import asyncio
from semantic_kernel import Kernel
from semantic_kernel.connectors.ai.open_ai import OpenAIChatCompletion
from semantic_kernel.contents import ChatHistory

async def main():
    # 1. 初始化 Kernel 容器
    kernel = Kernel()

    # 2. 注入大模型后端服务 (支持 OpenAI 或 Azure OpenAI)
    kernel.add_service(
        OpenAIChatCompletion(
            service_id="chat-service",
            ai_model_id="gpt-4o-mini",
            api_key="sk-your-openai-api-key",
        )
    )

    # 3. 按 service_id 获取聊天服务句柄
    chat = kernel.get_service("chat-service")
    
    # 4. 构建结构化多轮历史
    history = ChatHistory()
    history.add_user_message("请用一句话讲清楚 Semantic Kernel 的核心设计优势。")

    # 5. 执行调用
    settings = kernel.get_prompt_execution_settings_from_service_id("chat-service")
    result = await chat.get_chat_message_content(
        chat_history=history,
        settings=settings,
    )
    print(f"Agent 回复:\n{result}")

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

from typing import Annotated
from semantic_kernel.functions import kernel_function

class WeatherPlugin:
    """天气查询业务插件"""

    @kernel_function(
        name="get_current_weather",
        description="获取指定城市的实时天气数据",
    )
    def get_current_weather(
        self,
        city: Annotated[str, "需要查询的城市名称，例如：北京、上海、深圳"],
    ) -> Annotated[str, "天气描述字符串"]:
        data_mock = {
            "北京": "晴朗，气温 24°C，微风",
            "上海": "多云转小雨，气温 21°C",
            "深圳": "雷阵雨，气温 29°C",
        }
        return data_mock.get(city, f"{city}：暂未接入气象监控站点数据")

    @kernel_function(
        name="get_forecast",
        description="预测指定城市未来 3 天的天气走势",
    )
    def get_forecast(
        self,
        city: Annotated[str, "城市名称"],
    ) -> str:
        return f"{city} 未来 3 天趋势：第 1 天晴，第 2 天阴，第 3 天小雨"


class FinancialMathPlugin:
    """高精度企业计算插件"""

    @kernel_function(description="计算两个数值的和")
    def add(
        self,
        a: Annotated[float, "被加数"],
        b: Annotated[float, "加数"],
    ) -> float:
        return a + b

    @kernel_function(description="计算两数之积")
    def multiply(
        self,
        a: Annotated[float, "被乘数"],
        b: Annotated[float, "乘数"],
    ) -> float:
        return a * b

import asyncio
from semantic_kernel import Kernel

async def run_plugin():
    kernel = Kernel()

    # 注册插件实例到 Kernel 中
    kernel.add_plugin(WeatherPlugin(), plugin_name="Weather")
    kernel.add_plugin(FinancialMathPlugin(), plugin_name="Math")

    # 显式根据插件名和方法名调用
    weather_res = await kernel.invoke(
        kernel.get_function("Weather", "get_current_weather"),
        city="北京",
    )
    print("插件输出:", weather_res)  # 晴朗，气温 24°C，微风

    calc_res = await kernel.invoke(
        kernel.get_function("Math", "multiply"),
        a=12.5,
        b=8.0,
    )
    print("计算输出:", calc_res)   # 100.0

asyncio.run(run_plugin())

from semantic_kernel.functions import KernelFunctionFromPrompt

# 用 Prompt 模板生成标准函数
summarize_function = KernelFunctionFromPrompt(
    function_name="summarize_text",
    plugin_name="TextProcessing",
    prompt="""
请对以下输入文本进行结构化提炼，提取出 3 个核心要点，每点控制在 20 字以内：

文本内容：
{{$input}}

提炼结果：
""",
)

# 挂载到 kernel
kernel.add_function("TextProcessing", summarize_function)

# 执行语义函数
res = await kernel.invoke(
    summarize_function,
    input="AgentScope 是阿里巴巴达摩院开源的分布式 Agent 框架，主打消息一等公民和 RPC 解耦...",
)
print("语义摘要结果:\n", res)

import asyncio
from semantic_kernel import Kernel
from semantic_kernel.connectors.ai.open_ai import (
    OpenAIChatCompletion,
    OpenAIChatPromptExecutionSettings,
)
from semantic_kernel.connectors.ai.function_choice_behavior import FunctionChoiceBehavior
from semantic_kernel.contents import ChatHistory

async def auto_tool_calling():
    kernel = Kernel()
    kernel.add_service(
        OpenAIChatCompletion(service_id="chat", ai_model_id="gpt-4o")
    )

    # 注册可用插件
    kernel.add_plugin(WeatherPlugin(), plugin_name="Weather")
    kernel.add_plugin(FinancialMathPlugin(), plugin_name="Math")

    # 启用自动函数选择 (Auto Tool Calling)
    execution_settings = OpenAIChatPromptExecutionSettings(
        function_choice_behavior=FunctionChoiceBehavior.Auto(
            auto_invoke=True,  # 由 Kernel 自动拦截并执行函数，将结果送回 LLM 获得最终回答
            max_auto_invoke_attempts=5,
        )
    )

    history = ChatHistory()
    history.add_user_message("北京现在的天气如何？另外帮我算一下 24.5 乘以 4 等于多少？")

    chat_service = kernel.get_service("chat")
    response = await chat_service.get_chat_message_content(
        chat_history=history,
        settings=execution_settings,
        kernel=kernel,  # 将 kernel 传入上下文，赋予 LLM 检索和运行插件的权限
    )

    print("最终 Agent 综合回答:\n", response.content)

asyncio.run(auto_tool_calling())

from semantic_kernel.memory import SemanticTextMemory
from semantic_kernel.connectors.memory.chroma import ChromaMemoryStore
from semantic_kernel.connectors.ai.open_ai import OpenAITextEmbedding

# 构建持久化语义记忆系统
memory = SemanticTextMemory(
    storage=ChromaMemoryStore(persist_directory="./sk_chroma_db"),
    embeddings_generator=OpenAITextEmbedding(ai_model_id="text-embedding-3-small"),
)

# 写入业务文档
await memory.save_information(
    collection="enterprise_docs",
    id="policy_001",
    text="公司年假制度规定：入职满一年的员工每年享有 10 个工作日的全薪年假。",
)

# 检索最相关的事实
matches = await memory.search(
    collection="enterprise_docs",
    query="新员工入职第二年有多少天带薪假？",
    limit=2,
    min_relevance_score=0.7,
)

for m in matches:
    print(f"相似度: {m.relevance:.2f} | 内容: {m.text}")