{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Use this to design AI agent as a POC\n",
    "<h2>初始化</h2>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 读取环境变量\n",
    "from dotenv import load_dotenv\n",
    "load_dotenv()  # 加载 .env 文件"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<hr>\n",
    "<h2>定义使用搜索引擎</h2>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from langchain_community.tools.tavily_search import TavilySearchResults\n",
    "\n",
    "tool = TavilySearchResults(max_results=3)\n",
    "tools = [tool]\n",
    "# tool.invoke(\"What's a 'node' in LangGraph?\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<hr>\n",
    "<h2>使用通义千问作为LLM</h2>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 使用通义千问\n",
    "from langchain_community.llms import Tongyi\n",
    "\n",
    "import os\n",
    "\n",
    "# 初始化通义模型（以qwen-max为例）\n",
    "llm_tongyi = Tongyi(\n",
    "    model_name=\"qwen-turbo\",\n",
    "    dashscope_api_key=os.getenv(\"DASHSCOPE_API_KEY\")\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<h2>手动实现通义的工具调用流程</h2>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from typing import List, Dict, Any\n",
    "from langchain_core.messages import HumanMessage\n",
    "\n",
    "def manual_tool_invocation(llm: Tongyi, tools: List, query: str) -> str:\n",
    "\n",
    "    # 0. 调用搜索引擎\n",
    "    tool.invoke(query)\n",
    "\n",
    "    # 1. 构建工具描述\n",
    "    tool_descs = \"\\n\".join([\n",
    "        f\"{tool.name}: {tool.description}\\n参数: {tool.args}\"\n",
    "        for tool in tools\n",
    "    ])\n",
    "    \n",
    "    # 2. 构造特殊 prompt\n",
    "    prompt = f\"\"\"请根据问题决定是否需要使用工具。可用工具：\n",
    "    {tool_descs}\n",
    "    \n",
    "    问题：{query}\n",
    "    \n",
    "    如果需要使用工具，请严格按以下格式回复：\n",
    "    ```json\n",
    "    {{\"tool\": \"工具名\", \"args\": {{\"参数名\": \"参数值\"}}}}\n",
    "    ```\n",
    "    如果不需要工具，请直接回答问题。\"\"\"\n",
    "    \n",
    "    # 3. 调用模型\n",
    "    response = llm.invoke(prompt)\n",
    "    \n",
    "    # 4. 解析工具调用\n",
    "    if \"```json\" in response:\n",
    "        try:\n",
    "            import json\n",
    "            tool_call = json.loads(response.split(\"```json\")[1].split(\"```\")[0].strip())\n",
    "            selected_tool = next(t for t in tools if t.name == tool_call[\"tool\"])\n",
    "            return selected_tool.invoke(tool_call[\"args\"])\n",
    "        except:\n",
    "            return f\"工具调用失败：{response}\"\n",
    "    \n",
    "    return response\n",
    "\n",
    "# 使用示例\n",
    "# result = manual_tool_invocation(llm_tongyi, tools, \"What's a 'node' in LangGraph?\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<h2>定义智能体工作流</h2>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from typing import Annotated\n",
    "\n",
    "from typing_extensions import TypedDict\n",
    "\n",
    "from langgraph.graph import StateGraph, START, END\n",
    "from langgraph.graph.message import add_messages\n",
    "\n",
    "class State(TypedDict):\n",
    "    # Messages have the type \"list\". The `add_messages` function\n",
    "    # in the annotation defines how this state key should be updated\n",
    "    # (in this case, it appends messages to the list, rather than overwriting them)\n",
    "    messages: Annotated[list, add_messages]\n",
    "\n",
    "graph_builder = StateGraph(State)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<h2>添加智能体节点</h2>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "def chatbot(state: State):\n",
    "    # return {\"messages\": [llm_tongyi.invoke(state[\"messages\"])]}\n",
    "    return {\"messages\": [manual_tool_invocation(llm_tongyi, tools, \"What's a 'node' in LangGraph?\")]}\n",
    "\n",
    "# The first argument is the unique node name\n",
    "# The second argument is the function or object that will be called whenever\n",
    "# the node is used.\n",
    "graph_builder.add_node(\"chatbot\", chatbot)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<h2>添加工作流起点和终点</h2>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 定义起点\n",
    "graph_builder.add_edge(START, \"chatbot\")\n",
    "# 定义终点\n",
    "graph_builder.add_edge(\"chatbot\", END)\n",
    "\n",
    "# 完成编辑\n",
    "graph = graph_builder.compile()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<h2>展示工作流</h2>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from IPython.display import Image, display\n",
    "\n",
    "try:\n",
    "    display(Image(graph.get_graph().draw_mermaid_png()))\n",
    "except Exception:\n",
    "    # This requires some extra dependencies and is optional\n",
    "    pass"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<h2>执行工作流</h2>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 流式交互函数（适配通义模型）\n",
    "def stream_graph_updates(user_input: str):\n",
    "    state = {\"messages\": [{\"role\": \"user\", \"content\": user_input}]}\n",
    "\n",
    "    for event in graph.stream(state):\n",
    "        for value in event.values():\n",
    "            last_message = value[\"messages\"][-1]\n",
    "            print(f\"Assistant: {last_message}\")\n",
    "\n",
    "while True:\n",
    "    try:\n",
    "        user_input = input(\"User: \")\n",
    "        print(\"User: \", user_input)\n",
    "        if user_input.lower() in [\"quit\", \"exit\", \"q\"]:\n",
    "            print(\"Goodbye!\")\n",
    "            break\n",
    "        stream_graph_updates(user_input)\n",
    "    except:\n",
    "        # fallback if input() is not available\n",
    "        user_input = \"What do you know about LangGraph?\"\n",
    "        print(\"User: \" + user_input)\n",
    "        stream_graph_updates(user_input)\n",
    "        break"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.10.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}