ai-agents-for-beginners

🔍 探索 Microsoft Agent Framework - 基础代理 (.NET)

📋 学习目标

通过在 .NET 中实现一个基础代理，本示例探讨了 Microsoft Agent Framework 的基本概念。您将学习核心代理模式，并通过 C# 和 .NET 生态系统了解智能代理的工作原理。

您将学到的内容

• 🏗️ 代理架构：了解 .NET 中 AI 代理的基本结构
• 🛠️ 工具集成：代理如何使用外部函数扩展功能
• 💬 对话流程：通过线程管理处理多轮对话和上下文
• 🔧 配置模式：在 .NET 中设置和管理代理的最佳实践

🎯 涵盖的关键概念

代理框架原则

• 自主性：代理如何使用 .NET AI 抽象做出独立决策
• 反应性：响应环境变化和用户输入
• 主动性：根据目标和上下文采取行动
• 社交能力：通过自然语言与对话线程进行交互

技术组件

• AIAgent：核心代理编排和对话管理 (.NET)
• 工具函数：通过 C# 方法和属性扩展代理功能
• OpenAI 集成：通过标准化 .NET API 利用语言模型
• 安全配置：基于环境的 API 密钥管理

🔧 技术栈

核心技术

• Microsoft Agent Framework (.NET)
• GitHub Models API 集成
• OpenAI 兼容客户端模式
• 使用 DotNetEnv 的基于环境的配置

代理功能

• 自然语言理解和生成
• 使用 C# 属性进行函数调用和工具使用
• 基于对话线程的上下文感知响应
• 使用依赖注入模式的可扩展架构

📚 框架对比

本示例展示了 Microsoft Agent Framework 与其他代理框架的对比：

功能	Microsoft Agent Framework	其他框架
集成	原生 Microsoft 生态系统	兼容性各异
简洁性	简洁直观的 API	通常设置复杂
可扩展性	易于工具集成	依赖框架
企业级	为生产环境而设计	因框架而异

🚀 入门指南

前置条件

• .NET 10 SDK 或更高版本
• GitHub Models API 访问令牌

必需的环境变量

# zsh/bash
export GH_TOKEN=<your_github_token>
export GH_ENDPOINT=https://models.github.ai/inference
export GH_MODEL_ID=openai/gpt-5-mini

# PowerShell
$env:GH_TOKEN = "<your_github_token>"
$env:GH_ENDPOINT = "https://models.github.ai/inference"
$env:GH_MODEL_ID = "openai/gpt-5-mini"

示例代码

运行代码示例：

# zsh/bash
chmod +x ./02-dotnet-agent-framework.cs
./02-dotnet-agent-framework.cs

或者使用 dotnet CLI：

dotnet run ./02-dotnet-agent-framework.cs

查看完整代码 02-dotnet-agent-framework.cs。

#!/usr/bin/dotnet run

#:package Microsoft.Extensions.AI@10.*
#:package Microsoft.Agents.AI.OpenAI@1.*-*

using System.ClientModel;
using System.ComponentModel;

using Microsoft.Agents.AI;
using Microsoft.Extensions.AI;

using OpenAI;

// Tool Function: Random Destination Generator
// This static method will be available to the agent as a callable tool
// The [Description] attribute helps the AI understand when to use this function
// This demonstrates how to create custom tools for AI agents
[Description("Provides a random vacation destination.")]
static string GetRandomDestination()
{
    // List of popular vacation destinations around the world
    // The agent will randomly select from these options
    var destinations = new List<string>
    {
        "Paris, France",
        "Tokyo, Japan",
        "New York City, USA",
        "Sydney, Australia",
        "Rome, Italy",
        "Barcelona, Spain",
        "Cape Town, South Africa",
        "Rio de Janeiro, Brazil",
        "Bangkok, Thailand",
        "Vancouver, Canada"
    };

    // Generate random index and return selected destination
    // Uses System.Random for simple random selection
    var random = new Random();
    int index = random.Next(destinations.Count);
    return destinations[index];
}

// Extract configuration from environment variables
// Retrieve the GitHub Models API endpoint, defaults to https://models.github.ai/inference if not specified
// Retrieve the model ID, defaults to openai/gpt-5-mini if not specified
// Retrieve the GitHub token for authentication, throws exception if not specified
var github_endpoint = Environment.GetEnvironmentVariable("GH_ENDPOINT") ?? "https://models.github.ai/inference";
var github_model_id = Environment.GetEnvironmentVariable("GH_MODEL_ID") ?? "openai/gpt-5-mini";
var github_token = Environment.GetEnvironmentVariable("GH_TOKEN") ?? throw new InvalidOperationException("GH_TOKEN is not set.");

// Configure OpenAI Client Options
// Create configuration options to point to GitHub Models endpoint
// This redirects OpenAI client calls to GitHub's model inference service
var openAIOptions = new OpenAIClientOptions()
{
    Endpoint = new Uri(github_endpoint)
};

// Initialize OpenAI Client with GitHub Models Configuration
// Create OpenAI client using GitHub token for authentication
// Configure it to use GitHub Models endpoint instead of OpenAI directly
var openAIClient = new OpenAIClient(new ApiKeyCredential(github_token), openAIOptions);

// Define Agent Identity and Comprehensive Instructions
// Agent name for identification and logging purposes
var AGENT_NAME = "TravelAgent";

// Detailed instructions that define the agent's personality, capabilities, and behavior
// This system prompt shapes how the agent responds and interacts with users
var AGENT_INSTRUCTIONS = """
You are a helpful AI Agent that can help plan vacations for customers.

Important: When users specify a destination, always plan for that location. Only suggest random destinations when the user hasn't specified a preference.

When the conversation begins, introduce yourself with this message:
"Hello! I'm your TravelAgent assistant. I can help plan vacations and suggest interesting destinations for you. Here are some things you can ask me:
1. Plan a day trip to a specific location
2. Suggest a random vacation destination
3. Find destinations with specific features (beaches, mountains, historical sites, etc.)
4. Plan an alternative trip if you don't like my first suggestion

What kind of trip would you like me to help you plan today?"

Always prioritize user preferences. If they mention a specific destination like "Bali" or "Paris," focus your planning on that location rather than suggesting alternatives.
""";

// Create AI Agent with Advanced Travel Planning Capabilities
// Initialize complete agent pipeline: OpenAI client → Chat client → AI agent
// Configure agent with name, detailed instructions, and available tools
// This demonstrates the .NET agent creation pattern with full configuration
AIAgent agent = openAIClient
    .GetChatClient(github_model_id)
    .CreateAIAgent(
        name: AGENT_NAME,
        instructions: AGENT_INSTRUCTIONS,
        tools: [AIFunctionFactory.Create(GetRandomDestination)]
    );

// Create New Conversation Thread for Context Management
// Initialize a new conversation thread to maintain context across multiple interactions
// Threads enable the agent to remember previous exchanges and maintain conversational state
// This is essential for multi-turn conversations and contextual understanding
AgentThread thread = agent.GetNewThread();

// Execute Agent: First Travel Planning Request
// Run the agent with an initial request that will likely trigger the random destination tool
// The agent will analyze the request, use the GetRandomDestination tool, and create an itinerary
// Using the thread parameter maintains conversation context for subsequent interactions
await foreach (var update in agent.RunStreamingAsync("Plan me a day trip", thread))
{
    await Task.Delay(10);
    Console.Write(update);
}

Console.WriteLine();

// Execute Agent: Follow-up Request with Context Awareness
// Demonstrate contextual conversation by referencing the previous response
// The agent remembers the previous destination suggestion and will provide an alternative
// This showcases the power of conversation threads and contextual understanding in .NET agents
await foreach (var update in agent.RunStreamingAsync("I don't like that destination. Plan me another vacation.", thread))
{
    await Task.Delay(10);
    Console.Write(update);
}

🎓 关键收获

1. 代理架构：Microsoft Agent Framework 提供了一种简洁、类型安全的方式来在 .NET 中构建 AI 代理
2. 工具集成：使用 [Description] 属性装饰的函数成为代理可用的工具
3. 对话上下文：线程管理支持多轮对话并具备完整的上下文感知能力
4. 配置管理：环境变量和安全凭证处理遵循 .NET 的最佳实践
5. OpenAI 兼容性：GitHub Models 集成通过 OpenAI 兼容 API 无缝工作

🔗 其他资源

• Microsoft Agent Framework 文档
• GitHub Models Marketplace
• Microsoft.Extensions.AI
• .NET 单文件应用

---

免责声明：
本文档使用AI翻译服务Co-op Translator进行翻译。尽管我们努力确保准确性，但请注意，自动翻译可能包含错误或不准确之处。应以原始语言的文档为权威来源。对于关键信息，建议使用专业人工翻译。因使用本翻译而引起的任何误解或误读，我们概不负责。

This site is open source. Improve this page.