引言：AI時代下的聊天機器人開發範式演進

在生成式AI技術爆發的當下，基於大語言模型（LLM）的聊天機器人開發已形成標準化技術鏈路。本文將結合DeepSeek API與微軟Semantic Kernel框架，以C#語言實戰演示從基礎API調用到高級框架集成的完整開發流程。

環境準備與基礎配置

• .NET 9 SDK
• Visual Studio 2022或VSCode
• DeepSeek API密鑰 官網申請

DeepSeek API基礎調用

DeepSeek API的Endpoint地址為：https://api.deepseek.com/chat/completions，相關文檔可查看官方文檔

1. 單輪對話實現
   代碼示例

public async Task<ResponseBody> GetChatMessageContentsAsync(CancellationToken cancellationToken = new CancellationToken())
{
    var client = new HttpClient();
    var request = new HttpRequestMessage(HttpMethod.Post, _builder.Endpoint);
    request.Headers.Add("Accept", "application/json");
    request.Headers.Add("Authorization", $"Bearer {_builder.ApiKey}");

    _body.Stream = false;
    var content = new StringContent(_body.SerializeObject(), null, "application/json");
    request.Content = content;
    var response = await client.SendAsync(request, cancellationToken);
    var responseBody = await response.Content.ReadAsStringAsync(cancellationToken);
    return JsonConvert.DeserializeObject<ResponseBody>(responseBody) ?? new ResponseBody();
}

2. 流式響應處理
   代碼示例

public async IAsyncEnumerable<ResponseBody> GetStreamingChatMessageContentsAsync([EnumeratorCancellation] CancellationToken cancellationToken = new CancellationToken())
{
    var client = new HttpClient();
    var request = new HttpRequestMessage(HttpMethod.Post, _builder.Endpoint);
    request.Headers.Add("Accept", "application/json");
    request.Headers.Add("Authorization", $"Bearer {_builder.ApiKey}");
    
    _body.Stream = true;
    var content = new StringContent(_body.SerializeObject(), null, "application/json");
    request.Content = content;
    var response = await client.SendAsync(request, cancellationToken);
    var stream = await response.Content.ReadAsStreamAsync(cancellationToken);
    var reader = new StreamReader(stream);
    while (!reader.EndOfStream)
    {
        var line = await reader.ReadLineAsync(cancellationToken);
        if (string.IsNullOrEmpty(line) || line.StartsWith(":")) continue;
        if (line.StartsWith("data: "))
        {
            var jsonData = line["data: ".Length ..];
            if (jsonData == "[DONE]") break;
            yield return JsonConvert.DeserializeObject<ResponseBody>(jsonData) ?? new ResponseBody();
        }
    }
}

Semantic Kernel框架集成

Semantic Kernel是一種輕型開源開發工具包，可用於輕鬆生成 AI 代理並將最新的 AI 模型集成到 C#、Python 或 Java 代碼庫中。 它充當一個高效的中間件，可實現企業級解決方案的快速交付。
DeepSeek API與Semantic Kernel框架集成，可快速實現基於大語言模型的聊天機器人開發。由於DeepSeek API與OpenAI API的兼容性，因此DeepSeek API與Semantic Kernel框架的集成非常簡單。只需使用OpenAI的連接器，即可快速實現DeepSeek API與Semantic Kernel框架的集成。

1. NuGet包安裝

dotnet add package Microsoft.SemanticKernel

2. Semantic Kernel初始化

var openAIClientCredential = new ApiKeyCredential(apiKey);
var openAIClientOption = new OpenAIClientOptions
{
    Endpoint = new Uri("https://api.deepseek.com"),

};
var builder = Kernel.CreateBuilder()
    .AddOpenAIChatCompletion(modelId, new OpenAIClient(openAIClientCredential, openAIClientOption));

var kernel = builder.Build();

3. 聊天機器人開發

var chatCompletionService = kernel.GetRequiredService<IChatCompletionService>();

Console.WriteLine("😀User >> "+ chatHistory.Last().Content);
var result = chatCompletionService.GetStreamingChatMessageContentsAsync(
    chatHistory
);
Console.Write("👨Assistant >> ");
await foreach (var item in result)
{
    Thread.Sleep(200);
    Console.Write(item.Content);
}

代碼示例