Semantic Kernel

使用这个企业级编排框架，构建智能 AI 代理与多代理系统

什么是 Semantic Kernel？

Semantic Kernel 是一个模型无关的 SDK，它使开发者能够构建、编排和部署 AI 代理与多代理系统。无论你是构建一个简单的聊天机器人，还是一个复杂的多代理工作流，Semantic Kernel 都能提供你所需的企业级可靠性与灵活性工具。

系统要求

• Python: 3.10+
• .NET: .NET 10.0+
• Java: JDK 17+
• 操作系统支持: Windows, macOS, Linux

核心特性

• 模型灵活性: 连接任何大语言模型，内置支持 OpenAI、Azure OpenAI、Hugging Face、NVidia 等
• 代理框架: 构建模块化的 AI 代理，具备工具/插件、记忆和规划能力
• 多代理系统: 通过协作的专家代理编排复杂工作流
• 插件生态系统: 通过原生代码函数、提示词模板、OpenAPI 规范或模型上下文协议 (MCP) 进行扩展
• 向量数据库支持: 与 Azure AI Search、Elasticsearch、Chroma 等无缝集成
• 多模态支持: 处理文本、视觉和音频输入
• 本地部署: 支持使用 Ollama、LMStudio 或 ONNX 运行
• 流程框架: 使用结构化工作流方法对复杂业务流程进行建模
• 企业就绪: 为可观测性、安全性和稳定的 API 而构建

安装

首先，为你的 AI 服务设置环境变量：

Azure OpenAI:

export AZURE_OPENAI_API_KEY=AAA....

或直接使用 OpenAI:

export OPENAI_API_KEY=sk-...

Python

pip install semantic-kernel

.NET

dotnet add package Microsoft.SemanticKernel
dotnet add package Microsoft.SemanticKernel.Agents.Core

Java

请参阅 semantic-kernel-java 构建说明。

快速开始

基础代理 - Python

创建一个响应提示的简单助手：

import asyncio
from semantic_kernel.agents import ChatCompletionAgent
from semantic_kernel.connectors.ai.open_ai import AzureChatCompletion

async def main():
    # 使用基本指令初始化聊天代理
    agent = ChatCompletionAgent(
        service=AzureChatCompletion(),
        name="SK-Assistant",
        instructions="You are a helpful assistant.",
    )

    # 获取对用户消息的响应
    response = await agent.get_response(messages="Write a haiku about Semantic Kernel.")
    print(response.content)

asyncio.run(main()) 

# 输出:
# Language's essence,
# Semantic threads intertwine,
# Meaning's core revealed.

基础代理 - .NET

using Microsoft.SemanticKernel;
using Microsoft.SemanticKernel.Agents;

var builder = Kernel.CreateBuilder();
builder.AddAzureOpenAIChatCompletion(
                Environment.GetEnvironmentVariable("AZURE_OPENAI_DEPLOYMENT"),
                Environment.GetEnvironmentVariable("AZURE_OPENAI_ENDPOINT"),
                Environment.GetEnvironmentVariable("AZURE_OPENAI_API_KEY")
                );
var kernel = builder.Build();

ChatCompletionAgent agent =
    new()
    {
        Name = "SK-Agent",
        Instructions = "You are a helpful assistant.",
        Kernel = kernel,
    };

await foreach (AgentResponseItem<ChatMessageContent> response 
    in agent.InvokeAsync("Write a haiku about Semantic Kernel."))
{
    Console.WriteLine(response.Message);
}

// 输出:
// Language's essence,
// Semantic threads intertwine,
// Meaning's core revealed.

带插件的代理 - Python

使用自定义工具（插件）和结构化输出来增强你的代理：

import asyncio
from typing import Annotated
from pydantic import BaseModel
from semantic_kernel.agents import ChatCompletionAgent
from semantic_kernel.connectors.ai.open_ai import AzureChatCompletion, OpenAIChatPromptExecutionSettings
from semantic_kernel.functions import kernel_function, KernelArguments

class MenuPlugin:
    @kernel_function(description="Provides a list of specials from the menu.")
    def get_specials(self) -> Annotated[str, "Returns the specials from the menu."]:
        return """
        Special Soup: Clam Chowder
        Special Salad: Cobb Salad
        Special Drink: Chai Tea
        """

    @kernel_function(description="Provides the price of the requested menu item.")
    def get_item_price(
        self, menu_item: Annotated[str, "The name of the menu item."]
    ) -> Annotated[str, "Returns the price of the menu item."]:
        return "$9.99"

class MenuItem(BaseModel):
    price: float
    name: str

async def main():
    # 配置结构化输出格式
    settings = OpenAIChatPromptExecutionSettings()
    settings.response_format = MenuItem

    # 创建带有插件和设置的代理
    agent = ChatCompletionAgent(
        service=AzureChatCompletion(),
        name="SK-Assistant",
        instructions="You are a helpful assistant.",
        plugins=[MenuPlugin()],
        arguments=KernelArguments(settings)
    )

    response = await agent.get_response(messages="What is the price of the soup special?")
    print(response.content)

    # 输出:
    # The price of the Clam Chowder, which is the soup special, is $9.99.

asyncio.run(main()) 

带插件的代理 - .NET

using System.ComponentModel;
using Microsoft.SemanticKernel;
using Microsoft.SemanticKernel.Agents;
using Microsoft.SemanticKernel.ChatCompletion;

var builder = Kernel.CreateBuilder();
builder.AddAzureOpenAIChatCompletion(
                Environment.GetEnvironmentVariable("AZURE_OPENAI_DEPLOYMENT"),
                Environment.GetEnvironmentVariable("AZURE_OPENAI_ENDPOINT"),
                Environment.GetEnvironmentVariable("AZURE_OPENAI_API_KEY")
                );
var kernel = builder.Build();

kernel.Plugins.Add(KernelPluginFactory.CreateFromType<MenuPlugin>());

ChatCompletionAgent agent =
    new()
    {
        Name = "SK-Assistant",
        Instructions = "You are a helpful assistant.",
        Kernel = kernel,
        Arguments = new KernelArguments(new PromptExecutionSettings() { FunctionChoiceBehavior = FunctionChoiceBehavior.Auto() })

    };

await foreach (AgentResponseItem<ChatMessageContent> response 
    in agent.InvokeAsync("What is the price of the soup special?"))
{
    Console.WriteLine(response.Message);
}

sealed class MenuPlugin
{
    [KernelFunction, Description("Provides a list of specials from the menu.")]
    public string GetSpecials() =>
        """
        Special Soup: Clam Chowder
        Special Salad: Cobb Salad
        Special Drink: Chai Tea
        """;

    [KernelFunction, Description("Provides the price of the requested menu item.")]
    public string GetItemPrice(
        [Description("The name of the menu item.")]
        string menuItem) =>
        "$9.99";
}

多代理系统 - Python

构建一个可以协作的专家代理系统：

import asyncio
from semantic_kernel.agents import ChatCompletionAgent, ChatHistoryAgentThread
from semantic_kernel.connectors.ai.open_ai import AzureChatCompletion, OpenAIChatCompletion

billing_agent = ChatCompletionAgent(
    service=AzureChatCompletion(), 
    name="BillingAgent", 
    instructions="You handle billing issues like charges, payment methods, cycles, fees, discrepancies, and payment failures."
)

refund_agent = ChatCompletionAgent(
    service=AzureChatCompletion(),
    name="RefundAgent",
    instructions="Assist users with refund inquiries, including eligibility, policies, processing, and status updates.",
)

triage_agent = ChatCompletionAgent(
    service=OpenAIChatCompletion(),
    name="TriageAgent",
    instructions="Evaluate user requests and forward them to BillingAgent or RefundAgent for targeted assistance."
    " Provide the full answer to the user containing any information from the agents",
    plugins=[billing_agent, refund_agent],
)

thread: ChatHistoryAgentThread = None

async def main() -> None:
    print("Welcome to the chat bot!\n  Type 'exit' to exit.\n  Try to get some billing or refund help.")
    while True:
        user_input = input("User:> ")

        if user_input.lower().strip() == "exit":
            print("\n\nExiting chat...")
            return False

        response = await triage_agent.get_response(
            messages=user_input,
            thread=thread,
        )

        if response:
            print(f"Agent :> {response}")

# Agent :> I understand that you were charged twice for your subscription last month, and I'm here to assist you with resolving this issue. Here’s what we need to do next:

# 1. **Billing Inquiry**:
#    - Please provide the email address or account number associated with your subscription, the date(s) of the charges, and the amount charged. This will allow the billing team to investigate the discrepancy in the charges.

# 2. **Refund Process**:
#    - For the refund, please confirm your subscription type and the email address associated with your account.
#    - Provide the dates and transaction IDs for the charges you believe were duplicated.

# Once we have these details, we will be able to:

# - Check your billing history for any discrepancies.
# - Confirm any duplicate charges.
# - Initiate a refund for the duplicate payment if it qualifies. The refund process usually takes 5-10 business days after approval.

# Please provide the necessary details so we can proceed with resolving this issue for you.


if __name__ == "__main__":
    asyncio.run(main())

下一步

1. 📖 尝试我们的 快速入门指南 或了解 构建代理
2. 🔌 探索超过 100 个 详细示例
3. 💡 学习 Semantic Kernel 核心 概念

API 参考

• C# API 参考
• Python API 参考

故障排除

常见问题

• 认证错误: 检查你的 API 密钥环境变量是否正确设置
• 模型可用性: 验证你的 Azure OpenAI 部署或 OpenAI 模型访问权限

获取帮助

• 查看我们的 GitHub issues 了解已知问题
• 在 Discord 社区 中搜索解决方案
• 寻求帮助时，请包含你的 SDK 版本和完整的错误信息

加入社区

我们欢迎您为 SK 社区做出贡献和提出建议！参与 GitHub 仓库的讨论是最简单的参与方式之一。欢迎提交错误报告和修复！

对于新功能、组件或扩展，请在发送 PR 之前先开一个 issue 与我们讨论。这是为了避免被拒绝，因为我们可能正在将核心向不同方向发展，同时也需要考虑对更大生态系统的影响。

要了解更多信息并开始使用：

• 阅读 文档
• 学习如何 贡献 到项目
• 在 GitHub 讨论区 提问
• 在 Discord 社区 提问
• 参加 定期办公时间与 SK 社区活动
• 在我们的 博客 上关注团队

贡献者荣誉墙

行为准则

本项目采用了
Microsoft 开源行为准则。
欲了解更多信息，请参阅
行为准则 FAQ
或联系 opencode@microsoft.com
提出任何其他问题或意见。

许可证

版权所有 (c) Microsoft Corporation。保留所有权利。

根据 MIT 许可证授权。