前言
本章讲一下在Semantic Kernel中使用DependencyInject(依赖注入),在之前的章节我们都是通过手动创建Kernel对象来完成框架的初始化工作,今天我们用依赖注入的方式来实现。
实战
定义Native Plugins
我们用官网的LightPlugins插件来演示依赖注入在SK中的使用
public class LightPlugin { public bool IsOn { get; set; } = false; #pragma warning disable CA1024 // Use properties where appropriate [KernelFunction] [Description("Gets the state of the light.")] public string GetState() => IsOn ? "on" : "off"; #pragma warning restore CA1024 // Use properties where appropriate [KernelFunction] [Description("Changes the state of the light.'")] public string ChangeState(bool newState) { this.IsOn = newState; var state = GetState(); // Print the state to the console Console.WriteLine($"[Light is now {state}]"); return state; } }
这个插件有两个方法一个是获取当前灯的状态,第二个是改变灯的状态
创建kernel对象
在之前我们的演示中都是通过Kernel对象提供的CreateBuilder方法来创建Kernel对象。
var kernel = Kernel.CreateBuilder(). AddOpenAIChatCompletion(modelId: config.ModelId, apiKey: config.ApiKey) .Build();
在api项目的开发中,依靠依赖注入的方式更容易管理依赖项,以及对象的复用
依赖注入注入Kernel依赖
有两种方式可以用依赖注入创建Kernel对象,第一种是借助于KernelServiceCollectionExtensions累提供的AddKernel扩展方法,第二种就是自己Kernel kernel = new(services.BuildServiceProvider());或者services.AddTransient<Kernel>();
AddKernel源码
/// </returns> /// <remarks> /// Both services are registered as transient, as both objects are mutable. /// </remarks> public static IKernelBuilder AddKernel(this IServiceCollection services) { Verify.NotNull(services); // Register a KernelPluginCollection to be populated with any IKernelPlugins that have been // directly registered in DI. It's transient because the Kernel will store the collection // directly, and we don't want two Kernel instances to hold on to the same mutable collection. services.AddTransient<KernelPluginCollection>(); // Register the Kernel as transient. It's mutable and expected to be mutated by consumers, // such as via adding event handlers, adding plugins, storing state in its Data collection, etc. services.AddTransient<Kernel>(); // Create and return a builder that can be used for adding services and plugins // to the IServiceCollection. return new KernelBuilder(services); }
通过源码我们可以看出来,这两种方式基本上没区别,第二种AddKernel实际上是简化了我们第二种的步骤,我们就用第一种举例演示
//依赖注入 { IServiceCollection services = new ServiceCollection(); //会话服务注册到IOC容器 services.AddKernel().AddOpenAIChatCompletion(modelId: config.ModelId, apiKey: config.ApiKey, httpClient: client); services.AddSingleton<KernelPlugin>(sp => KernelPluginFactory.CreateFromType<LightPlugin>(serviceProvider: sp)); var kernel = services.BuildServiceProvider().GetRequiredService<Kernel>();
这就是在依赖注入中注册Kernel对象和插件的步骤,依赖项都会被注册到IServiceCollection中
Semantic Kernel使用的服务和插件通常作为Singleton单例注册到依赖注入容器中,以便它们可以在各种Kernel之间重用/共享。Kernel通常注册为Transient瞬态,以便每个实例不受处理其他任务的Kernel所做更改的影响。
在项目中使用时,我们可以通过在构造函数中获取Kernel对象的实例,用Kernel对象来获取服务实例
var chatCompletionService = kernel.GetRequiredService<IChatCompletionService>();
IChatCompletionService实例也可以通过IServiceProvider来获取,您可以灵活地使用更适合您要求的方法。
实战
我们用依赖注入跑一下LightPlugin插件
// Create chat history var history = new ChatHistory(); // Get chat completion service var chatCompletionService = kernel.GetRequiredService<IChatCompletionService>(); // Start the conversation Console.Write("User > "); string? userInput; while ((userInput = Console.ReadLine()) is not null) { // Add user input history.AddUserMessage(userInput); // Enable auto function calling OpenAIPromptExecutionSettings openAIPromptExecutionSettings = new() { ToolCallBehavior = ToolCallBehavior.AutoInvokeKernelFunctions }; // Get the response from the AI var result = await chatCompletionService.GetChatMessageContentAsync( history, executionSettings: openAIPromptExecutionSettings, kernel: kernel); // Print the results Console.WriteLine("Assistant > " + result); // Add the message from the agent to the chat history history.AddMessage(result.Role, result.Content ?? string.Empty); // Get user input again Console.Write("User > "); }
输出:
User > 当前灯光的状态 Assistant > 当前灯光的状态是关闭的。 User > 帮我开个灯 [Light is now on] Assistant > 已经成功为您点亮了灯。
最后
本文Demo用的大模型月之暗面的moonshot-v1-8k
"Endpoint": "https://api.moonshot.cn", "ModelId": "moonshot-v1-8k",
原则上任何支持OpenAI function calling 格式的都可以使用。
通过本章的学习,我们深入了解了在Semantic Kernel中利用依赖注入的方式来管理Kernel对象和插件,使得项目开发更加灵活和高效。
参考文献
Using Semantic Kernel with Dependency Injection
示例代码