{ "cells": [ { "cell_type": "markdown", "id": "261d91b6", "metadata": {}, "source": [ "快速入门指南\n", "===========" ] }, { "cell_type": "markdown", "id": "034c3c59", "metadata": {}, "source": [ "只需执行几个步骤，即可在 Python、.NET 或 Java 中使用语义内核生成第一个 AI 代理。以C#为例： \n", "\n", "+ 安装所需的包\n", "+ 使用 AI 创建来回对话\n", "+ 使 AI 代理能够运行代码\n", "+ 观看 AI 动态创建计划" ] }, { "cell_type": "markdown", "id": "70d6d69f", "metadata": {}, "source": [ "## 初始化基础设施" ] }, { "cell_type": "code", "execution_count": 37, "id": "becd4e47", "metadata": { "language_info": { "name": "polyglot-notebook" }, "polyglot_notebook": { "kernelName": "csharp" } }, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "初始化完毕！\r\n" ] } ], "source": [ "#!import ./ini.ipynb" ] }, { "cell_type": "markdown", "id": "f942ceb2", "metadata": {}, "source": [ "## 安装 SDK" ] }, { "cell_type": "markdown", "id": "145ba5d7", "metadata": {}, "source": [ "语义内核有多个 NuGet 包可用。但是，对于大多数方案，通常只需要 Microsoft.SemanticKernel。\n", "\n", "可以使用以下命令安装它：\n", "\n", "```Bash\n", " dotnet add package Microsoft.SemanticKernel\n", "```\n", "\n", "有关 Nuget 包的完整列表，请参阅 [支持的语言文章](https://learn.microsoft.com/zh-cn/semantic-kernel/get-started/supported-languages?pivots=programming-language-csharp)。" ] }, { "cell_type": "markdown", "id": "6e796f0a", "metadata": {}, "source": [ "## 快速开始使用笔记本" ] }, { "cell_type": "markdown", "id": "d3ecfe91", "metadata": {}, "source": [ "C# 开发人员，可以快速开始使用笔记本。这些笔记本提供了有关如何使用语义内核生成 AI 代理的分步指南。\n", "![快速开始](./Assets/Images/getting_started_notebooks.png)\n", "\n", "要开始，请按照以下步骤操作：\n", "\n", "1. 克隆[语义内核存储库](https://github.com/microsoft/semantic-kernel)\n", "2. 在 Visual Studio Code 中打开存储库\n", "3. 导航到 _/dotnet/notebooks\n", "4. 打开 00-getting-started.ipynb 以开始设置环境并创建第一个 AI 代理！\n", "\n" ] }, { "cell_type": "markdown", "id": "7e1ffa9a", "metadata": {}, "source": [ "## 编写第一个控制台应用" ] }, { "cell_type": "markdown", "id": "eeba44ee", "metadata": {}, "source": [ "1. 使用以下命令创建新的 .NET 控制台项目：\n", "\n", " ```bash\n", " dotnet new console\n", " ```" ] }, { "cell_type": "markdown", "id": "765fd394", "metadata": {}, "source": [ "2. 安装以下 .NET 依赖项：\n", "\n", " ```baSH\n", " dotnet add package Microsoft.SemanticKernel\n", " dotnet add package Microsoft.Extensions.Logging\n", " dotnet add package Microsoft.Extensions.Logging.Console\n", " ```" ] }, { "cell_type": "markdown", "id": "8c9b17d0", "metadata": {}, "source": [ "3. 将 Program.cs 文件的内容替换为以下代码：\n", "\n", "```csharp\n", " // Import packages\n", " using Microsoft.Extensions.DependencyInjection;\n", " using Microsoft.Extensions.Logging;\n", " using Microsoft.SemanticKernel;\n", " using Microsoft.SemanticKernel.ChatCompletion;\n", " using Microsoft.SemanticKernel.Connectors.OpenAI;\n", "\n", " // Populate values from your OpenAI deployment\n", " var modelId = \"\";\n", " var endpoint = \"\";\n", " var apiKey = \"\";\n", "\n", " // Create a kernel with Azure OpenAI chat completion\n", " var builder = Kernel.CreateBuilder().AddAzureOpenAIChatCompletion(modelId, endpoint, apiKey);\n", "\n", " // Add enterprise components\n", " builder.Services.AddLogging(services => services.AddConsole().SetMinimumLevel(LogLevel.Trace));\n", "\n", " // Build the kernel\n", " Kernel kernel = builder.Build();\n", " var chatCompletionService = kernel.GetRequiredService();\n", "\n", " // Add a plugin (the LightsPlugin class is defined below)\n", " kernel.Plugins.AddFromType(\"Lights\");\n", "\n", " // Enable planning\n", " OpenAIPromptExecutionSettings openAIPromptExecutionSettings = new() \n", " {\n", " FunctionChoiceBehavior = FunctionChoiceBehavior.Auto()\n", " };\n", "\n", " // Create a history store the conversation\n", " var history = new ChatHistory();\n", "\n", " // Initiate a back-and-forth chat\n", " string? userInput;\n", " do {\n", " // Collect user input\n", " Console.Write(\"User > \");\n", " userInput = Console.ReadLine();\n", "\n", " // Add user input\n", " history.AddUserMessage(userInput);\n", "\n", " // Get the response from the AI\n", " var result = await chatCompletionService.GetChatMessageContentAsync(\n", " history,\n", " executionSettings: openAIPromptExecutionSettings,\n", " kernel: kernel);\n", "\n", " // Print the results\n", " Console.WriteLine(\"Assistant > \" + result);\n", "\n", " // Add the message from the agent to the chat history\n", " history.AddMessage(result.Role, result.Content ?? string.Empty);\n", " } while (userInput is not null);\n", "\n", "```\n", "\n", "以下来回聊天应类似于你在控制台中看到的内容。下面添加了函数调用，以演示 AI 如何在后台利用插件。\n", "\n", "| 角色 | 消息 |\n", "| ------------------ - | ----------- |\n", "| 🔵 用户 | 请切换灯 |\n", "| 🔴 助手（函数调用） | LightsPlugin.GetState() |\n", "| 🟢 工具 | off |\n", "| 🔴 助手（函数调用） | LightsPlugin.ChangeState(true) |\n", "| 🟢 工具 | on |\n", "| 🔴 助手 | 灯现在打开 |\n", "\n", "如果有兴趣了解有关上述代码的详细信息，我们将在下一部分中将其分解。" ] }, { "cell_type": "markdown", "id": "b7e846a6", "metadata": {}, "source": [ "## 了解代码" ] }, { "cell_type": "markdown", "id": "31f3881d", "metadata": {}, "source": [ "为了更轻松地开始使用语义内核构建企业应用，我们创建了一个分步指导你完成创建内核的过程，并使用它与 AI 服务交互。\n", "\n", "![过程](./Assets/Images/dotnetmap.png)\n", "\n", "在以下部分中，我们将通过演练步骤 1、2、3、4、6、9 和 10 来解包上述示例。构建由 AI 服务提供支持且可以运行代码的简单代理所需的一切。\n", "+ 导入包\n", "+ 添加 AI 服务\n", "+ 企业组件\n", "+ 生成内核\n", "+ 添加内存（已跳过）\n", "+ 添加插件\n", "+ 创建内核参数（已跳过）\n", "+ 创建提示（跳过）\n", "+ 规划\n", "+ Invoke" ] }, { "cell_type": "markdown", "id": "d9b47cb5", "metadata": {}, "source": [ "### １导入包" ] }, { "cell_type": "markdown", "id": "382cfdf5", "metadata": {}, "source": [ "### ２添加 AI 服务" ] } ], "metadata": { "kernelspec": { "display_name": ".NET (C#)", "language": "C#", "name": ".net-csharp" }, "language_info": { "name": "polyglot-notebook" }, "polyglot_notebook": { "kernelInfo": { "defaultKernelName": "csharp", "items": [ { "aliases": [], "languageName": "csharp", "name": "csharp" } ] } } }, "nbformat": 4, "nbformat_minor": 5 }