{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "181574eb-16d2-4bae-bd4c-b583fbc01df3",
   "metadata": {},
   "source": [
    "# 使用 API 快速搭建你的第一个 AI 应用\n",
    "> [HW3：以 AI 搭建自己的應用](https://colab.research.google.com/drive/15jh4v_TBPsTyIBhi0Fz46gEkjvhzGaBR?usp=sharing)中文镜像版\n",
    "> \n",
    "> 指导文章：[02. 简单入门：通过 API 与 Gradio 构建 AI 应用](https://github.com/Hoper-J/LLM-Guide-and-Demos/blob/master/Guide/02.%20简单入门：通过%20API%20与%20Gradio%20构建%20AI%20应用.md)\n",
    "\n",
    "目标：了解如何通过使用 API 和 Prompt 来构建自己的语言模型应用。\n",
    "\n",
    "在线链接：[Colab](https://colab.research.google.com/drive/1xR9BS1sD7htg7gaoEtNSIO7Ir3qC7_ec?usp=sharing)\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "625d15d0-bc47-4c04-8eec-779df027a447",
   "metadata": {},
   "source": [
    "# 安装库"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "4392ac0a-4964-4801-bbff-bcf49af8049b",
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "%pip install openai\n",
    "%pip install gradio\n",
    "%pip install numpy"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4f4f1afe-84b6-4456-97e8-f74bae95bad1",
   "metadata": {},
   "source": [
    "# 导入与设置"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "ba854a1c-f84e-4f1d-b0c5-c0674a42dfa7",
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import json\n",
    "from typing import List, Dict, Tuple\n",
    "\n",
    "import openai\n",
    "import gradio as gr"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "68af81ab-e1bb-45d1-9393-f171e756146c",
   "metadata": {},
   "outputs": [],
   "source": [
    "# TODO: 设置你的 OPENAI API 密钥，这里假设 DashScope API 被配置在了 OPENAI_API_KEY 环境变量中\n",
    "OPENAI_API_KEY = \"\"\n",
    "# 不填写则默认使用环境变量\n",
    "if not OPENAI_API_KEY:\n",
    "    OPENAI_API_KEY = os.getenv('OPENAI_API_KEY')\n",
    "\n",
    "# 初始化 OpenAI 客户端，使用阿里云 DashScope API\n",
    "client = openai.OpenAI(\n",
    "    api_key=OPENAI_API_KEY,\n",
    "    base_url=\"https://dashscope.aliyuncs.com/compatible-mode/v1\",  # 阿里云的 API 地址\n",
    ")\n",
    "\n",
    "# 检查 API 设置是否正确\n",
    "try:\n",
    "    response = client.chat.completions.create(\n",
    "        model=\"qwen-turbo\",  # 使用通义千问-Turbo 大模型，可以替换为 Deepseek 系列：deepseek-v3 / deepseek-r1\n",
    "        messages=[{'role': 'user', 'content': \"测试\"}],\n",
    "        max_tokens=1,\n",
    "    )\n",
    "    print(\"API 设置成功！！\")\n",
    "except Exception as e:\n",
    "    print(f\"API 可能有问题，请检查：{e}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "60d0e850-70c4-4fae-a79c-5a9dae3200bb",
   "metadata": {},
   "source": [
    "## 第1部分：文章摘要（单轮对话应用）\n",
    "\n",
    "在此任务中，你需要将你的聊天机器人变为一个**摘要器**。它的工作是当用户输入一篇文章时，能够为用户总结该文章的内容。\n",
    "\n",
    "你需要完成以下步骤：\n",
    "\n",
    "1. 设计一个用于生成摘要的提示词，并填写在 **prompt_for_summarization** 中。\n",
    "2. **点击运行按钮**， 这将弹出一个可交互的界面。\n",
    "3. 你可以找到一篇文章或使用当前的示例文章：《从百草园到三味书屋》，并将其填写在标记为“文章”的输入框中。\n",
    "4. 点击“发送”按钮生成文章的摘要。（你可以使用“温度”滑块来控制输出的创造性，温度越高，输出越具创造性）。\n",
    "5. 如果你**想更改提示词**，可以停止单元格，返回到TODO部分进行更改，然后再次运行。\n",
    "7. 在你获得满意的结果后，点击“导出”按钮保存结果。文件列表中将出现一个名为 **part1.json** 的文件。\n",
    "\n",
    "注意：\n",
    "\n",
    "- **如果你再次点击“导出”按钮，之前的结果将被覆盖。**\n",
    "- **即使使用相同的提示词，输出的结果可能仍然不同。**\n",
    "\n",
    "------\n",
    "\n",
    "在运行此单元格之前，请确保已运行 **安装包** 和 **导入与设置**。\n",
    "\n",
    "**记得在进行下一步前停止此单元格。**\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "d7fb78cb-43a6-48cf-8d48-832ee7cbd1d3",
   "metadata": {},
   "outputs": [],
   "source": [
    "# TODO: 修改提示词以满足你的摘要需求\n",
    "PROMPT_FOR_SUMMARIZATION = \"请将以下文章概括成几句话。\"\n",
    "\n",
    "def reset():\n",
    "    \"\"\"\n",
    "    清空对话记录\n",
    "\n",
    "    返回:\n",
    "        List: 空的对话记录列表\n",
    "    \"\"\"\n",
    "    return []\n",
    "\n",
    "def interact_summarization(prompt, article, temp=1.0):\n",
    "    \"\"\"\n",
    "    调用模型生成摘要。\n",
    "\n",
    "    参数:\n",
    "        prompt (str): 用于摘要的提示词\n",
    "        article (str): 需要摘要的文章内容\n",
    "        temp (float): 模型温度，控制输出创造性（默认 1.0）\n",
    "\n",
    "    返回:\n",
    "        List[Tuple[str, str]]: 对话记录，包含输入文本与模型输出\n",
    "    \"\"\"\n",
    "    # 合成请求文本\n",
    "    input_text = f\"{prompt}\\n{article}\"\n",
    "    \n",
    "    response = client.chat.completions.create(\n",
    "        model=\"qwen-turbo\",  # 使用通义千问-Turbo大模型\n",
    "        messages=[{'role': 'user', 'content': input_text}],\n",
    "        temperature=temp,\n",
    "    )\n",
    "    return [(input_text, response.choices[0].message.content)]\n",
    "\n",
    "def export_summarization(chatbot, article):\n",
    "    \"\"\"\n",
    "    导出摘要任务的对话记录和文章内容到 JSON 文件。\n",
    "\n",
    "    参数:\n",
    "        chatbot (List[Tuple[str, str]]): 模型对话记录\n",
    "        article (str): 文章内容\n",
    "    \"\"\"\n",
    "    target = {\"chatbot\": chatbot, \"article\": article}\n",
    "    with open(\"files/part1.json\", \"w\", encoding=\"utf-8\") as file:\n",
    "        json.dump(target, file, ensure_ascii=False, indent=4)\n",
    "\n",
    "# 构建 Gradio UI 界面\n",
    "with gr.Blocks() as demo:\n",
    "    gr.Markdown(\"# 第1部分：摘要\\n填写任何你喜欢的文章，让聊天机器人为你总结！\")\n",
    "    chatbot = gr.Chatbot(type=\"tuples\")\n",
    "    prompt_textbox = gr.Textbox(label=\"提示词\", value=PROMPT_FOR_SUMMARIZATION, visible=False)\n",
    "    article_textbox = gr.Textbox(label=\"文章\", interactive=True, value=\"填充\")\n",
    "    \n",
    "    with gr.Column():\n",
    "        gr.Markdown(\"# 温度调节\\n温度用于控制聊天机器人的输出，温度越高，响应越具创造性。\")\n",
    "        temperature_slider = gr.Slider(0.0, 1.99, 1.0, step=0.01, label=\"温度\")\n",
    "    \n",
    "    with gr.Row():\n",
    "        send_button = gr.Button(value=\"发送\")\n",
    "        reset_button = gr.Button(value=\"重置\")\n",
    "    \n",
    "    with gr.Column():\n",
    "        gr.Markdown(\"# 保存结果\\n当你对结果满意后，点击导出按钮保存结果。\")\n",
    "        export_button = gr.Button(value=\"导出\")\n",
    "    \n",
    "    # 绑定按钮与回调函数\n",
    "    send_button.click(interact_summarization,\n",
    "                      inputs=[prompt_textbox, article_textbox, temperature_slider],\n",
    "                      outputs=[chatbot])\n",
    "    reset_button.click(reset, outputs=[chatbot])\n",
    "    export_button.click(export_summarization, inputs=[chatbot, article_textbox])\n",
    "\n",
    "# 启动 Gradio 应用\n",
    "demo.launch(debug=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5d239804-f4e2-46c8-b31c-83a21d7c0f46",
   "metadata": {},
   "source": [
    "### 检查并打印你的结果\n",
    "\n",
    "此部分用于检查你的“part1.json”文件是否包含所有正确的上下文，或者用于查看我们提供的示例文件。\n",
    "\n",
    "你需要：\n",
    "\n",
    "1. 确保文件列表中有你保存的 **part1.json** 文件。\n",
    "2. 点击运行按钮。它将显示一个**冻结的** Gradio 界面，重现你的摘要结果。\n",
    "\n",
    "---\n",
    "\n",
    "在运行此单元格之前，请确保已运行 **安装库** 和 **导入与设置**。\n",
    "\n",
    "**记得在进行下一步前停止此单元格。**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "d4a2fde6-aa69-41bc-bd59-27a87330d1fa",
   "metadata": {},
   "outputs": [],
   "source": [
    "# 加载对话记录的 JSON 文件\n",
    "with open(\"files/part1.json\", \"r\") as f:\n",
    "    context = json.load(f)\n",
    "\n",
    "chatbot = context['chatbot']  # 获取对话记录\n",
    "article = context['article']  # 获取原始文章\n",
    "summarization = chatbot[0][-1]  # 获取摘要结果\n",
    "\n",
    "# 构建 Gradio UI 界面\n",
    "with gr.Blocks() as demo:\n",
    "    gr.Markdown(\"# 第1部分：摘要\\n你可以查看文章和摘要！\")\n",
    "    chatbot = gr.Chatbot(type=\"tuples\", value=context['chatbot'])  # 加载对话历史\n",
    "    article_textbox = gr.Textbox(label=\"文章\", interactive=False, value=context['article'])  # 显示原始文章\n",
    "\n",
    "    # 构建展示摘要和原文的部分\n",
    "    with gr.Column():\n",
    "        gr.Markdown(\"# 只是一个检查\")\n",
    "        gr.Textbox(label=\"文章\", value=article, show_copy_button=True)  # 显示并允许复制原文\n",
    "        gr.Textbox(label=\"摘要\", value=summarization, show_copy_button=True)  # 显示并允许复制摘要\n",
    "\n",
    "# 启动 Gradio 应用\n",
    "demo.launch(debug=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9ba0c022-60e1-4efa-840e-158e41688ba1",
   "metadata": {},
   "source": [
    "## 第2部分：角色扮演（多轮对话应用）\n",
    "\n",
    "在此任务中，你需要将聊天机器人设定为**角色扮演模式**。你应该为它指定一个角色，然后通过提示让它进入该角色的状态。\n",
    "\n",
    "你需要完成以下步骤：\n",
    "\n",
    "1. 想出一个你希望聊天机器人扮演的**角色**，以及一个使聊天机器人进入该角色的提示词。在 **character_for_chatbot** 中填写角色，在 **prompt_for_roleplay** 中填写提示词。\n",
    "2. **点击运行按钮，界面将弹出一个可交互的界面。**\n",
    "3. **与聊天机器人进行** 2 轮 **互动**。在标为“输入”的框中输入你想说的话，然后点击“发送”按钮。（你可以使用“温度”滑块来控制输出的创造性。）\n",
    "4. 如果你**想更改提示词或角色**，可以停止单元格，返回TODO重新设置，然后重新运行单元格。\n",
    "5. 在你获得满意的结果后，点击“导出”按钮保存结果。文件列表中将出现一个名为 **part2.json** 的文件。\n",
    "\n",
    "注意：\n",
    "\n",
    "- **如果你再次点击“导出”按钮，之前的结果将被覆盖。**\n",
    "- **即使使用相同的提示词，输出的结果可能仍然不同。**\n",
    "\n",
    "------\n",
    "\n",
    "在运行此单元格之前，请确保已运行 **安装包** 和 **导入与设置**。\n",
    "\n",
    "**记得在进行下一步前停止此单元格。**\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "54852754-e02b-449d-8805-237ed1cb07fb",
   "metadata": {},
   "outputs": [],
   "source": [
    "# TODO: 修改以下变量以定义角色和角色提示词\n",
    "CHARACTER_FOR_CHATBOT = \"面试官\"  # 机器人扮演的角色，注意，真正起作用的实际是提示词，因为并没有预设 system 角色\n",
    "PROMPT_FOR_ROLEPLAY = \"我需要你面试我有关AI的知识，仅提出问题\"  # 指定角色提示词\n",
    "\n",
    "# 清除对话的函数\n",
    "def reset():\n",
    "    \"\"\"\n",
    "    清空对话记录。\n",
    "\n",
    "    返回:\n",
    "        List: 空的对话记录列表\n",
    "    \"\"\"\n",
    "    return []\n",
    "\n",
    "# 调用模型生成对话的函数\n",
    "def interact_roleplay(chatbot, user_input, temp=1.0):\n",
    "    \"\"\"\n",
    "    处理角色扮演多轮对话，调用模型生成回复。\n",
    "\n",
    "    参数:\n",
    "        chatbot (List[Tuple[str, str]]): 对话历史记录（用户与模型回复）\n",
    "        user_input (str): 当前用户输入\n",
    "        temp (float): 模型温度参数（默认 1.0）\n",
    "\n",
    "    返回:\n",
    "        List[Tuple[str, str]]: 更新后的对话记录\n",
    "    \"\"\"\n",
    "    try:\n",
    "        # 构建包含历史对话的消息列表\n",
    "        messages = []\n",
    "        for input_text, response_text in chatbot:\n",
    "            messages.append({'role': 'user', 'content': input_text})\n",
    "            messages.append({'role': 'assistant', 'content': response_text})\n",
    "        \n",
    "        # 添加当前用户输入\n",
    "        messages.append({'role': 'user', 'content': user_input})\n",
    "        \n",
    "        # 调用 API 获取回复\n",
    "        response = client.chat.completions.create(\n",
    "            model=\"qwen-turbo\",\n",
    "            messages=messages,\n",
    "            temperature=temp,\n",
    "        )\n",
    "        chatbot.append((user_input, response.choices[0].message.content))\n",
    "\n",
    "    except Exception as e:\n",
    "        print(f\"发生错误：{e}\")\n",
    "        chatbot.append((user_input, f\"抱歉，发生了错误：{e}\"))\n",
    "        \n",
    "    return chatbot\n",
    "\n",
    "def export_roleplay(chatbot, description):\n",
    "    \"\"\"\n",
    "    导出角色扮演对话记录及任务描述到 JSON 文件。\n",
    "\n",
    "    参数:\n",
    "        chatbot (List[Tuple[str, str]]): 对话记录\n",
    "        description (str): 任务描述\n",
    "    \"\"\"\n",
    "    target = {\"chatbot\": chatbot, \"description\": description}\n",
    "    with open(\"files/part2.json\", \"w\", encoding=\"utf-8\") as file:\n",
    "        json.dump(target, file, ensure_ascii=False, indent=4)\n",
    "\n",
    "# 进行第一次对话：设定角色提示\n",
    "first_dialogue = interact_roleplay([], PROMPT_FOR_ROLEPLAY)\n",
    "\n",
    "# 构建 Gradio UI 界面\n",
    "with gr.Blocks() as demo:\n",
    "    gr.Markdown(\"# 第2部分：角色扮演\\n与聊天机器人进行角色扮演互动！\")\n",
    "    chatbot = gr.Chatbot(type=\"tuples\", value=first_dialogue)\n",
    "    description_textbox = gr.Textbox(label=\"机器人扮演的角色\", interactive=False, value=CHARACTER_FOR_CHATBOT)\n",
    "    input_textbox = gr.Textbox(label=\"输入\", value=\"\")\n",
    "    \n",
    "    with gr.Column():\n",
    "        gr.Markdown(\"# 温度调节\\n温度控制聊天机器人的响应创造性。\")\n",
    "        temperature_slider = gr.Slider(0.0, 1.99, 1.0, step=0.01, label=\"温度\")\n",
    "    \n",
    "    with gr.Row():\n",
    "        send_button = gr.Button(value=\"发送\")\n",
    "        reset_button = gr.Button(value=\"重置\")\n",
    "    \n",
    "    with gr.Column():\n",
    "        gr.Markdown(\"# 保存结果\\n点击导出按钮保存对话记录。\")\n",
    "        export_button = gr.Button(value=\"导出\")\n",
    "        \n",
    "    # 绑定按钮与回调函数\n",
    "    send_button.click(interact_roleplay, inputs=[chatbot, input_textbox, temperature_slider], outputs=[chatbot])\n",
    "    reset_button.click(reset, outputs=[chatbot])\n",
    "    export_button.click(export_roleplay, inputs=[chatbot, description_textbox])\n",
    "    \n",
    "# 启动 Gradio 应用\n",
    "demo.launch(debug=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "852dee4b-c026-4afb-be0e-22feaf1953e7",
   "metadata": {},
   "source": [
    "### 检查并打印你的结果\n",
    "\n",
    "此部分用于检查你的 \"part2.json\" 文件是否包含所有正确的上下文信息，或者用于查看我们提供的示例文件。\n",
    "\n",
    "你需要：\n",
    "\n",
    "1. 确保文件列表中有你保存的 **part2.json** 文件。\n",
    "2. 点击运行按钮。它将显示一个**冻结的** Gradio 界面，重现你的摘要结果。\n",
    "\n",
    "------\n",
    "\n",
    "在运行此单元格之前，请确保已运行 **安装库** 和 **导入与设置**。\n",
    "\n",
    "**记得在进行下一步前停止此单元格。**\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "41090de2-f1e8-476f-a40f-edac95938772",
   "metadata": {},
   "outputs": [],
   "source": [
    "# 加载对话记录的 JSON 文件\n",
    "with open(\"files/part2.json\", \"r\") as f:\n",
    "    context = json.load(f)\n",
    "\n",
    "# 遍历对话记录并正确存储\n",
    "chatbot = context['chatbot']\n",
    "role = context['description']  # 机器人扮演的角色\n",
    "dialogue = \"\"  # 对话的完整记录\n",
    "for i, (user, bot) in enumerate(chatbot):\n",
    "    if i != 0:\n",
    "        dialogue += f\"用户: {user}\\n\"  # 用户输入\n",
    "    dialogue += f\"机器人: {bot}\\n\"  # 机器人回复\n",
    "\n",
    "# 构建 Gradio UI 界面\n",
    "with gr.Blocks() as demo:\n",
    "    gr.Markdown(f\"# 第2部分：角色扮演\\n聊天机器人和你玩了一个角色扮演游戏，查看结果吧！\")\n",
    "    chatbot = gr.Chatbot(type=\"tuples\", value=context['chatbot'])  # 加载之前的对话记录\n",
    "    description_textbox = gr.Textbox(label=\"机器人扮演的角色\", interactive=False, value=context['description'])  # 显示角色\n",
    "    with gr.Column():\n",
    "        gr.Markdown(\"# 只是一个检查\")\n",
    "        gr.Textbox(label=\"角色\", value=role, show_copy_button=True)  # 显示并允许复制角色信息\n",
    "        gr.Textbox(label=\"对话\", value=dialogue, show_copy_button=True)  # 显示并允许复制完整对话\n",
    "\n",
    "# 启动 Gradio 应用\n",
    "demo.launch(debug=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "197d2ec4-7353-40dd-9683-9ed9bf7122a9",
   "metadata": {},
   "source": [
    "## 第3部分：定制化任务（多轮对话应用）\n",
    "\n",
    "在本任务中，你需要提示聊天机器人执行某项特定任务。你应该首先设计一个你希望聊天机器人执行的任务，然后为它提供合适的提示词以引导其完成该任务。\n",
    "\n",
    "你需要完成以下步骤：\n",
    "\n",
    "1. 设计一个任务，并根据任务写一个提示词。将任务描述填写在 **chatbot_task** 中，并将提示词填写在 **prompt_for_task** 中。\n",
    "2. **点击运行按钮**，与聊天机器人进行互动。在 \"输入\" 框中输入你的内容，然后点击 \"发送\" 按钮。（你可以使用 \"温度\" 滑块来控制输出的创造性。）\n",
    "3. 如果你**想更改提示词或任务名称**，可以停止单元格，返回TODO重新设置，然后重新运行单元格。\n",
    "4. 在你获得满意的结果后，点击 \"导出\" 按钮保存结果。文件列表中将出现一个名为 **part3.json** 的文件。\n",
    "\n",
    "注意：\n",
    "\n",
    "- **如果你再次点击 \"导出\" 按钮，之前的结果将被覆盖。**\n",
    "- **即使使用相同的提示词，输出的结果可能仍然不同。**\n",
    "- 实际上，这和第2部分是完全一致的，只是将提前预进行的prompt延后到了用户点击发送的时候，使得更像是一个通用的范例，隐藏了角色扮演的属性，但仅仅是隐藏，本质并没有改变。\n",
    "\n",
    "---\n",
    "\n",
    "在运行此单元格之前，请确保已运行 **安装包** 和 **导入与设置**。\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "28501b1b-2218-47d2-949b-5dd922cd0194",
   "metadata": {},
   "outputs": [],
   "source": [
    "# TODO: 修改以下变量以定义任务描述与任务提示词\n",
    "CHATBOT_TASK = \"小学数学老师（输入“开始”）\"  # 用于告诉用户聊天机器人可以执行的任务\n",
    "PROMPT_FOR_TASK = \"现在开始，你将扮演一个出小学数学题的老师，当我说开始时提供一个简单的数学题，接收到正确回答后进行下一题，否则给我答案\"\n",
    "\n",
    "def reset():\n",
    "    \"\"\"\n",
    "    清空对话记录。\n",
    "\n",
    "    返回:\n",
    "        List: 空的对话记录列表\n",
    "    \"\"\"\n",
    "    return []\n",
    "\n",
    "def interact_customize(chatbot, prompt, user_input, temp=1.0):\n",
    "    \"\"\"\n",
    "    调用模型处理定制化任务对话。\n",
    "\n",
    "    参数:\n",
    "        chatbot (List[Tuple[str, str]]): 历史对话记录\n",
    "        prompt (str): 指定任务的提示词\n",
    "        user_input (str): 当前用户输入\n",
    "        temp (float): 模型温度参数（默认 1.0）\n",
    "\n",
    "    返回:\n",
    "        List[Tuple[str, str]]: 更新后的对话记录\n",
    "    \"\"\"\n",
    "    try:\n",
    "        messages = []\n",
    "        # 添加任务提示\n",
    "        messages.append({'role': 'user', 'content': prompt})\n",
    "        \n",
    "        # 构建历史对话记录\n",
    "        for input_text, response_text in chatbot:\n",
    "            messages.append({'role': 'user', 'content': input_text})\n",
    "            messages.append({'role': 'assistant', 'content': response_text})\n",
    "\n",
    "        # 添加当前用户输入\n",
    "        messages.append({'role': 'user', 'content': user_input})\n",
    "\n",
    "        response = client.chat.completions.create(\n",
    "            model=\"qwen-turbo\",\n",
    "            messages=messages,\n",
    "            temperature=temp,\n",
    "            max_tokens=200,  # 修改这个看看\n",
    "        )\n",
    "\n",
    "        chatbot.append((user_input, response.choices[0].message.content))\n",
    "\n",
    "    except Exception as e:\n",
    "        print(f\"发生错误：{e}\")\n",
    "        chatbot.append((user_input, f\"抱歉，发生了错误：{e}\"))\n",
    "    \n",
    "    return chatbot\n",
    "\n",
    "def export_customized(chatbot, description):\n",
    "    \"\"\"\n",
    "    导出定制化任务对话记录及任务描述到 JSON 文件。\n",
    "\n",
    "    参数:\n",
    "        chatbot (List[Tuple[str, str]]): 对话记录\n",
    "        description (str): 任务描述\n",
    "    \"\"\"\n",
    "    target = {\"chatbot\": chatbot, \"description\": description}\n",
    "    with open(\"files/part3.json\", \"w\", encoding=\"utf-8\") as file:\n",
    "        json.dump(target, file, ensure_ascii=False, indent=4)\n",
    "\n",
    "# 构建 Gradio UI 界面\n",
    "with gr.Blocks() as demo:\n",
    "    gr.Markdown(\"# 第3部分：定制化任务\\n聊天机器人可以执行某项任务，试着与它互动吧！\")\n",
    "    chatbot = gr.Chatbot(type=\"tuples\")\n",
    "    desc_textbox = gr.Textbox(label=\"任务描述\", value=CHATBOT_TASK, interactive=False)\n",
    "    prompt_textbox = gr.Textbox(label=\"提示词\", value=PROMPT_FOR_TASK, visible=False)\n",
    "    input_textbox = gr.Textbox(label=\"输入\", value=\"\")\n",
    "    \n",
    "    with gr.Column():\n",
    "        gr.Markdown(\"# 温度调节\\n温度用于控制聊天机器人的输出，温度越高响应越具创造性。\")\n",
    "        temperature_slider = gr.Slider(0.0, 1.99, 1.0, step=0.01, label=\"温度\")\n",
    "    \n",
    "    with gr.Row():\n",
    "        send_button = gr.Button(value=\"发送\")\n",
    "        reset_button = gr.Button(value=\"重置\")\n",
    "    \n",
    "    with gr.Column():\n",
    "        gr.Markdown(\"# 保存结果\\n当你对结果满意后，点击导出按钮保存结果。\")\n",
    "        export_button = gr.Button(value=\"导出\")\n",
    "    \n",
    "    # 绑定按钮与回调函数\n",
    "    send_button.click(interact_customize, inputs=[chatbot, prompt_textbox, input_textbox, temperature_slider], outputs=[chatbot])\n",
    "    reset_button.click(reset, outputs=[chatbot])\n",
    "    export_button.click(export_customized, inputs=[chatbot, desc_textbox])\n",
    "    \n",
    "# 启动 Gradio 应用\n",
    "demo.launch(debug=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a8173473-5912-46cd-9d2f-99389289ed7a",
   "metadata": {},
   "source": [
    "### 检查并打印你的结果\n",
    "\n",
    "此部分用于检查你的 \"part3.json\" 文件是否包含所有正确的上下文信息，或者用于查看我们提供的示例文件。\n",
    "\n",
    "你需要：\n",
    "\n",
    "1. 确保文件列表中有你保存的 **part3.json** 文件。\n",
    "2. 点击运行按钮。它将显示一个**冻结的** Gradio 界面，重现你的摘要结果。\n",
    "\n",
    "------\n",
    "\n",
    "在运行此单元格之前，请确保已运行 **安装库** 和 **导入与设置**。\n",
    "\n",
    "**记得在进行下一步前停止此单元格。**\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "29bb52a9-2a22-49c9-95db-7a6124fc54ee",
   "metadata": {},
   "outputs": [],
   "source": [
    "# 加载对话记录的 JSON 文件\n",
    "with open(\"files/part3.json\", \"r\") as f:\n",
    "    context = json.load(f)\n",
    "\n",
    "# 遍历并加载对话记录\n",
    "chatbot = context['chatbot']\n",
    "desc = context['description']\n",
    "dialogue = \"\"\n",
    "for user, bot in chatbot:\n",
    "    dialogue += f\"用户: {user}\\n\"\n",
    "    dialogue += f\"机器人: {bot}\\n\"\n",
    "\n",
    "# 构建 Gradio UI 界面\n",
    "with gr.Blocks() as demo:\n",
    "    gr.Markdown(\"# 第3部分：定制化任务\\n聊天机器人可以执行特定任务，试着与它互动吧！\")\n",
    "    chatbot = gr.Chatbot(type=\"tuples\", value=context['chatbot'])  # 显示聊天记录\n",
    "    desc_textbox = gr.Textbox(label=\"任务描述\", value=context['description'], interactive=False)  # 显示任务描述\n",
    "    \n",
    "    with gr.Column():\n",
    "        gr.Markdown(\"# 只是一个检查\")\n",
    "        gr.Textbox(label=\"任务描述\", value=desc, show_copy_button=True)  # 显示并允许复制任务描述\n",
    "        gr.Textbox(label=\"对话记录\", value=dialogue, show_copy_button=True)  # 显示并允许复制对话记录\n",
    "\n",
    "# 启动 Gradio 应用\n",
    "demo.launch(debug=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "aec5293d-0c75-4bee-b7d4-72f1dc08788e",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "ai",
   "language": "python",
   "name": "ai"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.12.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}