{ "cells": [ { "cell_type": "markdown", "metadata": {}, "source": [ "# 构建第一个标准化患者智能体 —— API编程实操\n", "\n", "> **第二章 第五节 实操课** | 魔塔平台CPU环境\n", "\n", "本Notebook包含三个部分：\n", "1. **Part 1**: 单次API调用测试 —— 验证工作流连通性\n", "2. **Part 2**: 10次单轮对话 —— 自动化护理评估问答，保存到 `result.txt`\n", "3. **Part 3**: 多轮对话 —— 模拟真实问诊，AI能记住上下文\n", "\n", "---" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## 环境准备\n", "\n", "安装必要的Python库（魔塔平台只需安装 `requests`）" ] }, { "cell_type": "code", "execution_count": 1, "metadata": { "execution": { "iopub.execute_input": "2026-04-10T05:14:06.267453Z", "iopub.status.busy": "2026-04-10T05:14:06.267287Z", "iopub.status.idle": "2026-04-10T05:14:09.272856Z", "shell.execute_reply": "2026-04-10T05:14:09.272088Z", "shell.execute_reply.started": "2026-04-10T05:14:06.267437Z" }, "tags": [] }, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "\u001b[33mWARNING: Running pip as the 'root' user can result in broken permissions and conflicting behaviour with the system package manager. It is recommended to use a virtual environment instead: https://pip.pypa.io/warnings/venv\u001b[0m\u001b[33m\n", "\u001b[0m\n", "\u001b[1m[\u001b[0m\u001b[34;49mnotice\u001b[0m\u001b[1;39;49m]\u001b[0m\u001b[39;49m A new release of pip is available: \u001b[0m\u001b[31;49m23.3.2\u001b[0m\u001b[39;49m -> \u001b[0m\u001b[32;49m26.0.1\u001b[0m\n", "\u001b[1m[\u001b[0m\u001b[34;49mnotice\u001b[0m\u001b[1;39;49m]\u001b[0m\u001b[39;49m To update, run: \u001b[0m\u001b[32;49mpip install --upgrade pip\u001b[0m\n" ] } ], "source": [ "!pip install requests -q" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## 配置信息\n", "\n", "**请将下面的配置替换为你自己的信息：**\n", "- `API_KEY`：在 https://www.coze.cn/open/oauth/pats 创建的个人访问令牌\n", "- `WORKFLOW_ID`：你的工作流URL中的 `workflow_id` 参数\n", "\n", "> 示例URL：`https://www.coze.cn/work_flow?workflow_id=7610xxx92420&space_id=74827xxxx99773491`\n", "> 其中 `76189xxx92420` 就是你的 workflow_id" ] }, { "cell_type": "code", "execution_count": 1, "metadata": { "ExecutionIndicator": { "show": false }, "execution": { "iopub.execute_input": "2026-04-10T07:01:13.196234Z", "iopub.status.busy": "2026-04-10T07:01:13.196062Z", "iopub.status.idle": "2026-04-10T07:01:13.234152Z", "shell.execute_reply": "2026-04-10T07:01:13.233621Z", "shell.execute_reply.started": "2026-04-10T07:01:13.196217Z" }, "tags": [] }, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "配置完成！\n", "API地址: https://api.coze.cn/v1/workflow/run\n", "工作流ID: 76267776xxx24089902\n", "API Key: pat_f76JZa8qYHtK57Ex...（已隐藏）\n" ] } ], "source": [ "import requests\n", "import json\n", "import time\n", "from datetime import datetime\n", "\n", "# ============================================================\n", "# 请修改以下配置为你自己的信息\n", "# ============================================================\n", "API_KEY = \"pat_f76JZa8qYHtK57ExxxxxtUkDguwkcNW6fNzvNfS2aV4\"\n", "WORKFLOW_ID = \"76267776xxx24089902\"\n", "# ============================================================\n", "\n", "API_URL = \"https://api.coze.cn/v1/workflow/run\"\n", "HEADERS = {\n", " \"Authorization\": f\"Bearer {API_KEY}\",\n", " \"Content-Type\": \"application/json\"\n", "}\n", "\n", "print(\"配置完成！\")\n", "print(f\"API地址: {API_URL}\")\n", "print(f\"工作流ID: {WORKFLOW_ID}\")\n", "print(f\"API Key: {API_KEY[:20]}...（已隐藏）\")" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "---\n", "\n", "## Part 1：单次调用测试\n", "\n", "先发一个问题，确认工作流能正常工作。" ] }, { "cell_type": "code", "execution_count": 3, "metadata": { "execution": { "iopub.execute_input": "2026-04-10T05:14:13.393584Z", "iopub.status.busy": "2026-04-10T05:14:13.393411Z", "iopub.status.idle": "2026-04-10T05:14:17.119718Z", "shell.execute_reply": "2026-04-10T05:14:17.118991Z", "shell.execute_reply.started": "2026-04-10T05:14:13.393567Z" }, "tags": [] }, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "护士：你好，请问你叫什么名字？\n", "患者：……李文清。（手指绞着衣角）\n", "\n", "护士：您平时吃什么药？\n", "患者：（挠挠头）有个叫氨氯地平的降压药，还有个治糖尿病的二甲双胍。\n" ] } ], "source": [ "def call_workflow(question):\n", " \"\"\"调用Coze工作流，发送一个问题并返回回答\"\"\"\n", " # 包装输入：明确告诉LLM这是护士的问题，必须回答\n", " wrapped_input = (\n", " f\"护士对你说：'{question}'\\n\"\n", " f\"请针对护士的这个问题回答，1-3句话，不要回答其他内容。\"\n", " )\n", " \n", " payload = {\n", " \"workflow_id\": WORKFLOW_ID,\n", " \"parameters\": {\n", " \"USER_INPUT\": wrapped_input\n", " }\n", " }\n", " \n", " try:\n", " response = requests.post(API_URL, headers=HEADERS, json=payload, timeout=60)\n", " result = response.json()\n", " \n", " if result.get(\"code\") == 0:\n", " parsed = json.loads(result[\"data\"])\n", " return parsed.get(\"data\") or parsed.get(\"result\") or parsed.get(\"output\") or str(parsed)\n", " else:\n", " return f\"[错误] code={result.get('code')}, msg={result.get('msg')}\"\n", " except requests.exceptions.Timeout:\n", " return \"[错误] 请求超时，请重试\"\n", " except Exception as e:\n", " return f\"[错误] {str(e)}\"\n", "\n", "\n", "# 测试一次调用\n", "test_question = \"你好，请问你叫什么名字？\"\n", "print(f\"护士：{test_question}\")\n", "print(f\"患者：{call_workflow(test_question)}\")\n", "print()\n", "test_question2 = \"您平时吃什么药？\"\n", "print(f\"护士：{test_question2}\")\n", "print(f\"患者：{call_workflow(test_question2)}\")" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "**检查点**：\n", "- 如果看到患者的回答 → 工作流连通成功，继续！\n", "- 如果看到 `[错误]` → 检查API Key和workflow_id是否正确，工作流是否已发布" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "---\n", "\n", "## Part 2：10次单轮对话\n", "\n", "用10个护理评估常见问题，逐一调用工作流，收集AI患者的回答。\n", "\n", "> **注意**：每次调用是独立的，AI不会记得前一次对话的内容。" ] }, { "cell_type": "code", "execution_count": 4, "metadata": { "execution": { "iopub.execute_input": "2026-04-10T05:14:20.413864Z", "iopub.status.busy": "2026-04-10T05:14:20.413703Z", "iopub.status.idle": "2026-04-10T05:14:20.417285Z", "shell.execute_reply": "2026-04-10T05:14:20.416667Z", "shell.execute_reply.started": "2026-04-10T05:14:20.413848Z" }, "tags": [] }, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "共 10 个问题，开始逐一调用...\n", "\n" ] } ], "source": [ "# 准备10个护理评估问题\n", "questions = [\n", " \"你好，我是你的责任护士小王，请问你叫什么名字？\",\n", " \"您今年多大了？平时一个人住吗？\",\n", " \"您来医院是看什么病的？\",\n", " \"您平时吃什么药？\",\n", " \"这些药一天吃几次？有没有什么副作用？\",\n", " \"有没有忘记吃药的时候？大概多久忘一次？\",\n", " \"您平时自己量血压吗？最近血压多少？\",\n", " \"最近有没有头晕、乏力或者其他不舒服？\",\n", " \"家里人知道您的病情吗？他们会帮您管药吗？\",\n", " \"您对自己的健康状况担心吗？最担心什么？\"\n", "]\n", "\n", "print(f\"共 {len(questions)} 个问题，开始逐一调用...\\n\")" ] }, { "cell_type": "code", "execution_count": 5, "metadata": { "execution": { "iopub.execute_input": "2026-04-10T05:14:23.674199Z", "iopub.status.busy": "2026-04-10T05:14:23.674040Z", "iopub.status.idle": "2026-04-10T05:14:50.524251Z", "shell.execute_reply": "2026-04-10T05:14:50.523548Z", "shell.execute_reply.started": "2026-04-10T05:14:23.674183Z" }, "tags": [] }, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "--- 第 1/10 轮 ---\n", "护士：你好，我是你的责任护士小王，请问你叫什么名字？\n", "患者：……李文清。（指尖轻轻摩挲着病号服衣角）\n", "\n", "--- 第 2/10 轮 ---\n", "护士：您今年多大了？平时一个人住吗？\n", "患者：我七十了。（低头）平时一个人住，孩子们都在外地。\n", "\n", "--- 第 3/10 轮 ---\n", "护士：您来医院是看什么病的？\n", "患者：（搓了搓衣角）我血压和血糖都不稳，过来调调。\n", "\n", "--- 第 4/10 轮 ---\n", "护士：您平时吃什么药？\n", "患者：吃氨氯地平，还有二甲双胍。（挠挠后颈）\n", "\n", "--- 第 5/10 轮 ---\n", "护士：这些药一天吃几次？有没有什么副作用？\n", "患者：氨氯地平一天一次，二甲双胍一天两次。（挠挠后颈）副作用倒是没太留意，就是偶尔有点头晕。\n", "\n", "--- 第 6/10 轮 ---\n", "护士：有没有忘记吃药的时候？大概多久忘一次？\n", "患者：……有时候忘，（挠挠后颈）大概两三天就会漏一次晚上那顿降糖药。\n", "\n", "--- 第 7/10 轮 ---\n", "护士：您平时自己量血压吗？最近血压多少？\n", "患者：（挠挠头）有时候量……最近一次好像是150/90？记不太准了。\n", "\n", "--- 第 8/10 轮 ---\n", "护士：最近有没有头晕、乏力或者其他不舒服？\n", "患者：（揉了揉后颈）这两天早上起来头有点发懵，腿也软乎乎的……没敢跟孩子说。\n", "\n", "--- 第 9/10 轮 ---\n", "护士：家里人知道您的病情吗？他们会帮您管药吗？\n", "患者：（低头抠衣角）知道是知道……孩子们忙，哪好意思让他们天天盯着我吃药啊。\n", "\n", "--- 第 10/10 轮 ---\n", "护士：您对自己的健康状况担心吗？最担心什么？\n", "患者：（叹气）挺担心的……就怕病情控制不好，给远在外地的儿女添麻烦。\n", "\n", "==================================================\n", "全部完成！共 10 轮对话\n" ] } ], "source": [ "# 循环调用工作流，收集回答\n", "results = []\n", "\n", "for i, question in enumerate(questions, 1):\n", " print(f\"--- 第 {i}/10 轮 ---\")\n", " print(f\"护士：{question}\")\n", " \n", " answer = call_workflow(question)\n", " print(f\"患者：{answer}\")\n", " print()\n", " \n", " results.append({\n", " \"round\": i,\n", " \"question\": question,\n", " \"answer\": answer\n", " })\n", " \n", " # 间隔1秒，避免请求过快\n", " if i < len(questions):\n", " time.sleep(1)\n", "\n", "print(\"=\"*50)\n", "print(f\"全部完成！共 {len(results)} 轮对话\")" ] }, { "cell_type": "code", "execution_count": 6, "metadata": { "execution": { "iopub.execute_input": "2026-04-10T05:14:57.968976Z", "iopub.status.busy": "2026-04-10T05:14:57.968819Z", "iopub.status.idle": "2026-04-10T05:14:57.987799Z", "shell.execute_reply": "2026-04-10T05:14:57.987335Z", "shell.execute_reply.started": "2026-04-10T05:14:57.968960Z" }, "tags": [] }, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "结果已保存到 result.txt\n", "\n", "--- 文件内容预览 ---\n", "\n", "============================================================\n", "标准化患者智能体 —— 单轮对话记录\n", "时间：2026-04-10 13:14:57\n", "工作流ID：7626777633124089902\n", "============================================================\n", "\n", "--- 第 1 轮对话 ---\n", "护士：你好，我是你的责任护士小王，请问你叫什么名字？\n", "患者：……李文清。（指尖轻轻摩挲着病号服衣角）\n", "\n", "--- 第 2 轮对话 ---\n", "护士：您今年多大了？平时一个人住吗？\n", "患者：我七十了。（低头）平时一个人住，孩子们都在外地。\n", "\n", "--- 第 3 轮对话 ---\n", "护士：您来医院是看什么病的？\n", "患者：（搓了搓衣角）我血压和血糖都不稳，过来调调。\n", "\n", "--- 第 4 轮对话 ---\n", "护士：您平时吃什么药？\n", "患者：吃氨氯地平，还有二甲双胍。（挠挠后颈）\n", "\n", "--- 第 5 轮对话 ---\n", "护士：这些药一天吃几次？有没有什么副作用？\n", "患者：氨氯地平一天一次，二甲双胍一天两次。（挠挠后颈）副作用倒是没太留意，就是偶尔有点头晕。\n", "\n", "--- 第 6 轮对话 ---\n", "护士：有没有忘记吃药的时候？大概多久忘一次？\n", "患者：……有时候忘，（挠挠后颈）大概两三天就会漏一次晚上那顿降糖药。\n", "\n", "--- 第 7 轮对话 ---\n", "护士：您平时自己量血压吗？最近血压多少？\n", "患者：（挠挠头）有时候量……最近一次好像是150/90？记不太准了。\n", "\n", "--- 第 8 轮对话 ---\n", "护士：最近有没有头晕、乏力或者其他不舒服？\n", "患者：（揉了揉后颈）这两天早上起来头有点发懵，腿也软乎乎的……没敢跟孩子说。\n", "\n", "--- 第 9 轮对话 ---\n", "护士：家里人知道您的病情吗？他们会帮您管药吗？\n", "患者：（低头抠衣角）知道是知道……孩子们忙，哪好意思让他们天天盯着我吃药啊。\n", "\n", "--- 第 10 轮对话 ---\n", "护士：您对自己的健康状况担心吗？最担心什么？\n", "患者：（叹气）挺担心的……就怕病情控制不好，给远在外地的儿女添麻烦。\n", "\n", "============================================================\n", "记录结束\n", "\n" ] } ], "source": [ "# 保存结果到 result.txt\n", "timestamp = datetime.now().strftime(\"%Y-%m-%d %H:%M:%S\")\n", "\n", "with open(\"result.txt\", \"w\", encoding=\"utf-8\") as f:\n", " f.write(\"=\" * 60 + \"\\n\")\n", " f.write(\"标准化患者智能体 —— 单轮对话记录\\n\")\n", " f.write(f\"时间：{timestamp}\\n\")\n", " f.write(f\"工作流ID：{WORKFLOW_ID}\\n\")\n", " f.write(\"=\" * 60 + \"\\n\\n\")\n", " \n", " for item in results:\n", " f.write(f\"--- 第 {item['round']} 轮对话 ---\\n\")\n", " f.write(f\"护士：{item['question']}\\n\")\n", " f.write(f\"患者：{item['answer']}\\n\\n\")\n", " \n", " f.write(\"=\" * 60 + \"\\n\")\n", " f.write(\"记录结束\\n\")\n", "\n", "print(\"结果已保存到 result.txt\")\n", "print(\"\\n--- 文件内容预览 ---\\n\")\n", "\n", "with open(\"result.txt\", \"r\", encoding=\"utf-8\") as f:\n", " print(f.read())" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "---\n", "\n", "## Part 3：多轮对话 —— 模拟真实问诊\n", "\n", "### 原理说明\n", "\n", "单轮对话中，每次调用都是独立的，AI不知道之前说了什么。\n", "\n", "多轮对话的核心思路：**每次调用时，把之前的对话历史一起发送给AI**。\n", "\n", "```\n", "第1轮输入: \"护士：你吃什么药？\"\n", "第1轮输出: \"降压药和二甲双胍\"\n", "\n", "第2轮输入: \"对话历史：\\n护士：你吃什么药？\\n患者：降压药和二甲双胍\\n---\\n护士：二甲双胍一天吃几次？\"\n", "第2轮输出: \"医生说一天两次……\"\n", "```\n", "\n", "这样AI就能\"看到\"之前的对话，实现上下文连贯。" ] }, { "cell_type": "code", "execution_count": 7, "metadata": { "execution": { "iopub.execute_input": "2026-04-10T05:15:03.054970Z", "iopub.status.busy": "2026-04-10T05:15:03.054812Z", "iopub.status.idle": "2026-04-10T05:15:03.060600Z", "shell.execute_reply": "2026-04-10T05:15:03.060053Z", "shell.execute_reply.started": "2026-04-10T05:15:03.054954Z" }, "tags": [] }, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "多轮对话函数已定义！\n" ] } ], "source": [ "def call_workflow_with_history(question, history):\n", " \"\"\"带对话历史的工作流调用（多轮对话）\n", " \n", " 参数：\n", " question: 当前护士的提问\n", " history: 之前的对话历史列表 [{\"role\":\"护士\",\"content\":\"...\"}, {\"role\":\"患者\",\"content\":\"...\"}]\n", " 返回：\n", " AI患者的回答\n", " \"\"\"\n", " if history:\n", " # 构建结构化输入：历史 + 新问题 + 明确指令\n", " lines = []\n", " lines.append(\"你是患者李文清。以下是你刚才和护士的对话：\")\n", " lines.append(\"\")\n", " for msg in history:\n", " lines.append(f\"{msg['role']}：{msg['content']}\")\n", " lines.append(\"\")\n", " lines.append(f\"护士现在又问你：'{question}'\")\n", " lines.append(\"\")\n", " lines.append(\"请针对护士的这个新问题回答，1-3句话。\")\n", " lines.append(\"如果之前提到过相关信息（如药名），请保持一致。\")\n", " full_input = \"\\n\".join(lines)\n", " else:\n", " # 第一轮：和单轮一样包装\n", " full_input = (\n", " f\"护士对你说：'{question}'\\n\"\n", " f\"请针对护士的这个问题回答，1-3句话，不要回答其他内容。\"\n", " )\n", " \n", " # 直接调用工作流\n", " payload = {\n", " \"workflow_id\": WORKFLOW_ID,\n", " \"parameters\": {\"USER_INPUT\": full_input}\n", " }\n", " try:\n", " response = requests.post(API_URL, headers=HEADERS, json=payload, timeout=60)\n", " result = response.json()\n", " if result.get(\"code\") == 0:\n", " parsed = json.loads(result[\"data\"])\n", " return parsed.get(\"data\") or parsed.get(\"result\") or parsed.get(\"output\") or str(parsed)\n", " else:\n", " return f\"[错误] code={result.get('code')}, msg={result.get('msg')}\"\n", " except Exception as e:\n", " return f\"[错误] {str(e)}\"\n", "\n", "\n", "print(\"多轮对话函数已定义！\")" ] }, { "cell_type": "code", "execution_count": 8, "metadata": { "execution": { "iopub.execute_input": "2026-04-10T05:15:06.014106Z", "iopub.status.busy": "2026-04-10T05:15:06.013921Z", "iopub.status.idle": "2026-04-10T05:15:23.489874Z", "shell.execute_reply": "2026-04-10T05:15:23.489345Z", "shell.execute_reply.started": "2026-04-10T05:15:06.014089Z" }, "tags": [] }, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "开始多轮对话（用药问诊场景）\n", "==================================================\n", "\n", "--- 第 1 轮 ---\n", "护士：你好李文清，我是护士小王。你平时在吃什么药啊？\n", "患者：吃氨氯地平降压，还有二甲双胍治糖尿病。（挠了挠后颈）\n", "\n", "--- 第 2 轮 ---\n", "护士：这个降压药叫什么名字，你记得吗？\n", "患者：氨氯地平，医生开的那个降压药。（指尖蹭了蹭衣角）\n", "\n", "--- 第 3 轮 ---\n", "护士：那糖尿病的药呢，叫什么？一天吃几次？\n", "患者：叫二甲双胍，一天吃两次。（低头抠了抠裤腿）\n", "\n", "--- 第 4 轮 ---\n", "护士：你刚才说的那个糖尿病的药，吃了有没有什么不舒服？\n", "患者：（叹气）没啥特别不舒服的……就是有时候胃里有点胀，忍忍就过去了。\n", "\n", "--- 第 5 轮 ---\n", "护士：那降压药呢，吃了之后感觉怎么样？有没有头晕？\n", "患者：（抿了抿嘴）降压药吃着还行……偶尔会有点头沉，歇会儿就好了。\n", "\n", "--- 第 6 轮 ---\n", "护士：你有没有忘记吃药的时候？一般什么时候容易忘？\n", "患者：……有时候会忘。（低头）晚上那顿治糖尿病的药，经常记不起来。\n", "\n", "--- 第 7 轮 ---\n", "护士：你刚才提到的那两种药，你能再说一遍名字吗？\n", "患者：氨氯地平，还有二甲双胍。（抠了抠手背）\n", "\n", "==================================================\n", "多轮对话完成！共 7 轮\n" ] } ], "source": [ "# 多轮对话：用药问诊场景\n", "multi_turn_questions = [\n", " \"你好李文清，我是护士小王。你平时在吃什么药啊？\",\n", " \"这个降压药叫什么名字，你记得吗？\",\n", " \"那糖尿病的药呢，叫什么？一天吃几次？\",\n", " \"你刚才说的那个糖尿病的药，吃了有没有什么不舒服？\",\n", " \"那降压药呢，吃了之后感觉怎么样？有没有头晕？\",\n", " \"你有没有忘记吃药的时候？一般什么时候容易忘？\",\n", " \"你刚才提到的那两种药，你能再说一遍名字吗？\"\n", "]\n", "\n", "# 执行多轮对话\n", "history = []\n", "multi_turn_results = []\n", "\n", "print(\"开始多轮对话（用药问诊场景）\")\n", "print(\"=\" * 50)\n", "\n", "for i, question in enumerate(multi_turn_questions, 1):\n", " print(f\"\\n--- 第 {i} 轮 ---\")\n", " print(f\"护士：{question}\")\n", " \n", " # 调用带历史的工作流\n", " answer = call_workflow_with_history(question, history)\n", " print(f\"患者：{answer}\")\n", " \n", " # 更新对话历史\n", " history.append({\"role\": \"护士\", \"content\": question})\n", " history.append({\"role\": \"患者\", \"content\": answer})\n", " \n", " multi_turn_results.append({\n", " \"round\": i,\n", " \"question\": question,\n", " \"answer\": answer\n", " })\n", " \n", " # 间隔1秒\n", " if i < len(multi_turn_questions):\n", " time.sleep(1)\n", "\n", "print(\"\\n\" + \"=\" * 50)\n", "print(f\"多轮对话完成！共 {len(multi_turn_results)} 轮\")" ] }, { "cell_type": "code", "execution_count": 9, "metadata": { "execution": { "iopub.execute_input": "2026-04-10T05:15:25.415468Z", "iopub.status.busy": "2026-04-10T05:15:25.415300Z", "iopub.status.idle": "2026-04-10T05:15:25.432513Z", "shell.execute_reply": "2026-04-10T05:15:25.431826Z", "shell.execute_reply.started": "2026-04-10T05:15:25.415452Z" }, "tags": [] }, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "多轮对话结果已保存到 result_multi_turn.txt\n", "\n", "--- 文件内容预览 ---\n", "\n", "============================================================\n", "标准化患者智能体 —— 多轮对话记录（用药问诊）\n", "时间：2026-04-10 13:15:25\n", "工作流ID：7626777633124089902\n", "============================================================\n", "\n", "--- 第 1 轮对话 ---\n", "护士：你好李文清，我是护士小王。你平时在吃什么药啊？\n", "患者：吃氨氯地平降压，还有二甲双胍治糖尿病。（挠了挠后颈）\n", "\n", "--- 第 2 轮对话 ---\n", "护士：这个降压药叫什么名字，你记得吗？\n", "患者：氨氯地平，医生开的那个降压药。（指尖蹭了蹭衣角）\n", "\n", "--- 第 3 轮对话 ---\n", "护士：那糖尿病的药呢，叫什么？一天吃几次？\n", "患者：叫二甲双胍，一天吃两次。（低头抠了抠裤腿）\n", "\n", "--- 第 4 轮对话 ---\n", "护士：你刚才说的那个糖尿病的药，吃了有没有什么不舒服？\n", "患者：（叹气）没啥特别不舒服的……就是有时候胃里有点胀，忍忍就过去了。\n", "\n", "--- 第 5 轮对话 ---\n", "护士：那降压药呢，吃了之后感觉怎么样？有没有头晕？\n", "患者：（抿了抿嘴）降压药吃着还行……偶尔会有点头沉，歇会儿就好了。\n", "\n", "--- 第 6 轮对话 ---\n", "护士：你有没有忘记吃药的时候？一般什么时候容易忘？\n", "患者：……有时候会忘。（低头）晚上那顿治糖尿病的药，经常记不起来。\n", "\n", "--- 第 7 轮对话 ---\n", "护士：你刚才提到的那两种药，你能再说一遍名字吗？\n", "患者：氨氯地平，还有二甲双胍。（抠了抠手背）\n", "\n", "============================================================\n", "\n", "观察要点：\n", "1. 第1轮提到的药物名称，后面几轮是否保持一致？\n", "2. 第4轮提到'刚才说的药'，AI能否正确关联？\n", "3. 第7轮要求复述药名，AI是否记得之前说过什么？\n", "4. 整体对话是否自然连贯，像真实问诊？\n", "\n" ] } ], "source": [ "# 保存多轮对话结果\n", "timestamp = datetime.now().strftime(\"%Y-%m-%d %H:%M:%S\")\n", "\n", "with open(\"result_multi_turn.txt\", \"w\", encoding=\"utf-8\") as f:\n", " f.write(\"=\" * 60 + \"\\n\")\n", " f.write(\"标准化患者智能体 —— 多轮对话记录（用药问诊）\\n\")\n", " f.write(f\"时间：{timestamp}\\n\")\n", " f.write(f\"工作流ID：{WORKFLOW_ID}\\n\")\n", " f.write(\"=\" * 60 + \"\\n\\n\")\n", " \n", " for item in multi_turn_results:\n", " f.write(f\"--- 第 {item['round']} 轮对话 ---\\n\")\n", " f.write(f\"护士：{item['question']}\\n\")\n", " f.write(f\"患者：{item['answer']}\\n\\n\")\n", " \n", " f.write(\"=\" * 60 + \"\\n\")\n", " f.write(\"\\n观察要点：\\n\")\n", " f.write(\"1. 第1轮提到的药物名称，后面几轮是否保持一致？\\n\")\n", " f.write(\"2. 第4轮提到'刚才说的药'，AI能否正确关联？\\n\")\n", " f.write(\"3. 第7轮要求复述药名，AI是否记得之前说过什么？\\n\")\n", " f.write(\"4. 整体对话是否自然连贯，像真实问诊？\\n\")\n", "\n", "print(\"多轮对话结果已保存到 result_multi_turn.txt\")\n", "print(\"\\n--- 文件内容预览 ---\\n\")\n", "\n", "with open(\"result_multi_turn.txt\", \"r\", encoding=\"utf-8\") as f:\n", " print(f.read())" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "---\n", "\n", "## 拓展：自由多轮对话（交互式）\n", "\n", "下面的代码让你可以像聊天一样，手动输入问题与AI患者对话。\n", "\n", "> 输入 `quit` 或 `退出` 结束对话。" ] }, { "cell_type": "code", "execution_count": 10, "metadata": { "execution": { "iopub.execute_input": "2026-04-10T05:15:30.042214Z", "iopub.status.busy": "2026-04-10T05:15:30.042047Z", "iopub.status.idle": "2026-04-10T05:15:42.930646Z", "shell.execute_reply": "2026-04-10T05:15:42.930071Z", "shell.execute_reply.started": "2026-04-10T05:15:30.042198Z" }, "tags": [] }, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "==================================================\n", "交互式多轮对话 —— 你扮演护士，AI扮演患者\n", "输入 quit 或 退出 结束对话\n", "==================================================\n" ] }, { "name": "stdin", "output_type": "stream", "text": [ "\n", "护士： 你好\n" ] }, { "name": "stdout", "output_type": "stream", "text": [ "患者：你好……（抬手挠了挠后颈）\n" ] }, { "name": "stdin", "output_type": "stream", "text": [ "\n", "护士： quit\n" ] }, { "name": "stdout", "output_type": "stream", "text": [ "\n", "对话结束。\n", "\n", "共进行了 1 轮对话\n" ] } ], "source": [ "# 交互式多轮对话\n", "print(\"=\" * 50)\n", "print(\"交互式多轮对话 —— 你扮演护士，AI扮演患者\")\n", "print(\"输入 quit 或 退出 结束对话\")\n", "print(\"=\" * 50)\n", "\n", "interactive_history = []\n", "round_num = 0\n", "\n", "while True:\n", " user_input = input(\"\\n护士：\").strip()\n", " \n", " if user_input.lower() in [\"quit\", \"exit\", \"退出\", \"q\"]:\n", " print(\"\\n对话结束。\")\n", " break\n", " \n", " if not user_input:\n", " print(\"（请输入你的问题）\")\n", " continue\n", " \n", " round_num += 1\n", " answer = call_workflow_with_history(user_input, interactive_history)\n", " print(f\"患者：{answer}\")\n", " \n", " interactive_history.append({\"role\": \"护士\", \"content\": user_input})\n", " interactive_history.append({\"role\": \"患者\", \"content\": answer})\n", "\n", "print(f\"\\n共进行了 {round_num} 轮对话\")" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "---\n", "\n", "## 实操完成！\n", "\n", "### 你今天学会了：\n", "1. 通过API调用Coze工作流（`requests.post`）\n", "2. 批量自动化对话并保存结果\n", "3. 通过拼接对话历史实现多轮对话\n", "\n", "### 生成的文件：\n", "- `result.txt` —— 10次单轮对话记录\n", "- `result_multi_turn.txt` —— 多轮用药问诊记录\n", "\n", "### 下一步：\n", "- 修改Prompt创建不同人格的患者\n", "- 设计更多问诊场景的多轮对话\n", "- 第三章将学习人格一致性评价方法" ] } ], "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": "3.11.11" } }, "nbformat": 4, "nbformat_minor": 4 }