minimal code changes. We can also use Semantic Kernel to accomplish these tasks.

Happy Programming