Add AI to Your .NET Application

// 1. Install packages
// dotnet add package Microsoft.SemanticKernel
// dotnet add package Microsoft.SemanticKernel.Connectors.OpenAI

// 2. Program.cs — register Semantic Kernel
builder.Services.AddSingleton(sp =>
{
    var kernel = Kernel.CreateBuilder()
        .AddAzureOpenAIChatCompletion(
            deploymentName: "gpt-4o",
            endpoint:       builder.Configuration["AzureOpenAI:Endpoint"]!,
            apiKey:         builder.Configuration["AzureOpenAI:ApiKey"]!
        )
        .Build();
    return kernel;
});

// ChatService.cs — injectable service for Blazor components
public class ChatService(Kernel kernel)
{
    private readonly List<ChatMessageContent> _history = [];

    public async Task<string> SendMessageAsync(string userMessage)
    {
        _history.Add(new ChatMessageContent(AuthorRole.User, userMessage));

        var chat = kernel.GetRequiredService<IChatCompletionService>();
        var settings = new OpenAIPromptExecutionSettings
        {
            MaxTokens   = 500,
            Temperature = 0.7
        };

        var response = await chat.GetChatMessageContentAsync(
            chatHistory: new ChatHistory(_history),
            executionSettings: settings,
            kernel: kernel
        );

        _history.Add(response);
        return response.Content ?? string.Empty;
    }
}

@* ChatWidget.razor — drop this anywhere in your Blazor app *@
@inject ChatService Chat

<div class="chat-container">
    @foreach (var msg in _messages)
    {
        <div class="message @(msg.IsUser ? "user" : "assistant")">
            @msg.Content
        </div>
    }

    <div class="input-row">
        <input @bind="_input" @onkeydown="HandleKey"
               placeholder="Ask anything..." />
        <button @onclick="Send" disabled="@_loading">
            @(_loading ? "Thinking..." : "Send")
        </button>
    </div>
</div>

@code {
    private string _input  = "";
    private bool   _loading = false;
    private List<(string Content, bool IsUser)> _messages = [];

    private async Task Send()
    {
        if (string.IsNullOrWhiteSpace(_input)) return;
        _messages.Add((_input, true));
        var prompt = _input;
        _input   = "";
        _loading = true;
        var reply = await Chat.SendMessageAsync(prompt);
        _messages.Add((reply, false));
        _loading = false;
    }

    private async Task HandleKey(KeyboardEventArgs e)
    {
        if (e.Key == "Enter") await Send();
    }
}

// RAG (Retrieval-Augmented Generation) — add your own data
// 1. Store your documents as embeddings
var memory = kernel.GetRequiredService<ISemanticTextMemory>();

await memory.SaveInformationAsync(
    collection: "product-docs",
    text:       "Our return policy allows returns within 30 days...",
    id:         "policy-001"
);

// 2. Retrieve relevant context before answering
var results = memory.SearchAsync("product-docs", userQuestion, limit: 3);
var context = string.Join("\n", await results.Select(r => r.Metadata.Text).ToListAsync());

// 3. Inject context into the prompt
var prompt = $"""
    Answer based only on this context:
    {context}

    Question: {userQuestion}
""";