C#接入AI操作步骤详解(deepseek接入)

在C#中接入AI有多种方式，下面我将为你详细介绍几种主流的方法和操作步骤：

1. 使用Azure AI服务（推荐新手）

准备工作

csharp

// 安装NuGet包
Install-Package Azure.AI.TextAnalytics
Install-Package Azure.AI.OpenAI

文本分析示例

csharp

using Azure;
using Azure.AI.TextAnalytics;

class Program
{
static async Task Main(string[] args)
{
// Azure认知服务配置
string endpoint = "你的终结点";
string apiKey = "你的API密钥";

var client = new TextAnalyticsClient(new Uri(endpoint), new AzureKeyCredential(apiKey));

// 情感分析
string document = "这个产品非常好用，我非常喜欢！";
DocumentSentiment documentSentiment = client.AnalyzeSentiment(document);

Console.WriteLine($"情感: {documentSentiment.Sentiment}");
Console.WriteLine($"正面评分: {documentSentiment.ConfidenceScores.Positive:P2}");
}
}

2. 使用OpenAI API

安装OpenAI包

csharp

Install-Package OpenAI

调用GPT模型示例

csharp

using OpenAI_API;
using OpenAI_API.Chat;

class OpenAIService
{
private readonly OpenAIAPI _api;

public OpenAIService(string apiKey)
{
_api = new OpenAIAPI(apiKey);
}

public async Task<string> GetChatResponse(string prompt)
{
var chat = _api.Chat.CreateConversation();
chat.AppendUserInput(prompt);

string response = await chat.GetResponseFromChatbotAsync();
return response;
}
}

// 使用示例
class Program
{
static async Task Main(string[] args)
{
var openAIService = new OpenAIService("你的OpenAI-API密钥");
string response = await openAIService.GetChatResponse("请解释一下人工智能");
Console.WriteLine(response);
}
}

3. 使用ML.NET进行本地AI开发

安装ML.NET

csharp

Install-Package Microsoft.ML

情感分析模型示例

csharp

using Microsoft.ML;
using Microsoft.ML.Data;

// 定义数据模型
public class SentimentData
{
[LoadColumn(0)] public string SentimentText;
[LoadColumn(1), ColumnName("Label")] public bool Sentiment;
}

public class SentimentPrediction
{
[ColumnName("PredictedLabel")] public bool Prediction { get; set; }
public float Probability { get; set; }
public float Score { get; set; }
}

class MLNetDemo
{
static void Main()
{
var mlContext = new MLContext();

// 加载数据
IDataView dataView = mlContext.Data.LoadFromTextFile<SentimentData>(
"sentiment-data.csv",
hasHeader: true,
separatorChar: ',');

// 数据预处理管道
var dataProcessPipeline = mlContext.Transforms.Text.FeaturizeText(
"Features",
nameof(SentimentData.SentimentText));

// 选择算法
var trainer = mlContext.BinaryClassification.Trainers.SdcaLogisticRegression();
var trainingPipeline = dataProcessPipeline.Append(trainer);

// 训练模型
var model = trainingPipeline.Fit(dataView);

// 预测
var predictionEngine = mlContext.Model.CreatePredictionEngine<SentimentData, SentimentPrediction>(model);

var sample = new SentimentData { SentimentText = "这是一个很棒的产品" };
var result = predictionEngine.Predict(sample);

Console.WriteLine($"预测结果: {(result.Prediction ? "正面" : "负面")}");
Console.WriteLine($"置信度: {result.Probability:P2}");
}
}

4. 集成TensorFlow.NET

安装TensorFlow.NET

csharp

Install-Package TensorFlow.NET
Install-Package SciSharp.TensorFlow.Redist

图像分类示例

csharp

using Tensorflow;
using static Tensorflow.Binding;

class TensorFlowDemo
{
public void Run()
{
// 加载预训练模型
var graph = new Graph();
var session = new Session(graph);

// 这里加载你的TensorFlow模型
// var model = graph.Import(); // 导入.pb文件

// 进行预测
// var results = session.Run(…);
}
}

5. 完整的AI服务封装示例

csharp

public interface IAIService
{
Task<string> AnalyzeTextAsync(string text);
Task<string> GenerateImageAsync(string prompt);
Task<object> ClassifyImageAsync(byte[] imageData);
}

public class AzureAIService : IAIService
{
private readonly TextAnalyticsClient _textClient;
private readonly OpenAIService _openAIService;

public AzureAIService(string textEndpoint, string textKey, string openAIKey)
{
_textClient = new TextAnalyticsClient(new Uri(textEndpoint), new AzureKeyCredential(textKey));
_openAIService = new OpenAIService(openAIKey);
}

public async Task<string> AnalyzeTextAsync(string text)
{
try
{
// 情感分析
var sentiment = await _textClient.AnalyzeSentimentAsync(text);

// 关键短语提取
var keyPhrases = await _textClient.ExtractKeyPhrasesAsync(text);

return $"情感: {sentiment.Value.Sentiment}, " +
$"关键短语: {string.Join(", ", keyPhrases.Value)}";
}
catch (Exception ex)
{
return $"分析失败: {ex.Message}";
}
}

public async Task<string> GenerateImageAsync(string prompt)
{
// 使用DALL-E生成图像
return await _openAIService.GenerateImage(prompt);
}

public Task<object> ClassifyImageAsync(byte[] imageData)
{
// 图像分类逻辑
throw new NotImplementedException();
}
}

6. 配置和最佳实践

appsettings.json配置

json

{
"AzureAI": {
"Endpoint": "https://your-resource.cognitiveservices.azure.com/",
"Key": "your-key"
},
"OpenAI": {
"ApiKey": "your-openai-key"
}
}

依赖注入配置

csharp

// Startup.cs或Program.cs
services.AddSingleton<IAIService>(provider =>
{
var config = provider.GetRequiredService<IConfiguration>();
return new AzureAIService(
config["AzureAI:Endpoint"],
config["AzureAI:Key"],
config["OpenAI:ApiKey"]
);
});

操作步骤总结

注意事项

• 妥善保管API密钥，不要硬编码在代码中

• 添加适当的错误处理和重试机制

• 考虑API调用频率限制

• 对于生产环境，使用配置管理工具管理密钥

• 监控API使用情况和成本

选择哪种方式取决于你的具体需求：云服务适合快速集成，本地ML.NET适合数据隐私要求高的场景，TensorFlow.NET适合已有TensorFlow模型的场景。

ML.NET 是完全免费和开源的。

下面为您详细解释一下它的免费模式以及可能涉及的潜在成本：

1. ML.NET 框架本身

• 许可证：它使用非常宽松的 MIT 许可证。这意味着您可以在个人项目、商业项目、开源或闭源项目中自由使用、修改和分发它，而无需支付任何授权费用。

• 开源：代码完全公开在 GitHub 上，由微软官方维护。

• 无运行时费用：一旦您的模型训练完成并集成到应用程序中，使用模型进行预测（推理）不会产生任何按调用次数计费的费用。这与调用云AI API（如OpenAI或Azure Cognitive Services）按调用量付费的模式有本质区别。

2. 潜在的间接成本

虽然框架免费，但在开发和使用过程中，可能会产生一些间接成本：

• 开发工具：

  • 免费：您可以使用完全免费的 Visual Studio Code 或 Visual Studio Community Edition 进行开发。

  • 付费：如果您选择使用付费版的 Visual Studio Professional 或 Enterprise，这会产生费用，但这与ML.NET本身无关，是开发工具的成本。

• 计算资源：

  • 训练模型：如果您的数据集非常庞大，训练一个复杂的模型（如图像分类、推荐系统）可能会消耗大量的CPU/GPU资源和时间。如果您在本地进行训练，成本就是电费和硬件折旧。如果您在云服务器（如Azure VM）上进行训练，则需要支付云服务器的费用。

  • 部署和运行：您的应用程序需要运行在某个服务器或计算机上。无论是本地服务器、Azure App Service、AWS EC2还是您用户的桌面电脑，这些环境都有其固有的硬件或托管成本。但模型的推理过程本身不产生额外许可费。

• 数据存储与准备：如果您的训练数据存储在云数据库或数据湖中，这些存储服务会产生费用。

ML.NET vs. 云AI服务的成本对比

为了更清晰，我们可以做一个对比：

ML.NET 是一个真正免费的框架，非常适合以下情况：

简单来说，ML.NET 让您用 前期开发成本和基础设施成本 换取了 零运行时许可费和绝对的数据控制权。对于预算有限、注重数据隐私的.NET开发者来说，它是一个极其强大的免费武器。

C# 接入 DeepSeek API 完整指南

下面详细介绍如何在 C# 中接入 DeepSeek API，包括控制台应用、Web API 和桌面应用的实现。

1. 准备工作

获取 API 密钥

安装必要的 NuGet 包

bash

Install-Package Newtonsoft.Json
Install-Package System.Net.Http

2. 基础实现 – 控制台应用

创建 API 请求模型

csharp

using Newtonsoft.Json;

public class DeepSeekRequest
{
[JsonProperty("model")]
public string Model { get; set; } = "deepseek-chat";

[JsonProperty("messages")]
public List<ChatMessage> Messages { get; set; } = new List<ChatMessage>();

[JsonProperty("max_tokens")]
public int MaxTokens { get; set; } = 2048;

[JsonProperty("temperature")]
public double Temperature { get; set; } = 0.7;

[JsonProperty("stream")]
public bool Stream { get; set; } = false;
}

public class ChatMessage
{
[JsonProperty("role")]
public string Role { get; set; } // "system", "user", "assistant"

[JsonProperty("content")]
public string Content { get; set; }
}

创建 API 响应模型

csharp

public class DeepSeekResponse
{
[JsonProperty("id")]
public string Id { get; set; }

[JsonProperty("object")]
public string Object { get; set; }

[JsonProperty("created")]
public long Created { get; set; }

[JsonProperty("model")]
public string Model { get; set; }

[JsonProperty("choices")]
public List<Choice> Choices { get; set; }

[JsonProperty("usage")]
public UsageInfo Usage { get; set; }
}

public class Choice
{
[JsonProperty("index")]
public int Index { get; set; }

[JsonProperty("message")]
public ChatMessage Message { get; set; }

[JsonProperty("finish_reason")]
public string FinishReason { get; set; }
}

public class UsageInfo
{
[JsonProperty("prompt_tokens")]
public int PromptTokens { get; set; }

[JsonProperty("completion_tokens")]
public int CompletionTokens { get; set; }

[JsonProperty("total_tokens")]
public int TotalTokens { get; set; }
}

实现 DeepSeek 服务类

csharp

using System;
using System.Collections.Generic;
using System.Net.Http;
using System.Text;
using System.Threading.Tasks;
using Newtonsoft.Json;

public class DeepSeekService
{
private readonly string _apiKey;
private readonly string _apiUrl = "https://api.deepseek.com/v1/chat/completions";
private readonly HttpClient _httpClient;

public DeepSeekService(string apiKey)
{
_apiKey = apiKey;
_httpClient = new HttpClient();
_httpClient.DefaultRequestHeaders.Add("Authorization", $"Bearer {_apiKey}");
}

public async Task<string> SendMessageAsync(string userMessage, string systemMessage = null)
{
var request = new DeepSeekRequest
{
Messages = new List<ChatMessage>()
};

// 添加系统消息（如果有）
if (!string.IsNullOrEmpty(systemMessage))
{
request.Messages.Add(new ChatMessage
{
Role = "system",
Content = systemMessage
});
}

// 添加用户消息
request.Messages.Add(new ChatMessage
{
Role = "user",
Content = userMessage
});

try
{
var jsonContent = JsonConvert.SerializeObject(request);
var content = new StringContent(jsonContent, Encoding.UTF8, "application/json");

var response = await _httpClient.PostAsync(_apiUrl, content);
response.EnsureSuccessStatusCode();

var responseContent = await response.Content.ReadAsStringAsync();
var deepSeekResponse = JsonConvert.DeserializeObject<DeepSeekResponse>(responseContent);

return deepSeekResponse.Choices[0].Message.Content;
}
catch (Exception ex)
{
return $"错误: {ex.Message}";
}
}

public async Task<string> SendMessageWithHistoryAsync(List<ChatMessage> messageHistory)
{
var request = new DeepSeekRequest
{
Messages = messageHistory
};

try
{
var jsonContent = JsonConvert.SerializeObject(request);
var content = new StringContent(jsonContent, Encoding.UTF8, "application/json");

var response = await _httpClient.PostAsync(_apiUrl, content);
response.EnsureSuccessStatusCode();

var responseContent = await response.Content.ReadAsStringAsync();
var deepSeekResponse = JsonConvert.DeserializeObject<DeepSeekResponse>(responseContent);

return deepSeekResponse.Choices[0].Message.Content;
}
catch (Exception ex)
{
return $"错误: {ex.Message}";
}
}

public void Dispose()
{
_httpClient?.Dispose();
}
}

控制台应用示例

csharp

class Program
{
static async Task Main(string[] args)
{
// 替换为你的实际 API 密钥
string apiKey = "你的DeepSeek-API密钥";

using var deepSeekService = new DeepSeekService(apiKey);

Console.WriteLine("DeepSeek AI 聊天助手 (输入 '退出' 结束对话)");
Console.WriteLine("===========================================");

var conversationHistory = new List<ChatMessage>();

while (true)
{
Console.Write("你: ");
var userInput = Console.ReadLine();

if (userInput?.ToLower() == "退出" || userInput?.ToLower() == "exit")
break;

if (string.IsNullOrWhiteSpace(userInput))
continue;

// 添加用户消息到历史
conversationHistory.Add(new ChatMessage
{
Role = "user",
Content = userInput
});

Console.Write("AI: ");
var response = await deepSeekService.SendMessageWithHistoryAsync(conversationHistory);
Console.WriteLine(response);

// 添加AI回复到历史
conversationHistory.Add(new ChatMessage
{
Role = "assistant",
Content = response
});

Console.WriteLine();
}
}
}

3. Web API 集成

创建 ASP.NET Core Web API 控制器

csharp

[ApiController]
[Route("api/[controller]")]
public class AIController : ControllerBase
{
private readonly DeepSeekService _deepSeekService;

public AIController(IConfiguration configuration)
{
var apiKey = configuration["DeepSeek:ApiKey"];
_deepSeekService = new DeepSeekService(apiKey);
}

[HttpPost("chat")]
public async Task<ActionResult<ChatResponse>> Chat([FromBody] ChatRequest request)
{
try
{
var response = await _deepSeekService.SendMessageAsync(request.Message, request.SystemMessage);

return Ok(new ChatResponse
{
Success = true,
Message = response,
Timestamp = DateTime.UtcNow
});
}
catch (Exception ex)
{
return BadRequest(new ChatResponse
{
Success = false,
Message = $"请求失败: {ex.Message}",
Timestamp = DateTime.UtcNow
});
}
}

[HttpPost("chat-with-history")]
public async Task<ActionResult<ChatResponse>> ChatWithHistory([FromBody] ChatHistoryRequest request)
{
try
{
var messages = request.Messages.Select(m => new ChatMessage
{
Role = m.Role,
Content = m.Content
}).ToList();

var response = await _deepSeekService.SendMessageWithHistoryAsync(messages);

return Ok(new ChatResponse
{
Success = true,
Message = response,
Timestamp = DateTime.UtcNow
});
}
catch (Exception ex)
{
return BadRequest(new ChatResponse
{
Success = false,
Message = $"请求失败: {ex.Message}",
Timestamp = DateTime.UtcNow
});
}
}
}

// DTO 类
public class ChatRequest
{
public string Message { get; set; }
public string SystemMessage { get; set; }
}

public class ChatHistoryRequest
{
public List<MessageDto> Messages { get; set; }
}

public class MessageDto
{
public string Role { get; set; }
public string Content { get; set; }
}

public class ChatResponse
{
public bool Success { get; set; }
public string Message { get; set; }
public DateTime Timestamp { get; set; }
}

appsettings.json 配置

json

{
"DeepSeek": {
"ApiKey": "你的DeepSeek-API密钥",
"ApiUrl": "https://api.deepseek.com/v1/chat/completions"
},
"Logging": {
"LogLevel": {
"Default": "Information",
"Microsoft.AspNetCore": "Warning"
}
},
"AllowedHosts": "*"
}

4. WPF 桌面应用集成

MainWindow.xaml

xml

<Window x:Class="DeepSeekChat.MainWindow"
xmlns="http://schemas.microsoft.com/winfx/2006/xaml/presentation"
xmlns:x="http://schemas.microsoft.com/winfx/2006/xaml"
Title="DeepSeek AI 聊天助手" Height="600" Width="800">
<Grid>
<Grid.RowDefinitions>
<RowDefinition Height="*"/>
<RowDefinition Height="Auto"/>
</Grid.RowDefinitions>

<ScrollViewer Grid.Row="0" VerticalScrollBarVisibility="Auto">
<ItemsControl x:Name="MessagesItemsControl">
<ItemsControl.ItemTemplate>
<DataTemplate>
<Border Margin="5" Padding="10"
Background="{Binding BackgroundColor}"
CornerRadius="10">
<TextBlock Text="{Binding Content}"
TextWrapping="Wrap"
Foreground="{Binding TextColor}"/>
</Border>
</DataTemplate>
</ItemsControl.ItemTemplate>
</ItemsControl>
</ScrollViewer>

<Grid Grid.Row="1" Margin="10">
<Grid.ColumnDefinitions>
<ColumnDefinition Width="*"/>
<ColumnDefinition Width="Auto"/>
</Grid.ColumnDefinitions>

<TextBox x:Name="MessageTextBox"
Grid.Column="0"
Height="60"
TextWrapping="Wrap"
VerticalScrollBarVisibility="Auto"
KeyDown="MessageTextBox_KeyDown"/>

<Button x:Name="SendButton"
Grid.Column="1"
Content="发送"
Width="60"
Height="60"
Margin="10,0,0,0"
Click="SendButton_Click"/>
</Grid>
</Grid>
</Window>

MainWindow.xaml.cs

csharp

using System;
using System.Collections.ObjectModel;
using System.Windows;
using System.Windows.Input;
using System.Windows.Media;

namespace DeepSeekChat
{
public partial class MainWindow : Window
{
private readonly DeepSeekService _deepSeekService;
private readonly ObservableCollection<MessageDisplay> _messages;

public MainWindow()
{
InitializeComponent();

string apiKey = "你的DeepSeek-API密钥"; // 应该从配置文件中读取
_deepSeekService = new DeepSeekService(apiKey);
_messages = new ObservableCollection<MessageDisplay>();
MessagesItemsControl.ItemsSource = _messages;
}

private async void SendButton_Click(object sender, RoutedEventArgs e)
{
await SendMessage();
}

private async void MessageTextBox_KeyDown(object sender, KeyEventArgs e)
{
if (e.Key == Key.Enter && Keyboard.Modifiers.HasFlag(ModifierKeys.Control))
{
await SendMessage();
}
}

private async Task SendMessage()
{
var userMessage = MessageTextBox.Text.Trim();
if (string.IsNullOrEmpty(userMessage))
return;

// 添加用户消息到界面
_messages.Add(new MessageDisplay
{
Content = $"你: {userMessage}",
BackgroundColor = Brushes.LightBlue,
TextColor = Brushes.Black
});

MessageTextBox.Clear();

try
{
// 显示加载状态
var loadingMessage = new MessageDisplay
{
Content = "AI: 思考中…",
BackgroundColor = Brushes.LightGray,
TextColor = Brushes.Black
};
_messages.Add(loadingMessage);

// 发送请求到 DeepSeek
var response = await _deepSeekService.SendMessageAsync(userMessage);

// 移除加载消息，添加实际回复
_messages.Remove(loadingMessage);
_messages.Add(new MessageDisplay
{
Content = $"AI: {response}",
BackgroundColor = Brushes.LightGreen,
TextColor = Brushes.Black
});

// 滚动到底部
ScrollToBottom();
}
catch (Exception ex)
{
_messages.Add(new MessageDisplay
{
Content = $"错误: {ex.Message}",
BackgroundColor = Brushes.LightCoral,
TextColor = Brushes.DarkRed
});
}
}

private void ScrollToBottom()
{
if (MessagesItemsControl.Items.Count > 0)
{
var border = (Border)VisualTreeHelper.GetChild(MessagesItemsControl, 0);
var scrollViewer = (ScrollViewer)VisualTreeHelper.GetChild(border, 0);
scrollViewer.ScrollToBottom();
}
}
}

public class MessageDisplay
{
public string Content { get; set; }
public Brush BackgroundColor { get; set; }
public Brush TextColor { get; set; }
}
}

5. 高级功能 – 流式响应

流式响应实现

csharp

public class DeepSeekStreamService
{
private readonly string _apiKey;
private readonly string _apiUrl = "https://api.deepseek.com/v1/chat/completions";
private readonly HttpClient _httpClient;

public DeepSeekStreamService(string apiKey)
{
_apiKey = apiKey;
_httpClient = new HttpClient();
_httpClient.DefaultRequestHeaders.Add("Authorization", $"Bearer {_apiKey}");
_httpClient.Timeout = TimeSpan.FromMinutes(5); // 设置较长的超时时间
}

public async IAsyncEnumerable<string> SendMessageStreamAsync(string userMessage, string systemMessage = null)
{
var request = new DeepSeekRequest
{
Messages = new List<ChatMessage>(),
Stream = true // 启用流式响应
};

if (!string.IsNullOrEmpty(systemMessage))
{
request.Messages.Add(new ChatMessage
{
Role = "system",
Content = systemMessage
});
}

request.Messages.Add(new ChatMessage
{
Role = "user",
Content = userMessage
});

var jsonContent = JsonConvert.SerializeObject(request);
var content = new StringContent(jsonContent, Encoding.UTF8, "application/json");

using var response = await _httpClient.PostAsync(_apiUrl, content, HttpCompletionOption.ResponseHeadersRead);
response.EnsureSuccessStatusCode();

using var stream = await response.Content.ReadAsStreamAsync();
using var reader = new StreamReader(stream);

while (!reader.EndOfStream)
{
var line = await reader.ReadLineAsync();
if (string.IsNullOrEmpty(line) || !line.StartsWith("data: "))
continue;

var data = line.Substring(6); // 移除 "data: " 前缀

if (data == "[DONE]")
yield break;

try
{
var streamResponse = JsonConvert.DeserializeObject<DeepSeekStreamResponse>(data);
var contentPiece = streamResponse?.Choices?[0]?.Delta?.Content;

if (!string.IsNullOrEmpty(contentPiece))
yield return contentPiece;
}
catch (JsonException)
{
// 忽略解析错误，继续处理下一行
}
}
}
}

// 流式响应模型
public class DeepSeekStreamResponse
{
[JsonProperty("id")]
public string Id { get; set; }

[JsonProperty("object")]
public string Object { get; set; }

[JsonProperty("created")]
public long Created { get; set; }

[JsonProperty("model")]
public string Model { get; set; }

[JsonProperty("choices")]
public List<StreamChoice> Choices { get; set; }
}

public class StreamChoice
{
[JsonProperty("index")]
public int Index { get; set; }

[JsonProperty("delta")]
public Delta Delta { get; set; }

[JsonProperty("finish_reason")]
public string FinishReason { get; set; }
}

public class Delta
{
[JsonProperty("role")]
public string Role { get; set; }

[JsonProperty("content")]
public string Content { get; set; }
}

6. 配置和最佳实践

依赖注入配置 (ASP.NET Core)

csharp

// Program.cs
builder.Services.AddSingleton<DeepSeekService>(provider =>
{
var configuration = provider.GetRequiredService<IConfiguration>();
var apiKey = configuration["DeepSeek:ApiKey"];
return new DeepSeekService(apiKey);
});

builder.Services.AddSingleton<DeepSeekStreamService>(provider =>
{
var configuration = provider.GetRequiredService<IConfiguration>();
var apiKey = configuration["DeepSeek:ApiKey"];
return new DeepSeekStreamService(apiKey);
});

错误处理和重试机制

csharp

public class ResilientDeepSeekService
{
private readonly DeepSeekService _deepSeekService;
private readonly int _maxRetries = 3;
private readonly TimeSpan _delay = TimeSpan.FromSeconds(1);

public ResilientDeepSeekService(string apiKey)
{
_deepSeekService = new DeepSeekService(apiKey);
}

public async Task<string> SendMessageWithRetryAsync(string userMessage, string systemMessage = null)
{
for (int i = 0; i < _maxRetries; i++)
{
try
{
return await _deepSeekService.SendMessageAsync(userMessage, systemMessage);
}
catch (HttpRequestException ex) when (i < _maxRetries – 1)
{
await Task.Delay(_delay * (i + 1));
}
}

throw new Exception("所有重试尝试都失败了");
}
}

使用注意事项

这个完整的指南应该能帮助你在各种 C# 应用中成功集成 DeepSeek API！