commit 27fba7a7cca19bc7c607a01f3d64d5e5ca4654a7
Author: ZhuJW <1421267742@qq.com>
Date:   Thu Mar 12 13:27:03 2026 +0800

    第一次提交

diff --git a/README.md b/README.md
new file mode 100644
index 0000000..9a9746f
--- /dev/null
+++ b/README.md
@@ -0,0 +1,285 @@
+# AutoGen SDLC 多智能体协同系统
+
+基于 **Microsoft AutoGen** + **阿里云 Qwen3.5-flash** 的端到端软件交付协同网络。
+
+## 🎯 项目概述
+
+本系统实现了多智能体协同的完整 SDLC（软件开发生命周期）流程：
+
+1. **PM Agent** - 产品经理：需求分析与 SRS 生成
+2. **QA Agent** - 测试工程师：测试用例设计与 TDD 实践
+3. **Dev Agent** - 开发工程师：代码实现与修复
+4. **Orchestrator Agent** - 协调器：流程调度与最终验证
+
+## 🚀 快速启动
+
+### 1. 环境准备
+
+```bash
+# Python 版本要求
+Python >= 3.9
+
+# 安装依赖
+pip install -r requirements.txt
+```
+
+### 2. 配置 API Key
+
+```bash
+# Windows (PowerShell)
+$env:DASHSCOPE_API_KEY="your_api_key_here"
+
+# Linux/Mac
+export DASHSCOPE_API_KEY="your_api_key_here"
+```
+
+获取 API Key: [阿里云 DashScope 控制台](https://dashscope.console.aliyun.com/)
+
+### 3. 运行演示
+
+#### 方式 1: 命令行模式
+
+```bash
+python autogen_sdls_system.py
+```
+
+#### 方式 2: Streamlit 前端界面
+
+```bash
+streamlit run frontend/streamlit_app.py
+```
+
+#### 方式 3: 自愈功能演示
+
+```bash
+python autogen_self_healing_demo.py
+```
+
+## 📁 项目结构
+
+```
+project/
+├── agents/                 # Agent 模块
+│   ├── pm_agent.py        # 产品经理 Agent
+│   ├── qa_agent.py        # 测试工程师 Agent
+│   ├── dev_agent.py       # 开发工程师 Agent
+│   └── orchestrator.py    # 协调器 Agent
+├── config/                # 配置文件
+│   └── llm_config.py     # LLM 配置与提示词模板
+├── frontend/             # 前端界面
+│   └── streamlit_app.py # Streamlit 实时聊天界面
+├── utils/                # 工具类
+│   ├── logger.py        # 日志记录
+│   └── callback_handler.py # 回调处理
+├── tests/                # 单元测试
+│   └── test_agents.py
+├── workspace/            # 工作目录（Agent 输出文件）
+├── logs/                 # 日志目录
+├── autogen_sdls_system.py      # 主程序入口
+├── autogen_self_healing_demo.py # 自愈演示
+├── usage_examples.py     # 使用示例
+└── requirements.txt      # 依赖包列表
+```
+
+## 💡 核心功能
+
+### 1. 端到端自动化流程
+
+```
+用户需求 → PM 分析 → QA 设计测试 → Dev 编码 → 自动测试 → 验证输出
+```
+
+### 2. TDD（测试驱动开发）
+
+- QA Agent 先于代码生成测试用例
+- Dev Agent 编写代码确保测试通过
+- 测试失败时触发自动修复循环
+
+### 3. 自愈能力
+
+系统检测到测试失败时，自动：
+1. 分析错误日志
+2. 定位 bug 原因
+3. 生成修复代码
+4. 重新验证测试
+
+### 4. 人机协同
+
+关键节点支持人工确认：
+- SRS 文档审核
+- 测试用例评审
+- 最终代码验收
+
+### 5. 实时可视化
+
+Streamlit 前端提供：
+- 实时对话展示
+- Agent 状态监控
+- 工作流进度跟踪
+- 对话历史导出
+
+## 🔧 配置说明
+
+### 模型配置
+
+在 `config/llm_config.py` 中配置：
+
+```python
+DASHSCOPE_API_KEY = "your_api_key"
+DASHSCOPE_BASE_URL = "https://dashscope.aliyuncs.com/compatible-mode/v1"
+```
+
+### 支持的模型
+
+- `qwen3.5-flash` (推荐，速度快)
+- `qwen-max` (最强性能)
+- `qwen-plus` (平衡性能和成本)
+- `qwen-turbo` (最快速度)
+
+### Agent 系统提示词
+
+每个 Agent 都有专属的系统提示词，定义在 `config/llm_config.py` 中：
+
+- `PM_PROMPT` - 产品经理提示词
+- `QA_PROMPT` - 测试工程师提示词
+- `DEV_PROMPT` - 开发工程师提示词
+- `ORCH_PROMPT` - 协调器提示词
+
+## 📖 使用示例
+
+### 示例 1: 使用命令行
+
+```python
+from autogen_sdls_system import AutoGenSDLCSystem
+
+system = AutoGenSDLCSystem(api_key="your_key")
+
+requirement = "开发一个电池健康状态预测 API"
+result = system.run_workflow(requirement, max_round=15)
+
+print(result["summary"])
+```
+
+### 示例 2: 单独调用 Agent
+
+```python
+from agents import ProductManagerAgent
+
+pm = ProductManagerAgent(api_key="your_key")
+srs = pm.generate_srs("我需要一个车载蓝牙管理模块")
+print(srs)
+```
+
+### 示例 3: 自定义配置
+
+```python
+from config.llm_config import get_llm_config
+
+llm_config = get_llm_config(
+    model="qwen-max",
+    temperature=0.5,
+    max_tokens=4096
+)
+```
+
+## 🎬 Workshop 演示脚本
+
+### 场景：电池健康状态 (SOH) 预测 API
+
+1. **输入模糊需求**
+   ```
+   "我需要一个电池健康预测功能"
+   ```
+
+2. **PM Agent 输出** - 完整 SRS 文档
+   - 功能性需求
+   - 非功能性需求
+   - 验收标准
+
+3. **QA Agent 输出** - Pytest 测试脚本
+   - 测试函数
+   - BDD 场景描述
+
+4. **Dev Agent 输出** - 可运行的 API 代码
+   - 核心业务逻辑
+   - 错误处理
+
+5. **自动测试** - 展示测试执行和修复循环
+
+6. **最终输出** - 可部署的代码包 + 测试报告
+
+## ✅ 演示亮点
+
+- ✅ 实时展示 Agent 协作对话
+- ✅ 人机协同确认节点
+- ✅ 测试失败→自动修复→通过的闭环
+- ✅ 符合汽车行业标准（MISRA-C、ISO 26262）
+
+## 🐛 常见问题
+
+### Q1: API Key 无效
+
+**解决**: 检查环境变量是否正确设置
+
+```bash
+# 验证环境变量
+echo $DASHSCOPE_API_KEY  # Linux/Mac
+echo %DASHSCOPE_API_KEY% # Windows CMD
+```
+
+### Q2: 模型响应超时
+
+**解决**: 增加超时时间或降低 `max_tokens`
+
+```python
+llm_config = get_llm_config(timeout=180)
+```
+
+### Q3: 测试执行失败
+
+**解决**: 检查工作目录权限和 pytest 安装
+
+```bash
+pip install pytest
+chmod +x workspace/
+```
+
+## 📊 日志与导出
+
+### 对话日志
+
+所有对话自动保存在 `logs/` 目录：
+
+- JSONL 格式：`session_YYYYMMDD_HHMMSS.jsonl`
+- JSON 导出：`conversation_*.json`
+- Markdown 导出：`conversation_*.md`
+
+### 导出方法
+
+```python
+system.export_conversation("my_conversation.json")
+system.export_report("my_report.md")
+```
+
+## 🔐 安全建议
+
+1. **不要硬编码 API Key** - 使用环境变量
+2. **审查生成的代码** - AI 可能产生不安全代码
+3. **限制执行权限** - 禁用 Docker 或使用沙箱
+4. **定期清理 workspace** - 避免敏感数据积累
+
+## 📝 许可证
+
+MIT License
+
+## 🤝 贡献
+
+欢迎提交 Issue 和 Pull Request！
+
+## 📧 联系方式
+
+如有问题请提交 Issue 或联系开发团队。
+
+---
+
+**🤖 Generated with AutoGen SDLC System**
diff --git a/STREAMLIT_V3_GUIDE.md b/STREAMLIT_V3_GUIDE.md
new file mode 100644
index 0000000..c188721
--- /dev/null
+++ b/STREAMLIT_V3_GUIDE.md
@@ -0,0 +1,208 @@
+# Streamlit v3 - Agent 对话流可视化版本
+
+## 🎯 核心特性
+
+这个全新版本**专门突出显示**:
+1. ✅ **每个 Agent 当前在做什么** (任务标签实时显示)
+2. ✅ **Agent 之间的对话流** (类似微信/QQ 聊天界面)
+3. ✅ **当前发言的 Agent** (绿色高亮 + 脉冲动画)
+4. ✅ **对话传递关系** (箭头指示下一个 Agent)
+
+## 🚀 启动方法
+
+```bash
+streamlit run frontend/streamlit_app_v3.py
+```
+
+## 🎨 界面说明
+
+### 1. Agent 状态行 (顶部)
+```
+┌─────────────┬─────────────┬─────────────┬─────────────┬─────────────┐
+│  📋 产品经理 │  ✅ 测试工程 │  💻 开发工程 │  🎯 协调器  │  👤 用户代理 │
+│   ⚪ 待命    │   ⚪ 待命    │   ⚪ 待命    │   ⚪ 待命    │   ⚪ 待命    │
+│  📝 需求分析 │  📝 测试设计 │  📝 代码实现 │  📝 流程协调 │  📝 提出需求 │
+│  💬 0 条消息  │  💬 0 条消息  │  💬 0 条消息  │  💬 0 条消息  │  💬 0 条消息  │
+└─────────────┴─────────────┴─────────────┴─────────────┴─────────────┘
+```
+
+**特点**:
+- 🔵 **绿色边框 + 脉冲动画** = 当前正在发言的 Agent
+- ⚪ **灰色** = 等待中的 Agent
+- 📝 **任务标签** = 实时显示该 Agent 在做什么
+
+### 2. 当前发言横幅
+```
+🟢 当前发言：产品经理 - 需求分析
+```
+绿色横幅，明确告诉您现在哪个 Agent 在说话！
+
+### 3. Agent 对话流 (核心区域)
+```
+┌─────────────────────────────────────────────────┐
+│ 📋 产品经理  [需求分析]  11:09:30               │
+│                                                 │
+│ 收到！开始分析需求...                           │
+│ 功能性需求：                                    │
+│ 1. 接收电压/电流/温度数据                       │
+│ 2. 计算 SOH 百分比                              │
+│ 3. 异常检测                                     │
+└─────────────────────────────────────────────────┘
+                    ⬇️
+          传递给 测试工程师
+                    ⬇️
+┌─────────────────────────────────────────────────┐
+│ ✅ 测试工程师  [测试设计]  11:09:45              │
+│                                                 │
+│ 基于 SRS 设计测试用例...                        │
+│ test_normal_operation()                         │
+│ test_zero_voltage()                             │
+└─────────────────────────────────────────────────┘
+```
+
+**特点**:
+- 💬 **彩色气泡** = 不同 Agent 有不同颜色
+- 📝 **任务徽章** = 每个消息都显示当前任务
+- ⬇️ **传递箭头** = 显示工作流转向下一个 Agent
+- 🕐 **时间戳** = 精确到秒的发言时间
+
+## 🎭 Agent 颜色方案
+
+| Agent | 颜色 | 气泡样式 |
+|-------|------|---------|
+| 📋 PM Agent | 🔵 蓝色 | 蓝色渐变 |
+| ✅ QA Agent | 🟢 绿色 | 绿色渐变 |
+| 💻 Dev Agent | 🟠 橙色 | 橙色渐变 |
+| 🎯 Orchestrator | 🟣 紫色 | 紫色渐变 |
+| 👤 User Proxy | 🟤 棕色 | 棕色渐变 |
+
+## 📊 实时任务标签
+
+每个 Agent 的任务会动态更新：
+
+### PM Agent 任务流
+1. `待命` → 初始状态
+2. `启动流程` → 收到 Orchestrator 指令
+3. `需求分析` → 分析用户需求
+4. `生成 SRS` → 编写需求文档
+5. `待命` → 完成任务
+
+### QA Agent 任务流
+1. `待命` → 初始状态
+2. `测试设计` → 阅读 SRS
+3. `编写测试` → 创建 test_*.py
+4. `TDD 实践` → 确保测试先行
+5. `待命` → 完成任务
+
+### Dev Agent 任务流
+1. `待命` → 初始状态
+2. `代码实现` → 阅读 SRS 和测试
+3. `编写代码` → 创建 src_*.py
+4. `修复 bug` → 测试失败时
+5. `待命` → 完成任务
+
+## 🔍 对话流示例
+
+完整的 SDLC 对话流会是这样：
+
+```
+👤 用户代理 [提出需求]
+  "我需要一个电池健康状态预测 API"
+        ⬇️ 传递给 协调器
+
+🎯 协调器 [启动流程]
+  "🚀 启动 SDLC 工作流！开始协作..."
+        ⬇️ 传递给 产品经理
+
+📋 产品经理 [需求分析]
+  "收到！分析需求中...
+   功能性需求：
+   1. 接收电压/电流/温度数据
+   2. 计算 SOH 百分比
+   3. 异常检测..."
+        ⬇️ 传递给 测试工程师
+
+✅ 测试工程师 [测试设计]
+  "基于 SRS 设计测试用例...
+   test_normal_operation()
+   test_zero_voltage()..."
+        ⬇️ 传递给 开发工程师
+
+💻 开发工程师 [代码实现]
+  "开始编写代码...
+  实现 calculate_soh() 函数...
+  符合 PEP8 规范..."
+        ⬇️ 传递给 用户代理
+
+👤 用户代理 [测试执行]
+  "执行测试...
+  ✅ test_normal_operation PASSED
+  ✅ test_zero_voltage PASSED..."
+        ⬇️ 传递给 协调器
+
+🎯 协调器 [总结完成]
+  "✅ SDLC 流程完成！
+  生成文件：SRS.md, test_battery_health.py..."
+```
+
+## 💡 使用技巧
+
+### 1. 实时追踪特定 Agent
+- 看**顶部状态行** → 哪个 Agent 绿色高亮就是在发言
+- 看**任务标签** → 了解该 Agent 具体在做什么
+- 看**消息数量** → 知道哪个 Agent 最活跃
+
+### 2. 理解对话流
+- 看**气泡颜色** → 快速识别是哪个 Agent
+- 看**传递箭头** → 理解工作流方向
+- 看**时间戳** → 了解执行顺序
+
+### 3. 调试问题
+- 如果某个 Agent **一直不说话** → 可能是模型配置问题
+- 如果**对话卡住** → 检查 API Key 和网络
+- 如果**任务标签不更新** → 刷新页面重试
+
+## 🆚 与 v2 的区别
+
+| 特性 | v2 | v3 (新版) |
+|------|----|-----------|
+| Agent 状态 | 静态卡片 | 动态高亮 + 任务标签 |
+| 对话展示 | 普通列表 | 聊天气泡流 |
+| 任务追踪 | ❌ 无 | ✅ 实时显示 |
+| 传递关系 | ❌ 无 | ✅ 箭头指示 |
+| 当前发言 | ⚪ 小圆点 | 🟢 绿色高亮 + 横幅 |
+
+## 🎬 演示场景
+
+### 场景 1: 电池健康预测 API
+1. 输入需求："我需要一个电池健康状态预测 API"
+2. 点击"启动对话流"
+3. 观察 5 个 Agent 的协作过程
+4. 查看每个 Agent 的任务变化
+5. 最终生成 3 个文件
+
+### 场景 2: 车牌识别系统
+1. 输入需求："开发一个车牌识别系统"
+2. 启动对话流
+3. 看 PM 如何分析需求
+4. 看 QA 如何设计测试
+5. 看 Dev 如何实现代码
+
+## 📥 导出功能
+
+点击侧边栏的"📥 导出对话"，可以保存完整的对话历史为 JSON 格式，包含：
+- 每个 Agent 的发言内容
+- 发言时间戳
+- 当时的任务状态
+- 对话顺序
+
+## 🐛 注意事项
+
+1. **确保安装依赖**: `pip install streamlit pyautogen dashscope`
+2. **设置 API Key**: `$env:DASHSCOPE_API_KEY="your_key"`
+3. **网络稳定**: 需要稳定的网络连接访问阿里云 API
+4. **耐心等待**: 大模型生成需要时间，每轮约 10-30 秒
+
+---
+
+**🎉 享受全新的 Agent 对话流可视化体验！**
diff --git a/TROUBLESHOOTING.md b/TROUBLESHOOTING.md
new file mode 100644
index 0000000..496295b
--- /dev/null
+++ b/TROUBLESHOOTING.md
@@ -0,0 +1,156 @@
+# AutoGen SDLC 系统 - 问题修复说明
+
+## 🔧 已修复的问题
+
+### 问题：Streamlit 运行时报错 `'TERMINAL'`
+
+**错误现象**:
+```
+错误：'TERMINAL'
+```
+
+**根本原因**:
+`UserProxyAgent` 的 `human_input_mode="TERMINAL"` 设置会在 Web 环境下尝试使用终端输入，导致 Streamlit 报错。
+
+**解决方案**:
+将所有文件中的 `human_input_mode="TERMINAL"` 改为 `human_input_mode="NEVER"`。
+
+### 修改的文件
+
+1. ✅ `frontend/streamlit_app_v2.py` (第 224 行)
+2. ✅ `frontend/streamlit_app.py` (第 104 行)
+3. ✅ `autogen_sdls_system.py` (第 115 行)
+
+---
+
+## 🚀 现在可以正常运行了
+
+### 方式 1: Streamlit 增强版界面（推荐）
+
+```bash
+streamlit run frontend/streamlit_app_v2.py
+```
+
+**功能特色**:
+- 🎭 Agent 实时状态面板
+- 🔄 Mermaid 工作流程图
+- ⏱️ 执行时间线
+- 💬 对话实录展示
+- 📁 生成文件预览
+- 📊 统计仪表盘
+
+### 方式 2: Streamlit 经典界面
+
+```bash
+streamlit run frontend/streamlit_app.py
+```
+
+### 方式 3: 命令行完整工作流
+
+```bash
+python autogen_sdls_system.py
+```
+
+### 方式 4: HTML 独立演示（无需依赖）
+
+```bash
+# 直接在浏览器打开
+demo_visualization.html
+```
+
+---
+
+## ⚙️ 配置说明
+
+### 环境变量设置
+
+**Windows PowerShell**:
+```powershell
+$env:DASHSCOPE_API_KEY="your_api_key_here"
+```
+
+**Linux/Mac**:
+```bash
+export DASHSCOPE_API_KEY="your_api_key_here"
+```
+
+### 依赖安装
+
+```bash
+pip install -r requirements.txt
+```
+
+或单独安装：
+```bash
+pip install streamlit autogen dashscope pytest
+```
+
+---
+
+## 🎯 Agent 说明
+
+| Agent | 职责 | 图标 |
+|-------|------|------|
+| PM Agent | 产品经理 - 需求分析与 SRS 生成 | 📋 |
+| QA Agent | 测试工程师 - 测试用例设计 | ✅ |
+| Dev Agent | 开发工程师 - 代码实现 | 💻 |
+| Orchestrator | 协调器 - 流程调度与验证 | 🎯 |
+| User Proxy | 用户代理 - 人机交互 | 👤 |
+
+---
+
+## 📊 工作流程
+
+```
+用户需求 
+  ↓
+📋 PM Agent → SRS 文档
+  ↓
+✅ QA Agent → 测试用例
+  ↓
+💻 Dev Agent → 源代码
+  ↓
+🔧 自动测试
+  ↓ (失败)
+💻 Dev Agent (修复)
+  ↓ (通过)
+🎯 Orchestrator → 最终报告
+```
+
+---
+
+## 💡 使用提示
+
+1. **API Key 获取**: 访问 [阿里云 DashScope 控制台](https://dashscope.console.aliyun.com/)
+2. **模型选择**: 推荐使用 `qwen3.5-flash`（速度快、成本低）
+3. **工作目录**: 所有生成的文件保存在 `workspace/` 目录
+4. **日志位置**: 对话历史保存在 `logs/` 目录
+5. **人机协同**: 如需人工确认，可在代码中添加 Streamlit 弹窗
+
+---
+
+## 🐛 常见问题
+
+### Q1: 仍然报错怎么办？
+
+**检查清单**:
+- [ ] API Key 是否正确设置
+- [ ] 依赖包是否安装完整
+- [ ] Python 版本是否 >= 3.9
+- [ ] 工作目录是否有写权限
+
+### Q2: 如何查看详细的错误信息？
+
+在 Streamlit 界面查看底部的红色错误提示，或在命令行运行查看完整堆栈。
+
+### Q3: 生成的代码在哪里？
+
+所有生成的文件都在 `workspace/` 目录下，可以在 Streamlit 界面的"生成的文件"区域查看。
+
+---
+
+## 📞 技术支持
+
+如有其他问题，请查看：
+- [AutoGen 官方文档](https://microsoft.github.io/autogen/)
+- [阿里云 DashScope 文档](https://help.aliyun.com/zh/dashscope/)
diff --git a/agents/__init__.py b/agents/__init__.py
new file mode 100644
index 0000000..d9f291d
--- /dev/null
+++ b/agents/__init__.py
@@ -0,0 +1,16 @@
+"""Agent 模块初始化文件"""
+from .pm_agent import ProductManagerAgent, create_pm_agent
+from .qa_agent import QAAgent, create_qa_agent
+from .dev_agent import DevAgent, create_dev_agent
+from .orchestrator import OrchestratorAgent, create_orchestrator_agent
+
+__all__ = [
+    "ProductManagerAgent",
+    "create_pm_agent",
+    "QAAgent",
+    "create_qa_agent",
+    "DevAgent",
+    "create_dev_agent",
+    "OrchestratorAgent",
+    "create_orchestrator_agent"
+]
diff --git a/agents/dev_agent.py b/agents/dev_agent.py
new file mode 100644
index 0000000..7792b87
--- /dev/null
+++ b/agents/dev_agent.py
@@ -0,0 +1,256 @@
+"""
+Dev Agent - 开发工程师智能体
+负责代码实现和测试驱动开发
+"""
+from autogen import AssistantAgent
+from typing import Dict, Any, Optional, List
+import os
+from pathlib import Path
+
+from config.llm_config import get_agent_llm_config, DEV_PROMPT
+
+
+class DevAgent:
+    """开发工程师 Agent，负责编写高质量代码"""
+    
+    def __init__(self, llm_config: Optional[Dict] = None):
+        """
+        初始化 Dev Agent
+        
+        Args:
+            llm_config: LLM 配置
+        """
+        self.llm_config = llm_config or get_agent_llm_config("Dev_Agent")
+        
+        self.agent = AssistantAgent(
+            name="Dev_Agent",
+            system_message=DEV_PROMPT,
+            llm_config=self.llm_config,
+            description="资深软件工程师，专注于汽车嵌入式 C++/Python 开发",
+            human_input_mode="NEVER"
+        )
+        
+        self.workspace_dir = Path("workspace")
+        self.workspace_dir.mkdir(exist_ok=True)
+        
+        # 代码历史用于迭代修复
+        self.code_history: List[str] = []
+    
+    def generate_code(self, srs_content: str, test_code: str) -> str:
+        """
+        根据 SRS 和测试用例生成实现代码
+        
+        Args:
+            srs_content: SRS 文档内容
+            test_code: 测试代码
+        
+        Returns:
+            生成的源代码内容
+        """
+        prompt = f"""
+请根据以下 SRS 和测试用例编写实现代码：
+
+【SRS 需求】
+{self._truncate(srs_content, 2500)}
+
+【测试用例】
+{self._truncate(test_code, 2000)}
+
+请编写：
+1. 完整的实现代码
+2. 符合 MISRA-C/PEP8 规范
+3. 包含详细的 docstring 和类型注解
+4. 确保所有测试用例通过
+
+输出 Python 代码，保存为 src_battery_health.py。
+"""
+        
+        response = self.agent.generate_reply(
+            messages=[{"role": "user", "content": prompt}]
+        )
+        
+        code_content = response if isinstance(response, str) else str(response)
+        
+        # 提取代码块（如果包含 Markdown 标记）
+        code_content = self._extract_code(code_content)
+        
+        # 保存到文件
+        code_file = self.workspace_dir / "src_battery_health.py"
+        with open(code_file, 'w', encoding='utf-8') as f:
+            f.write(code_content)
+        
+        self.code_history.append(code_content)
+        print(f"✅ 源代码已生成：{code_file}")
+        return code_content
+    
+    def fix_code(self, srs_content: str, test_code: str, error_log: str) -> str:
+        """
+        根据测试错误修复代码
+        
+        Args:
+            srs_content: SRS 文档
+            test_code: 测试代码
+            error_log: 测试失败日志
+        
+        Returns:
+            修复后的代码
+        """
+        previous_code = self.code_history[-1] if self.code_history else ""
+        
+        prompt = f"""
+代码测试失败，请根据错误日志修复代码：
+
+【SRS 需求】
+{self._truncate(srs_content, 1500)}
+
+【测试代码】
+{self._truncate(test_code, 1500)}
+
+【之前的代码】
+{self._truncate(previous_code, 2000)}
+
+【错误日志】
+{self._truncate(error_log, 1500)}
+
+请分析错误原因并修复代码，确保：
+1. 所有测试用例通过
+2. 保持代码质量
+3. 不引入新的问题
+
+输出完整的修复后代码。
+"""
+        
+        response = self.agent.generate_reply(
+            messages=[{"role": "user", "content": prompt}]
+        )
+        
+        fixed_code = response if isinstance(response, str) else str(response)
+        fixed_code = self._extract_code(fixed_code)
+        
+        # 更新文件
+        code_file = self.workspace_dir / "src_battery_health.py"
+        with open(code_file, 'w', encoding='utf-8') as f:
+            f.write(fixed_code)
+        
+        self.code_history.append(fixed_code)
+        print(f"✅ 代码已修复：{code_file}")
+        return fixed_code
+    
+    def optimize_code(self, code: str, optimization_goal: str) -> str:
+        """
+        优化现有代码
+        
+        Args:
+            code: 原始代码
+            optimization_goal: 优化目标（性能、可读性等）
+        
+        Returns:
+            优化后的代码
+        """
+        prompt = f"""
+请优化以下代码：
+
+【原始代码】
+{code}
+
+【优化目标】
+{optimization_goal}
+
+请保持功能不变，提升代码质量。
+"""
+        
+        response = self.agent.generate_reply(
+            messages=[{"role": "user", "content": prompt}]
+        )
+        
+        optimized_code = response if isinstance(response, str) else str(response)
+        optimized_code = self._extract_code(optimized_code)
+        
+        return optimized_code
+    
+    def add_documentation(self, code: str) -> str:
+        """
+        为代码添加完整文档
+        
+        Args:
+            code: 代码
+        
+        Returns:
+            带完整文档的代码
+        """
+        prompt = f"""
+请为以下代码添加完整的文档：
+
+【代码】
+{code}
+
+请添加：
+1. 模块级 docstring
+2. 类 docstring
+3. 函数 docstring（包含参数说明和返回值）
+4. 类型注解
+5. 关键步骤的注释
+
+保持代码功能不变。
+"""
+        
+        response = self.agent.generate_reply(
+            messages=[{"role": "user", "content": prompt}]
+        )
+        
+        documented_code = response if isinstance(response, str) else str(response)
+        documented_code = self._extract_code(documented_code)
+        
+        return documented_code
+    
+    def _extract_code(self, text: str) -> str:
+        """从文本中提取代码（移除 Markdown 标记）"""
+        lines = text.split('\n')
+        code_lines = []
+        in_code_block = False
+        
+        for line in lines:
+            if line.strip().startswith('```python'):
+                in_code_block = True
+                continue
+            elif line.strip().startswith('```'):
+                in_code_block = False
+                continue
+            
+            if in_code_block or not any(line.strip().startswith(x) for x in ['```']):
+                code_lines.append(line)
+        
+        # 如果没有找到代码块，返回原文
+        if not code_lines:
+            return text
+        
+        return '\n'.join(code_lines)
+    
+    def _truncate(self, text: str, max_length: int) -> str:
+        """截断文本"""
+        if len(text) <= max_length:
+            return text
+        return text[:max_length] + "... [内容已截断]"
+
+
+def create_dev_agent(llm_config: Optional[Dict] = None) -> AssistantAgent:
+    """
+    创建 Dev Agent（AutoGen 原生格式）
+    
+    Args:
+        llm_config: LLM 配置
+    
+    Returns:
+        AutoGen AssistantAgent 实例
+    """
+    config = llm_config or get_agent_llm_config("Dev_Agent")
+    
+    agent = AssistantAgent(
+        name="Dev_Agent",
+        system_message=DEV_PROMPT,
+        llm_config=config,
+        description="资深软件工程师",
+        human_input_mode="NEVER"
+    )
+    
+    return agent
diff --git a/agents/orchestrator.py b/agents/orchestrator.py
new file mode 100644
index 0000000..a87c88e
--- /dev/null
+++ b/agents/orchestrator.py
@@ -0,0 +1,292 @@
+"""
+Orchestrator Agent - 协调器智能体
+负责流程调度与最终验证
+"""
+from autogen import AssistantAgent
+from typing import Dict, Any, Optional, List
+import os
+from pathlib import Path
+
+from config.llm_config import get_agent_llm_config, ORCH_PROMPT
+
+
+class OrchestratorAgent:
+    """协调器 Agent，负责多智能体协同和流程控制"""
+    
+    def __init__(self, llm_config: Optional[Dict] = None):
+        """
+        初始化 Orchestrator Agent
+        
+        Args:
+            llm_config: LLM 配置
+        """
+        self.llm_config = llm_config or get_agent_llm_config("Orchestrator")
+        
+        self.agent = AssistantAgent(
+            name="Orchestrator",
+            system_message=ORCH_PROMPT,
+            llm_config=self.llm_config,
+            description="多智能体系统协调器",
+            human_input_mode="NEVER"
+        )
+        
+        self.workspace_dir = Path("workspace")
+        self.workspace_dir.mkdir(exist_ok=True)
+        
+        # 流程状态
+        self.workflow_state: Dict[str, Any] = {
+            "current_step": 0,
+            "total_steps": 5,
+            "status": "pending",
+            "artifacts": {}
+        }
+    
+    def start_workflow(self, user_requirement: str) -> Dict[str, Any]:
+        """
+        启动完整的工作流程
+        
+        Args:
+            user_requirement: 用户需求
+        
+        Returns:
+            工作流状态
+        """
+        self.workflow_state = {
+            "current_step": 1,
+            "total_steps": 5,
+            "status": "in_progress",
+            "user_requirement": user_requirement,
+            "artifacts": {}
+        }
+        
+        return self.workflow_state
+    
+    def validate_srs(self, srs_content: str) -> Dict[str, Any]:
+        """
+        验证 SRS 文档的完整性
+        
+        Args:
+            srs_content: SRS 文档内容
+        
+        Returns:
+            验证结果
+        """
+        prompt = f"""
+请验证以下 SRS 文档的完整性：
+
+{self._truncate(srs_content, 3000)}
+
+检查清单：
+1. ✅ 包含功能性需求列表
+2. ✅ 包含非功能性需求
+3. ✅ 包含验收标准
+4. ✅ 包含风险分析
+5. ✅ 需求具有唯一 ID
+
+请输出验证报告，指出缺失或不完整的部分。
+"""
+        
+        response = self.agent.generate_reply(
+            messages=[{"role": "user", "content": prompt}]
+        )
+        
+        validation_report = response if isinstance(response, str) else str(response)
+        
+        # 保存验证报告
+        report_file = self.workspace_dir / "srs_validation.md"
+        with open(report_file, 'w', encoding='utf-8') as f:
+            f.write(validation_report)
+        
+        return {
+            "valid": "✅" in validation_report and "❌" not in validation_report,
+            "report": validation_report,
+            "file": str(report_file)
+        }
+    
+    def validate_tests(self, test_code: str, srs_content: str) -> Dict[str, Any]:
+        """
+        验证测试用例的覆盖率
+        
+        Args:
+            test_code: 测试代码
+            srs_content: SRS 文档
+        
+        Returns:
+            验证结果
+        """
+        prompt = f"""
+请验证测试用例是否覆盖了所有 SRS 需求：
+
+【SRS 需求】
+{self._truncate(srs_content, 2000)}
+
+【测试代码】
+{self._truncate(test_code, 2000)}
+
+检查清单：
+1. ✅ 每个功能需求都有对应测试
+2. ✅ 包含边界情况测试
+3. ✅ 包含异常场景测试
+4. ✅ 测试可执行且独立
+5. ✅ 遵循 TDD 原则
+
+请输出验证报告。
+"""
+        
+        response = self.agent.generate_reply(
+            messages=[{"role": "user", "content": prompt}]
+        )
+        
+        validation_report = response if isinstance(response, str) else str(response)
+        
+        return {
+            "valid": "✅" in validation_report,
+            "report": validation_report
+        }
+    
+    def validate_code(self, code: str, srs_content: str, test_result: Dict) -> Dict[str, Any]:
+        """
+        验证代码质量
+        
+        Args:
+            code: 源代码
+            srs_content: SRS 文档
+            test_result: 测试结果
+        
+        Returns:
+            验证结果
+        """
+        prompt = f"""
+请验证代码质量：
+
+【SRS 需求】
+{self._truncate(srs_content, 1500)}
+
+【代码】
+{self._truncate(code, 2000)}
+
+【测试结果】
+{test_result}
+
+检查清单：
+1. ✅ 实现所有功能需求
+2. ✅ 通过所有测试用例
+3. ✅ 代码符合规范（MISRA-C/PEP8）
+4. ✅ 包含完整文档
+5. ✅ 无安全漏洞
+6. ✅ 性能满足要求
+
+请输出代码质量验证报告。
+"""
+        
+        response = self.agent.generate_reply(
+            messages=[{"role": "user", "content": prompt}]
+        )
+        
+        validation_report = response if isinstance(response, str) else str(response)
+        
+        return {
+            "valid": test_result.get("success", False) and "✅" in validation_report,
+            "report": validation_report
+        }
+    
+    def generate_final_report(self) -> str:
+        """
+        生成最终项目总结报告
+        
+        Returns:
+            最终报告内容
+        """
+        prompt = """
+请生成项目最终总结报告，包含：
+
+1. 项目概述
+2. 交付物清单：
+   - SRS 文档
+   - 测试用例
+   - 源代码
+3. 质量指标：
+   - 测试覆盖率
+   - 代码质量评分
+4. 合规性说明：
+   - ISO 26262
+   - MISRA-C
+   - ASPICE
+5. 后续建议
+
+请基于 workspace 目录下的所有文件生成完整报告。
+"""
+        
+        response = self.agent.generate_reply(
+            messages=[{"role": "user", "content": prompt}]
+        )
+        
+        final_report = response if isinstance(response, str) else str(response)
+        
+        # 保存最终报告
+        report_file = self.workspace_dir / "FINAL_REPORT.md"
+        with open(report_file, 'w', encoding='utf-8') as f:
+            f.write(final_report)
+        
+        print(f"✅ 最终报告已生成：{report_file}")
+        return final_report
+    
+    def request_human_approval(
+        self,
+        approval_type: str,
+        description: str,
+        data: Dict[str, Any]
+    ) -> bool:
+        """
+        请求人工确认（需要前端配合实现）
+        
+        Args:
+            approval_type: 确认类型
+            description: 确认描述
+            data: 相关数据
+        
+        Returns:
+            用户是否批准
+        """
+        # 这里只是标记，实际的前端交互由 Streamlit 处理
+        print(f"\n⚠️ 需要人工确认：{description}")
+        print(f"类型：{approval_type}")
+        print(f"数据：{data}\n")
+        
+        # 在命令行模式下，可以简单询问用户
+        # 在 GUI 模式下，这会触发前端弹窗
+        try:
+            response = input("是否批准？(y/n): ").lower()
+            return response == 'y'
+        except:
+            # 非交互模式下默认批准
+            return True
+    
+    def _truncate(self, text: str, max_length: int) -> str:
+        """截断文本"""
+        if len(text) <= max_length:
+            return text
+        return text[:max_length] + "... [内容已截断]"
+
+
+def create_orchestrator_agent(llm_config: Optional[Dict] = None) -> AssistantAgent:
+    """
+    创建 Orchestrator Agent（AutoGen 原生格式）
+    
+    Args:
+        llm_config: LLM 配置
+    
+    Returns:
+        AutoGen AssistantAgent 实例
+    """
+    config = llm_config or get_agent_llm_config("Orchestrator")
+    
+    agent = AssistantAgent(
+        name="Orchestrator",
+        system_message=ORCH_PROMPT,
+        llm_config=config,
+        description="多智能体系统协调器",
+        human_input_mode="NEVER"
+    )
+    
+    return agent
diff --git a/agents/pm_agent.py b/agents/pm_agent.py
new file mode 100644
index 0000000..bd68684
--- /dev/null
+++ b/agents/pm_agent.py
@@ -0,0 +1,139 @@
+"""
+PM Agent - 产品经理智能体
+负责需求消歧与规格生成
+"""
+from autogen import AssistantAgent
+from typing import Dict, Any, Optional
+import os
+from pathlib import Path
+
+from config.llm_config import get_agent_llm_config, PM_PROMPT
+
+
+class ProductManagerAgent:
+    """产品经理 Agent，负责生成软件需求规格说明书"""
+    
+    def __init__(self, llm_config: Optional[Dict] = None):
+        """
+        初始化 PM Agent
+        
+        Args:
+            llm_config: LLM 配置，为 None 时使用默认配置
+        """
+        self.llm_config = llm_config or get_agent_llm_config("PM_Agent")
+        
+        # 创建 AutoGen AssistantAgent
+        self.agent = AssistantAgent(
+            name="PM_Agent",
+            system_message=PM_PROMPT,
+            llm_config=self.llm_config,
+            description="资深软件产品经理，专注于汽车嵌入式系统领域",
+            human_input_mode="NEVER"  # 全自动模式
+        )
+        
+        self.workspace_dir = Path("workspace")
+        self.workspace_dir.mkdir(exist_ok=True)
+    
+    def generate_srs(self, user_requirement: str) -> str:
+        """
+        根据用户需求生成 SRS 文档
+        
+        Args:
+            user_requirement: 用户输入的原始需求
+        
+        Returns:
+            生成的 SRS 文档内容
+        """
+        prompt = f"""
+请根据以下用户需求生成完整的《软件需求规格说明书 (SRS)》：
+
+用户需求：{user_requirement}
+
+请确保输出包含：
+1. 文档标题和版本信息
+2. 功能性需求列表（FR-001, FR-002...）
+3. 非功能性需求（NFR-001, NFR-002...）
+4. 验收标准（AC-001, AC-002...）
+5. 潜在风险与边缘情况
+
+请以 Markdown 格式输出完整文档。
+"""
+        
+        # 调用 Agent 生成 SRS
+        response = self.agent.generate_reply(
+            messages=[{"role": "user", "content": prompt}]
+        )
+        
+        srs_content = response if isinstance(response, str) else str(response)
+        
+        # 保存到文件
+        srs_file = self.workspace_dir / "SRS.md"
+        with open(srs_file, 'w', encoding='utf-8') as f:
+            f.write(srs_content)
+        
+        print(f"✅ SRS 文档已生成：{srs_file}")
+        return srs_content
+    
+    def refine_requirements(
+        self,
+        original_srs: str,
+        feedback: str
+    ) -> str:
+        """
+        根据反馈优化需求
+        
+        Args:
+            original_srs: 原始 SRS 文档
+            feedback: 反馈意见
+        
+        Returns:
+            优化后的 SRS 文档
+        """
+        prompt = f"""
+请根据以下反馈优化现有的 SRS 文档：
+
+原始 SRS:
+{original_srs[:2000]}...  # 限制长度避免超出上下文
+
+反馈意见：
+{feedback}
+
+请输出优化后的完整 SRS 文档。
+"""
+        
+        response = self.agent.generate_reply(
+            messages=[{"role": "user", "content": prompt}]
+        )
+        
+        refined_srs = response if isinstance(response, str) else str(response)
+        
+        # 更新文件
+        srs_file = self.workspace_dir / "SRS.md"
+        with open(srs_file, 'w', encoding='utf-8') as f:
+            f.write(refined_srs)
+        
+        print(f"✅ SRS 文档已更新：{srs_file}")
+        return refined_srs
+
+
+def create_pm_agent(llm_config: Optional[Dict] = None) -> AssistantAgent:
+    """
+    创建 PM Agent（AutoGen 原生格式）
+    
+    Args:
+        llm_config: LLM 配置
+    
+    Returns:
+        AutoGen AssistantAgent 实例
+    """
+    config = llm_config or get_agent_llm_config("PM_Agent")
+    
+    agent = AssistantAgent(
+        name="PM_Agent",
+        system_message=PM_PROMPT,
+        llm_config=config,
+        description="资深软件产品经理",
+        human_input_mode="NEVER"
+    )
+    
+    return agent
diff --git a/agents/qa_agent.py b/agents/qa_agent.py
new file mode 100644
index 0000000..8e28551
--- /dev/null
+++ b/agents/qa_agent.py
@@ -0,0 +1,221 @@
+"""
+QA Agent - 测试工程师智能体
+负责测试用例创建和 TDD 实践
+"""
+from autogen import AssistantAgent
+from typing import Dict, Any, Optional
+import os
+from pathlib import Path
+
+from config.llm_config import get_agent_llm_config, QA_PROMPT
+
+
+class QAAgent:
+    """测试工程师 Agent，负责生成测试用例和测试脚本"""
+    
+    def __init__(self, llm_config: Optional[Dict] = None):
+        """
+        初始化 QA Agent
+        
+        Args:
+            llm_config: LLM 配置
+        """
+        self.llm_config = llm_config or get_agent_llm_config("QA_Agent")
+        
+        self.agent = AssistantAgent(
+            name="QA_Agent",
+            system_message=QA_PROMPT,
+            llm_config=self.llm_config,
+            description="资深测试工程师，专注于自动化测试和 TDD 实践",
+            human_input_mode="NEVER"
+        )
+        
+        self.workspace_dir = Path("workspace")
+        self.workspace_dir.mkdir(exist_ok=True)
+    
+    def generate_test_cases(self, srs_content: str) -> str:
+        """
+        根据 SRS 生成测试用例
+        
+        Args:
+            srs_content: SRS 文档内容
+        
+        Returns:
+            生成的测试用例内容
+        """
+        prompt = f"""
+请根据以下 SRS 文档生成完整的测试用例：
+
+{self._truncate(srs_content, 3000)}
+
+请生成：
+1. Pytest 测试脚本（包含完整的测试函数）
+2. BDD 风格的测试场景描述
+3. 测试数据准备
+4. 预期结果验证
+
+确保遵循 TDD 原则，测试先于代码存在。
+"""
+        
+        response = self.agent.generate_reply(
+            messages=[{"role": "user", "content": prompt}]
+        )
+        
+        test_content = response if isinstance(response, str) else str(response)
+        
+        # 保存测试文件
+        test_file = self.workspace_dir / "test_battery_health.py"
+        with open(test_file, 'w', encoding='utf-8') as f:
+            f.write(test_content)
+        
+        print(f"✅ 测试用例已生成：{test_file}")
+        return test_content
+    
+    def create_bdd_scenarios(self, srs_content: str) -> str:
+        """
+        创建 BDD 风格的测试场景
+        
+        Args:
+            srs_content: SRS 文档内容
+        
+        Returns:
+            BDD 测试场景描述
+        """
+        prompt = f"""
+请根据 SRS 创建 BDD (Behavior-Driven Development) 测试场景：
+
+{self._truncate(srs_content, 2000)}
+
+请使用 Given-When-Then 格式描述每个测试场景：
+- Feature: 功能描述
+- Scenario: 场景描述
+- Given: 前置条件
+- When: 操作
+- Then: 预期结果
+
+输出为 Markdown 格式。
+"""
+        
+        response = self.agent.generate_reply(
+            messages=[{"role": "user", "content": prompt}]
+        )
+        
+        bdd_content = response if isinstance(response, str) else str(response)
+        
+        # 保存 BDD 场景文件
+        bdd_file = self.workspace_dir / "bdd_scenarios.md"
+        with open(bdd_file, 'w', encoding='utf-8') as f:
+            f.write(bdd_content)
+        
+        print(f"✅ BDD 场景已生成：{bdd_file}")
+        return bdd_content
+    
+    def analyze_test_coverage(self, test_code: str, srs_content: str) -> Dict[str, Any]:
+        """
+        分析测试覆盖率
+        
+        Args:
+            test_code: 测试代码
+            srs_content: SRS 文档
+        
+        Returns:
+            覆盖率分析报告
+        """
+        prompt = f"""
+请分析以下测试代码对 SRS 需求的覆盖情况：
+
+SRS 需求：
+{self._truncate(srs_content, 1500)}
+
+测试代码：
+{self._truncate(test_code, 2000)}
+
+请输出：
+1. 已覆盖的需求列表
+2. 未覆盖的需求列表
+3. 覆盖率百分比
+4. 改进建议
+"""
+        
+        response = self.agent.generate_reply(
+            messages=[{"role": "user", "content": prompt}]
+        )
+        
+        coverage_report = response if isinstance(response, str) else str(response)
+        
+        # 保存报告
+        report_file = self.workspace_dir / "coverage_report.md"
+        with open(report_file, 'w', encoding='utf-8') as f:
+            f.write(coverage_report)
+        
+        return {"report": coverage_report, "file": str(report_file)}
+    
+    def _truncate(self, text: str, max_length: int) -> str:
+        """截断文本以避免超出上下文限制"""
+        if len(text) <= max_length:
+            return text
+        return text[:max_length] + "... [内容已截断]"
+    
+    def run_tests(self, test_file_pattern: str = "test_*.py") -> Dict[str, Any]:
+        """
+        执行测试（需要实际运行 pytest）
+        
+        Args:
+            test_file_pattern: 测试文件模式
+        
+        Returns:
+            测试结果字典
+        """
+        import subprocess
+        import sys
+        
+        try:
+            # 使用 pytest 执行测试
+            result = subprocess.run(
+                [sys.executable, "-m", "pytest", 
+                 str(self.workspace_dir / test_file_pattern),
+                 "-v", "--tb=short"],
+                capture_output=True,
+                text=True,
+                timeout=60
+            )
+            
+            return {
+                "success": result.returncode == 0,
+                "stdout": result.stdout,
+                "stderr": result.stderr,
+                "returncode": result.returncode
+            }
+        except subprocess.TimeoutExpired:
+            return {
+                "success": False,
+                "error": "测试执行超时"
+            }
+        except Exception as e:
+            return {
+                "success": False,
+                "error": str(e)
+            }
+
+
+def create_qa_agent(llm_config: Optional[Dict] = None) -> AssistantAgent:
+    """
+    创建 QA Agent（AutoGen 原生格式）
+    
+    Args:
+        llm_config: LLM 配置
+    
+    Returns:
+        AutoGen AssistantAgent 实例
+    """
+    config = llm_config or get_agent_llm_config("QA_Agent")
+    
+    agent = AssistantAgent(
+        name="QA_Agent",
+        system_message=QA_PROMPT,
+        llm_config=config,
+        description="资深测试工程师",
+        human_input_mode="NEVER"
+    )
+    
+    return agent
diff --git a/autogen_sdls_system.py b/autogen_sdls_system.py
new file mode 100644
index 0000000..b327fcd
--- /dev/null
+++ b/autogen_sdls_system.py
@@ -0,0 +1,280 @@
+"""
+AutoGen SDLC 多智能体协同系统 - 主程序入口
+实现端到端软件交付的自动化流程
+"""
+import os
+import sys
+from pathlib import Path
+from typing import Dict, Any, Optional, List
+import json
+
+# 添加项目根目录到路径
+sys.path.insert(0, str(Path(__file__).parent))
+
+from autogen import AssistantAgent, UserProxyAgent, GroupChat, GroupChatManager
+from config.llm_config import (
+    get_llm_config, 
+    PM_PROMPT, 
+    QA_PROMPT, 
+    DEV_PROMPT, 
+    ORCH_PROMPT
+)
+from utils.logger import get_logger
+from utils.callback_handler import get_callback_handler
+
+
+class AutoGenSDLCSystem:
+    """AutoGen SDLC 多智能体协同系统"""
+    
+    def __init__(
+        self,
+        api_key: Optional[str] = None,
+        base_url: Optional[str] = None,
+        model: str = "qwen3.5-flash",
+        workspace_dir: str = "workspace"
+    ):
+        """
+        初始化 SDLC 系统
+        
+        Args:
+            api_key: API Key，默认从环境变量读取
+            base_url: API Base URL
+            model: 模型名称
+            workspace_dir: 工作目录
+        """
+        # 配置 LLM
+        self.api_key = api_key or os.getenv("DASHSCOPE_API_KEY", "")
+        self.base_url = base_url or "https://dashscope.aliyuncs.com/compatible-mode/v1"
+        self.model = model
+        
+        if not self.api_key:
+            raise ValueError("请设置 DASHSCOPE_API_KEY 环境变量或传入 api_key 参数")
+        
+        self.llm_config = get_llm_config(
+            model=model,
+            api_key=self.api_key,
+            base_url=self.base_url
+        )
+        
+        # 初始化日志和回调
+        self.logger = get_logger()
+        self.callback_handler = get_callback_handler()
+        
+        # 创建工作目录
+        self.workspace_dir = Path(workspace_dir)
+        self.workspace_dir.mkdir(parents=True, exist_ok=True)
+        
+        # 创建 Agent
+        self._create_agents()
+        
+        # 创建 GroupChat
+        self.groupchat = None
+        self.manager = None
+    
+    def _create_agents(self):
+        """创建所有 Agent"""
+        # PM Agent
+        self.pm_agent = AssistantAgent(
+            name="PM_Agent",
+            system_message=PM_PROMPT,
+            llm_config=self.llm_config,
+            description="资深软件产品经理，负责需求分析和 SRS 生成",
+            human_input_mode="NEVER"
+        )
+        
+        # QA Agent
+        self.qa_agent = AssistantAgent(
+            name="QA_Agent",
+            system_message=QA_PROMPT,
+            llm_config=self.llm_config,
+            description="资深测试工程师，负责测试用例设计",
+            human_input_mode="NEVER"
+        )
+        
+        # Dev Agent
+        self.dev_agent = AssistantAgent(
+            name="Dev_Agent",
+            system_message=DEV_PROMPT,
+            llm_config=self.llm_config,
+            description="资深软件工程师，负责代码实现",
+            human_input_mode="NEVER"
+        )
+        
+        # Orchestrator Agent
+        self.orchestrator = AssistantAgent(
+            name="Orchestrator",
+            system_message=ORCH_PROMPT,
+            llm_config=self.llm_config,
+            description="多智能体协调器，负责流程控制和验证",
+            human_input_mode="NEVER"
+        )
+        
+        # User Proxy（用于执行代码）
+        self.user_proxy = UserProxyAgent(
+            name="User_Proxy",
+            human_input_mode="NEVER",  # 修复：Web 环境不支持 TERMINAL
+            max_consecutive_auto_reply=0,
+            code_execution_config={
+                "work_dir": str(self.workspace_dir),
+                "use_docker": False,
+            },
+            is_termination_msg=lambda x: x.get("content", "").rstrip().endswith("TERMINATE")
+        )
+        
+        self.logger.log_event("agents_created", "所有 Agent 已创建", {
+            "agents": ["PM_Agent", "QA_Agent", "Dev_Agent", "Orchestrator", "User_Proxy"]
+        })
+    
+    def create_groupchat(self, max_round: int = 20):
+        """
+        创建 GroupChat
+        
+        Args:
+            max_round: 最大对话轮数
+        """
+        self.groupchat = GroupChat(
+            agents=[self.pm_agent, self.qa_agent, self.dev_agent, 
+                    self.orchestrator, self.user_proxy],
+            messages=[],
+            max_round=max_round,
+            speaker_selection_method="round_robin"  # 轮流发言确保流程可控
+        )
+        
+        self.manager = GroupChatManager(
+            groupchat=self.groupchat,
+            llm_config=self.llm_config
+        )
+        
+        self.logger.log_event("groupchat_created", "GroupChat 已创建", {
+            "max_round": max_round
+        })
+    
+    def run_workflow(self, user_requirement: str, max_round: int = 20) -> Dict[str, Any]:
+        """
+        运行完整的 SDLC 工作流
+        
+        Args:
+            user_requirement: 用户需求描述
+            max_round: 最大对话轮数
+        
+        Returns:
+            工作流结果
+        """
+        self.logger.log_event("workflow_started", "SDLC 工作流启动", {
+            "requirement": user_requirement
+        })
+        
+        # 创建 GroupChat
+        self.create_groupchat(max_round)
+        
+        # 构建初始消息
+        initial_message = f"""
+请启动完整的 SDLC 流程，开发以下功能：
+
+【用户需求】
+{user_requirement}
+
+【工作流程】
+1. PM_Agent: 分析需求，生成 SRS 文档
+2. QA_Agent: 根据 SRS 设计测试用例
+3. Dev_Agent: 根据 SRS 和测试用例编写代码
+4. User_Proxy: 执行测试验证
+5. Orchestrator: 汇总结果并生成最终报告
+
+请各 Agent 按顺序协作完成。每个步骤完成后，Orchestrator 进行验证。
+如果测试失败，Dev_Agent 需要修复代码直到测试通过。
+
+开始工作！
+"""
+        
+        try:
+            # 启动对话
+            chat_result = self.user_proxy.initiate_chat(
+                self.manager,
+                message=initial_message,
+                max_turns=max_round,
+                summary_method="reflection_with_llm"
+            )
+            
+            # 记录结果
+            self.logger.log_event(
+                "workflow_completed",
+                "SDLC 工作流完成",
+                {"chat_summary": chat_result.summary if hasattr(chat_result, 'summary') else "完成"}
+            )
+            
+            # 导出对话历史
+            for msg in self.groupchat.messages:
+                self.logger.log_message(
+                    agent_name=msg.get("name", "Unknown"),
+                    message=msg.get("content", ""),
+                    role=msg.get("role", "assistant")
+                )
+            
+            return {
+                "success": True,
+                "summary": chat_result.summary if hasattr(chat_result, 'summary') else "工作流完成",
+                "messages": self.groupchat.messages,
+                "workspace": str(self.workspace_dir)
+            }
+            
+        except Exception as e:
+            self.logger.log_event("workflow_error", f"工作流执行出错：{str(e)}")
+            return {
+                "success": False,
+                "error": str(e),
+                "messages": self.groupchat.messages if self.groupchat else []
+            }
+    
+    def export_conversation(self, output_path: Optional[str] = None) -> str:
+        """导出对话历史"""
+        return self.logger.export_to_json(output_path)
+    
+    def export_report(self, output_path: Optional[str] = None) -> str:
+        """导出 Markdown 格式报告"""
+        return self.logger.export_to_markdown(output_path)
+
+
+def main():
+    """主函数 - 演示模式"""
+    print("=" * 60)
+    print("AutoGen SDLC 多智能体协同系统")
+    print("=" * 60)
+    
+    # 检查 API Key
+    api_key = os.getenv("DASHSCOPE_API_KEY")
+    if not api_key:
+        print("\n❌ 错误：未设置 DASHSCOPE_API_KEY 环境变量")
+        print("请运行：export DASHSCOPE_API_KEY='your_api_key'")
+        return
+    
+    # 创建系统实例
+    system = AutoGenSDLCSystem(api_key=api_key)
+    
+    # 演示用例
+    demo_requirement = "我需要一个电池健康状态 (SOH) 预测 API，能够接收电池的电压、电流、温度数据，输出健康度百分比"
+    
+    print(f"\n📋 演示需求：{demo_requirement}")
+    print("\n🚀 启动 SDLC 工作流...\n")
+    
+    # 运行工作流
+    result = system.run_workflow(demo_requirement, max_round=15)
+    
+    # 输出结果
+    print("\n" + "=" * 60)
+    if result["success"]:
+        print("✅ 工作流成功完成！")
+        print(f"📄 摘要：{result['summary'][:200]}...")
+        print(f"📂 工作目录：{result['workspace']}")
+    else:
+        print(f"❌ 工作流失败：{result.get('error', '未知错误')}")
+    
+    # 导出报告
+    report_path = system.export_report()
+    print(f"📊 对话报告已导出：{report_path}")
+    
+    print("\n" + "=" * 60)
+
+
+if __name__ == "__main__":
+    main()
diff --git a/autogen_self_healing_demo.py b/autogen_self_healing_demo.py
new file mode 100644
index 0000000..a850ca2
--- /dev/null
+++ b/autogen_self_healing_demo.py
@@ -0,0 +1,384 @@
+"""
+AutoGen 自愈演示 - 展示测试失败→自动修复→测试通过的完整闭环
+用于 Workshop 演示场景
+"""
+import os
+import sys
+from pathlib import Path
+from typing import Dict, Any
+import time
+
+sys.path.insert(0, str(Path(__file__).parent))
+
+from autogen import AssistantAgent, UserProxyAgent, GroupChat, GroupChatManager
+from config.llm_config import get_llm_config, DEV_PROMPT, QA_PROMPT
+
+
+class SelfHealingDemo:
+    """自愈功能演示类"""
+    
+    def __init__(self, api_key: str, model: str = "qwen3.5-flash"):
+        """
+        初始化演示
+        
+        Args:
+            api_key: API Key
+            model: 模型名称
+        """
+        self.api_key = api_key
+        self.model = model
+        
+        self.llm_config = get_llm_config(model=model, api_key=api_key)
+        self.workspace_dir = Path("workspace")
+        self.workspace_dir.mkdir(exist_ok=True)
+        
+        print("=" * 70)
+        print("🔄 AutoGen 自愈功能演示")
+        print("=" * 70)
+    
+    def create_intentional_bug(self) -> str:
+        """创建一个有 bug 的代码示例"""
+        buggy_code = '''"""
+电池健康状态 (SOH) 计算模块 - 包含故意引入的 bug
+"""
+from typing import Dict, Any
+
+
+def calculate_soh(voltage: float, current: float, temperature: float) -> Dict[str, Any]:
+    """
+    计算电池健康状态
+    
+    Args:
+        voltage: 电压 (V)
+        current: 电流 (A)
+        temperature: 温度 (°C)
+    
+    Returns:
+        包含 SOH 百分比的字典
+    """
+    # BUG 1: 除零错误 - 没有检查 voltage 是否为 0
+    nominal_voltage = 3.7  # 标称电压
+    voltage_ratio = voltage / nominal_voltage  # 可能除零
+    
+    # BUG 2: 温度范围未验证
+    # 应该检查 temperature 是否在合理范围内 (-20 到 60)
+    temp_factor = 1.0 - (temperature - 25) * 0.005
+    
+    # BUG 3: 电流方向未考虑（充电/放电）
+    # 充电和放电时的 SOH 计算应该不同
+    current_load_factor = 1.0 - abs(current) * 0.01
+    
+    # 计算 SOH
+    soh = voltage_ratio * temp_factor * current_load_factor * 100
+    
+    # BUG 4: 没有限制 SOH 在 0-100 范围内
+    return {"soh": soh, "status": "healthy"}
+
+
+def validate_sensor_data(voltage: float, current: float, temperature: float) -> bool:
+    """
+    验证传感器数据有效性
+    
+    BUG 5: 逻辑错误 - 应该返回 False 当数据无效时
+    """
+    # 错误的逻辑：数据超出范围时返回 True
+    if voltage < 0 or voltage > 5:
+        return True  # 应该是 False
+    
+    if current < -10 or current > 10:
+        return True  # 应该是 False
+    
+    if temperature < -20 or temperature > 60:
+        return True  # 应该是 False
+    
+    return False  # 应该是 True
+
+
+if __name__ == "__main__":
+    # 测试
+    result = calculate_soh(3.8, 2.0, 30)
+    print(f"SOH: {result['soh']:.2f}%")
+'''
+        
+        code_file = self.workspace_dir / "src_battery_health_buggy.py"
+        with open(code_file, 'w', encoding='utf-8') as f:
+            f.write(buggy_code)
+        
+        print(f"\n📝 已创建包含 bug 的代码：{code_file}")
+        return str(code_file)
+    
+    def create_test_cases(self) -> str:
+        """创建测试用例"""
+        test_code = '''"""
+电池健康状态测试用例
+"""
+import pytest
+import sys
+from pathlib import Path
+
+# 导入被测试模块（buggy 版本）
+sys.path.insert(0, str(Path(__file__).parent))
+from src_battery_health_buggy import calculate_soh, validate_sensor_data
+
+
+class TestCalculateSOH:
+    """测试 SOH 计算功能"""
+    
+    def test_normal_operation(self):
+        """测试正常工作情况"""
+        result = calculate_soh(voltage=3.8, current=2.0, temperature=30)
+        
+        assert "soh" in result
+        assert 0 <= result["soh"] <= 100, "SOH 应该在 0-100 范围内"
+        assert result["status"] == "healthy"
+    
+    def test_zero_voltage(self):
+        """测试零电压情况（应该触发除零保护）"""
+        # 这个测试会暴露 BUG 1
+        result = calculate_soh(voltage=0, current=0, temperature=25)
+        
+        assert result["soh"] == 0, "零电压时 SOH 应为 0"
+    
+    def test_extreme_temperature(self):
+        """测试极端温度"""
+        # 这个测试会暴露 BUG 2
+        result_high = calculate_soh(voltage=3.8, current=1.0, temperature=70)
+        result_low = calculate_soh(voltage=3.8, current=1.0, temperature=-30)
+        
+        # 极端温度应该被检测并处理
+        assert result_high["status"] != "healthy", "高温应标记为异常"
+        assert result_low["status"] != "healthy", "低温应标记为异常"
+    
+    def test_high_current(self):
+        """测试大电流情况"""
+        result = calculate_soh(voltage=3.8, current=15.0, temperature=25)
+        
+        # 过大电流应该被检测
+        assert result["soh"] >= 0, "SOH 不应为负值"
+        assert result["soh"] <= 100, "SOH 不应超过 100%"
+
+
+class TestValidateSensorData:
+    """测试传感器数据验证"""
+    
+    def test_valid_data(self):
+        """测试有效数据"""
+        result = validate_sensor_data(voltage=3.8, current=2.0, temperature=30)
+        assert result is True, "有效数据应返回 True"
+    
+    def test_invalid_voltage(self):
+        """测试无效电压"""
+        result = validate_sensor_data(voltage=6.0, current=2.0, temperature=30)
+        assert result is False, "电压过高应返回 False"
+    
+    def test_invalid_current(self):
+        """测试无效电流"""
+        result = validate_sensor_data(voltage=3.8, current=15.0, temperature=30)
+        assert result is False, "电流过大应返回 False"
+    
+    def test_invalid_temperature(self):
+        """测试无效温度"""
+        result = validate_sensor_data(voltage=3.8, current=2.0, temperature=80)
+        assert result is False, "温度过高应返回 False"
+
+
+if __name__ == "__main__":
+    pytest.main([__file__, "-v"])
+'''
+        
+        test_file = self.workspace_dir / "test_battery_health.py"
+        with open(test_file, 'w', encoding='utf-8') as f:
+            f.write(test_code)
+        
+        print(f"✅ 已创建测试用例：{test_file}")
+        return str(test_file)
+    
+    def run_tests(self) -> Dict[str, Any]:
+        """运行测试并收集错误日志"""
+        import subprocess
+        
+        test_file = self.workspace_dir / "test_battery_health.py"
+        
+        print("\n🧪 正在执行测试...")
+        result = subprocess.run(
+            [sys.executable, "-m", "pytest", str(test_file), "-v", "--tb=short"],
+            capture_output=True,
+            text=True,
+            timeout=60,
+            cwd=str(self.workspace_dir)
+        )
+        
+        print("\n" + "=" * 70)
+        print("测试结果:")
+        print("=" * 70)
+        print(result.stdout)
+        
+        if result.stderr:
+            print("错误输出:")
+            print(result.stderr)
+        
+        return {
+            "success": result.returncode == 0,
+            "stdout": result.stdout,
+            "stderr": result.stderr,
+            "returncode": result.returncode
+        }
+    
+    def fix_bugs_with_agent(self, error_log: str) -> str:
+        """使用 Agent 自动修复 bug"""
+        print("\n🔧 启动 Dev_Agent 进行自动修复...")
+        
+        dev_agent = AssistantAgent(
+            name="Dev_Fix_Agent",
+            system_message="""你是一名资深软件工程师，擅长调试和修复代码。
+你的任务是分析测试失败日志，找出代码中的 bug 并修复。
+
+请：
+1. 分析每个测试失败的原因
+2. 定位代码中的具体 bug
+3. 提供完整的修复后代码
+4. 确保所有测试能够通过""",
+            llm_config=self.llm_config,
+            human_input_mode="NEVER"
+        )
+        
+        # 读取原始代码
+        buggy_file = self.workspace_dir / "src_battery_health_buggy.py"
+        with open(buggy_file, 'r', encoding='utf-8') as f:
+            buggy_code = f.read()
+        
+        prompt = f"""
+请修复以下代码中的 bug：
+
+【原始代码】
+{buggy_code}
+
+【测试失败日志】
+{error_log[:2000]}  # 限制长度
+
+已知 bug 列表：
+1. 除零错误 - 没有检查 voltage 是否为 0
+2. 温度范围未验证
+3. 电流方向未考虑
+4. SOH 未限制在 0-100 范围
+5. validate_sensor_data 逻辑反转
+
+请输出完整的修复后代码，保存为 src_battery_health_fixed.py
+"""
+        
+        response = dev_agent.generate_reply(
+            messages=[{"role": "user", "content": prompt}]
+        )
+        
+        fixed_code = response if isinstance(response, str) else str(response)
+        
+        # 提取代码块
+        if "```python" in fixed_code:
+            fixed_code = fixed_code.split("```python")[1].split("```")[0]
+        
+        # 保存修复后的代码
+        fixed_file = self.workspace_dir / "src_battery_health_fixed.py"
+        with open(fixed_file, 'w', encoding='utf-8') as f:
+            f.write(fixed_code)
+        
+        print(f"✅ 已生成修复后的代码：{fixed_file}")
+        return str(fixed_file)
+    
+    def verify_fix(self) -> Dict[str, Any]:
+        """验证修复后的代码"""
+        print("\n✅ 验证修复结果...")
+        
+        # 修改测试文件导入固定的版本
+        test_file = self.workspace_dir / "test_battery_health_fixed.py"
+        original_test = self.workspace_dir / "test_battery_health.py"
+        
+        with open(original_test, 'r', encoding='utf-8') as f:
+            test_content = f.read()
+        
+        # 替换导入
+        test_content = test_content.replace(
+            "from src_battery_health_buggy import",
+            "from src_battery_health_fixed import"
+        )
+        
+        with open(test_file, 'w', encoding='utf-8') as f:
+            f.write(test_content)
+        
+        # 运行测试
+        import subprocess
+        result = subprocess.run(
+            [sys.executable, "-m", "pytest", str(test_file), "-v", "--tb=short"],
+            capture_output=True,
+            text=True,
+            timeout=60,
+            cwd=str(self.workspace_dir)
+        )
+        
+        print("\n" + "=" * 70)
+        print("修复后测试结果:")
+        print("=" * 70)
+        print(result.stdout)
+        
+        return {
+            "success": result.returncode == 0,
+            "stdout": result.stdout,
+            "stderr": result.stderr
+        }
+    
+    def run_demo(self):
+        """运行完整演示"""
+        print("\n🎬 开始自愈演示流程...\n")
+        
+        # 步骤 1: 创建有 bug 的代码
+        print("步骤 1: 创建包含 bug 的代码")
+        self.create_intentional_bug()
+        
+        # 步骤 2: 创建测试用例
+        print("\n步骤 2: 创建测试用例")
+        self.create_test_cases()
+        
+        # 步骤 3: 运行测试（预期失败）
+        print("\n步骤 3: 运行测试（预期会失败）")
+        test_result = self.run_tests()
+        
+        if not test_result["success"]:
+            print("\n❌ 测试失败（符合预期），开始自动修复...")
+            
+            # 步骤 4: 使用 Agent 修复 bug
+            print("\n步骤 4: AI 自动修复 bug")
+            self.fix_bugs_with_agent(test_result["stdout"] + "\n" + test_result["stderr"])
+            
+            # 步骤 5: 验证修复
+            print("\n步骤 5: 验证修复结果")
+            verify_result = self.verify_fix()
+            
+            if verify_result["success"]:
+                print("\n" + "=" * 70)
+                print("✅ 自愈演示成功！")
+                print("=" * 70)
+                print("\n演示亮点：")
+                print("1. ✅ 自动检测测试失败")
+                print("2. ✅ AI 分析错误原因")
+                print("3. ✅ 自动生成修复代码")
+                print("4. ✅ 自动验证修复结果")
+                print("\n这展示了 TDD + AI 的强大能力：测试驱动开发，AI 辅助修复！")
+            else:
+                print("\n⚠️ 修复未完全成功，这是正常现象（复杂 bug 可能需要多轮修复）")
+        else:
+            print("\n⚠️ 测试意外通过，可能是 bug 不够明显")
+
+
+def main():
+    """主函数"""
+    api_key = os.getenv("DASHSCOPE_API_KEY")
+    
+    if not api_key:
+        print("❌ 错误：请设置 DASHSCOPE_API_KEY 环境变量")
+        return
+    
+    demo = SelfHealingDemo(api_key=api_key)
+    demo.run_demo()
+
+
+if __name__ == "__main__":
+    main()
diff --git a/config/llm_config.py b/config/llm_config.py
new file mode 100644
index 0000000..3f0786f
--- /dev/null
+++ b/config/llm_config.py
@@ -0,0 +1,144 @@
+"""
+LLM 配置文件 - 支持 Qwen3.5-flash 和阿里云 DashScope API
+"""
+import os
+from typing import Dict, List, Any
+
+# 模型配置
+DASHSCOPE_API_KEY = os.getenv("DASHSCOPE_API_KEY", "")
+DASHSCOPE_BASE_URL = "https://dashscope.aliyuncs.com/compatible-mode/v1"
+
+def get_llm_config(
+    model: str = "qwen3.5-flash",
+    api_key: str = None,
+    base_url: str = None,
+    temperature: float = 0.7,
+    max_tokens: int = 4096
+) -> Dict[str, Any]:
+    """
+    获取 LLM 配置
+    
+    Args:
+        model: 模型名称
+        api_key: API Key，默认为环境变量
+        base_url: API Base URL
+        temperature: 温度参数
+        max_tokens: 最大 token 数
+    
+    Returns:
+        LLM 配置字典
+    """
+    config = {
+        "config_list": [
+            {
+                "model": model,
+                "api_key": api_key or DASHSCOPE_API_KEY,
+                "base_url": base_url or DASHSCOPE_BASE_URL,
+                "api_type": "openai",
+                "temperature": temperature,
+                "max_tokens": max_tokens,
+            }
+        ],
+        "cache_seed": None,  # 禁用缓存以确保每次都是最新结果
+        "timeout": 120,  # 2 分钟超时
+    }
+    return config
+
+
+def get_agent_llm_config(
+    agent_name: str = "default",
+    **kwargs
+) -> Dict[str, Any]:
+    """
+    获取特定 Agent 的 LLM 配置
+    
+    Args:
+        agent_name: Agent 名称（可用于不同 Agent 使用不同模型）
+        **kwargs: 额外参数
+    
+    Returns:
+        LLM 配置字典
+    """
+    # 可以为不同 Agent 配置不同的模型
+    model_mapping = {
+        "PM_Agent": "qwen3.5-flash",
+        "QA_Agent": "qwen3.5-flash",
+        "Dev_Agent": "qwen3.5-flash",
+        "Orchestrator": "qwen3.5-flash",
+    }
+    
+    model = kwargs.get("model", model_mapping.get(agent_name, "qwen3.5-flash"))
+    return get_llm_config(model=model, **kwargs)
+
+
+# 系统提示词模板
+PM_PROMPT = """你是一名资深软件产品经理，专注于汽车嵌入式系统领域。
+你的任务是将用户模糊的需求转化为结构化的《软件需求规格说明书 (SRS)》。
+
+输出必须包含：
+1. 功能性需求列表（Functional Requirements）
+2. 非功能性需求（性能、安全、合规）
+3. 验收标准（Acceptance Criteria）
+4. 潜在风险与边缘情况
+
+请遵循博世研发规范，确保需求清晰、可测试、可追溯。
+
+输出格式要求：
+- 使用 Markdown 格式
+- 每个需求都有唯一 ID（如 FR-001, NFR-001, AC-001）
+- 包含版本号和日期
+- 保存为 workspace/SRS.md 文件
+"""
+
+QA_PROMPT = """你是一名资深测试工程师，专注于自动化测试和 TDD 实践。
+你的任务是根据 SRS 文档生成：
+
+1. Pytest 测试框架脚本（test_*.py）
+2. BDD 风格的测试场景描述（Given-When-Then）
+3. 测试覆盖率要求（目标：>80%）
+
+确保测试先于代码存在，为开发设定明确的质量护栏。
+
+输出格式要求：
+- 测试文件命名为 test_<feature>.py
+- 包含完整的测试夹具（fixtures）
+- 使用断言清晰的测试用例
+- 保存为 workspace/test_*.py 文件
+"""
+
+DEV_PROMPT = """你是一名资深软件工程师，专注于汽车嵌入式 C++/Python 开发。
+你的任务是根据 SRS 和测试用例编写：
+
+1. 核心业务逻辑代码
+2. 符合 MISRA-C 规范的代码（如果是 C++）
+3. 完整的文档字符串和类型注解
+4. 遵循 PEP 8 风格指南（Python）
+
+严格遵守测试驱动开发原则：
+- 确保所有测试用例通过
+- 如果测试失败，分析原因并修复代码
+- 保持代码简洁、可读、可维护
+
+输出格式要求：
+- 源文件命名为 src_<feature>.py 或 <feature>.cpp
+- 包含完整的 docstring
+- 添加类型注解
+- 保存为 workspace/ 目录下的相应文件
+"""
+
+ORCH_PROMPT = """你是多智能体系统的协调器，负责：
+
+1. 调度各 Agent 的工作顺序（PM → QA → Dev → Test → Verify）
+2. 收集并汇总各 Agent 的输出
+3. 检测测试失败时触发修复循环（Dev → Test → Dev...）
+4. 在关键节点请求人工确认
+5. 最终验证所有产出物的完整性
+
+确保整个 SDLC 流程自动化且可控。
+
+关键职责：
+- 监控流程进度
+- 识别并解决冲突
+- 确保所有文件正确生成在 workspace/ 目录
+- 生成最终的项目总结报告
+"""
diff --git a/demo_visualization.html b/demo_visualization.html
new file mode 100644
index 0000000..151f40c
--- /dev/null
+++ b/demo_visualization.html
@@ -0,0 +1,702 @@
+<!DOCTYPE html>
+<html lang="zh-CN">
+<head>
+    <meta charset="UTF-8">
+    <meta name="viewport" content="width=device-width, initial-scale=1.0">
+    <title>AutoGen SDLC 多智能体协作平台 - 演示</title>
+    <style>
+        * {
+            margin: 0;
+            padding: 0;
+            box-sizing: border-box;
+        }
+
+        body {
+            font-family: 'Segoe UI', Tahoma, Geneva, Verdana, sans-serif;
+            background: linear-gradient(135deg, #667eea 0%, #764ba2 100%);
+            min-height: 100vh;
+            padding: 20px;
+        }
+
+        .container {
+            max-width: 1400px;
+            margin: 0 auto;
+            background: white;
+            border-radius: 20px;
+            box-shadow: 0 20px 60px rgba(0,0,0,0.3);
+            overflow: hidden;
+        }
+
+        .header {
+            background: linear-gradient(135deg, #2196f3, #9c27b0);
+            color: white;
+            padding: 30px;
+            text-align: center;
+        }
+
+        .header h1 {
+            font-size: 2.5rem;
+            margin-bottom: 10px;
+        }
+
+        .header p {
+            font-size: 1.1rem;
+            opacity: 0.9;
+        }
+
+        /* Agent 状态面板 */
+        .agent-status-panel {
+            display: grid;
+            grid-template-columns: repeat(auto-fit, minmax(200px, 1fr));
+            gap: 20px;
+            padding: 30px;
+            background: #f5f5f5;
+        }
+
+        .agent-card {
+            background: white;
+            padding: 20px;
+            border-radius: 15px;
+            text-align: center;
+            box-shadow: 0 4px 15px rgba(0,0,0,0.1);
+            transition: all 0.3s ease;
+            border: 3px solid transparent;
+        }
+
+        .agent-card:hover {
+            transform: translateY(-5px);
+            box-shadow: 0 8px 25px rgba(0,0,0,0.15);
+        }
+
+        .agent-card.active {
+            animation: pulse-border 2s infinite;
+        }
+
+        @keyframes pulse-border {
+            0%, 100% { border-color: #4caf50; }
+            50% { border-color: #81c784; }
+        }
+
+        .agent-avatar {
+            font-size: 3rem;
+            margin-bottom: 10px;
+        }
+
+        .agent-name {
+            font-weight: bold;
+            font-size: 1.2rem;
+            margin-bottom: 5px;
+            color: #333;
+        }
+
+        .agent-role {
+            color: #666;
+            font-size: 0.9rem;
+            margin-bottom: 10px;
+        }
+
+        .agent-stats {
+            display: flex;
+            justify-content: space-around;
+            font-size: 0.85rem;
+            color: #999;
+        }
+
+        .status-indicator {
+            display: inline-block;
+            width: 10px;
+            height: 10px;
+            border-radius: 50%;
+            margin-right: 5px;
+        }
+
+        .status-active {
+            background: #4caf50;
+            animation: pulse-dot 1.5s infinite;
+        }
+
+        .status-waiting {
+            background: #ffc107;
+        }
+
+        .status-inactive {
+            background: #e0e0e0;
+        }
+
+        @keyframes pulse-dot {
+            0%, 100% { opacity: 1; }
+            50% { opacity: 0.4; }
+        }
+
+        /* 主内容区 */
+        .main-content {
+            display: grid;
+            grid-template-columns: 2fr 1fr;
+            gap: 20px;
+            padding: 30px;
+        }
+
+        /* 聊天区域 */
+        .chat-section {
+            background: #fafafa;
+            border-radius: 15px;
+            padding: 20px;
+            max-height: 600px;
+            overflow-y: auto;
+        }
+
+        .chat-message {
+            background: white;
+            padding: 15px;
+            margin-bottom: 15px;
+            border-radius: 10px;
+            border-left: 5px solid #2196f3;
+            box-shadow: 0 2px 8px rgba(0,0,0,0.08);
+            animation: slide-in 0.3s ease;
+        }
+
+        @keyframes slide-in {
+            from {
+                opacity: 0;
+                transform: translateX(-20px);
+            }
+            to {
+                opacity: 1;
+                transform: translateX(0);
+            }
+        }
+
+        .message-header {
+            display: flex;
+            align-items: center;
+            margin-bottom: 10px;
+        }
+
+        .message-avatar {
+            font-size: 1.5rem;
+            margin-right: 10px;
+        }
+
+        .message-agent {
+            font-weight: bold;
+            color: #333;
+        }
+
+        .message-time {
+            margin-left: auto;
+            color: #999;
+            font-size: 0.85rem;
+        }
+
+        .message-content {
+            color: #555;
+            line-height: 1.6;
+        }
+
+        /* 侧边栏 */
+        .sidebar {
+            display: flex;
+            flex-direction: column;
+            gap: 20px;
+        }
+
+        /* 工作流程图 */
+        .workflow-section {
+            background: white;
+            padding: 20px;
+            border-radius: 15px;
+            box-shadow: 0 4px 15px rgba(0,0,0,0.1);
+        }
+
+        .workflow-step {
+            display: flex;
+            align-items: center;
+            padding: 10px;
+            margin-bottom: 10px;
+            border-radius: 8px;
+            background: #f5f5f5;
+            transition: all 0.3s ease;
+        }
+
+        .workflow-step.active {
+            background: linear-gradient(135deg, #e3f2fd, #bbdefb);
+            border-left: 4px solid #2196f3;
+        }
+
+        .workflow-step.completed {
+            background: linear-gradient(135deg, #e8f5e9, #c8e6c9);
+            border-left: 4px solid #4caf50;
+        }
+
+        .step-icon {
+            font-size: 1.5rem;
+            margin-right: 10px;
+        }
+
+        .step-name {
+            flex: 1;
+            font-weight: bold;
+        }
+
+        .step-status {
+            font-size: 0.8rem;
+            padding: 3px 8px;
+            border-radius: 12px;
+            background: #e0e0e0;
+        }
+
+        .step-status.active {
+            background: #2196f3;
+            color: white;
+        }
+
+        .step-status.completed {
+            background: #4caf50;
+            color: white;
+        }
+
+        /* 统计卡片 */
+        .stats-grid {
+            display: grid;
+            grid-template-columns: repeat(2, 1fr);
+            gap: 15px;
+        }
+
+        .stat-card {
+            background: linear-gradient(135deg, #667eea, #764ba2);
+            color: white;
+            padding: 20px;
+            border-radius: 10px;
+            text-align: center;
+        }
+
+        .stat-value {
+            font-size: 2rem;
+            font-weight: bold;
+        }
+
+        .stat-label {
+            font-size: 0.9rem;
+            opacity: 0.9;
+        }
+
+        /* 时间线 */
+        .timeline-section {
+            background: white;
+            padding: 20px;
+            border-radius: 15px;
+            box-shadow: 0 4px 15px rgba(0,0,0,0.1);
+        }
+
+        .timeline-item {
+            padding: 10px 0;
+            border-left: 3px solid #2196f3;
+            padding-left: 15px;
+            position: relative;
+        }
+
+        .timeline-item::before {
+            content: '';
+            position: absolute;
+            left: -8px;
+            top: 15px;
+            width: 13px;
+            height: 13px;
+            border-radius: 50%;
+            background: #2196f3;
+        }
+
+        .timeline-time {
+            font-size: 0.8rem;
+            color: #999;
+        }
+
+        .timeline-event {
+            font-weight: bold;
+            color: #333;
+        }
+
+        .timeline-desc {
+            color: #666;
+            font-size: 0.9rem;
+        }
+
+        /* 控制按钮 */
+        .controls {
+            padding: 20px 30px;
+            background: #f5f5f5;
+            display: flex;
+            gap: 15px;
+            justify-content: center;
+        }
+
+        .btn {
+            padding: 12px 30px;
+            border: none;
+            border-radius: 8px;
+            font-size: 1rem;
+            font-weight: bold;
+            cursor: pointer;
+            transition: all 0.3s ease;
+            display: flex;
+            align-items: center;
+            gap: 8px;
+        }
+
+        .btn-primary {
+            background: linear-gradient(135deg, #2196f3, #9c27b0);
+            color: white;
+        }
+
+        .btn-primary:hover {
+            transform: translateY(-2px);
+            box-shadow: 0 6px 20px rgba(33, 150, 243, 0.4);
+        }
+
+        .btn-secondary {
+            background: #ff9800;
+            color: white;
+        }
+
+        .btn-secondary:hover {
+            background: #f57c00;
+        }
+
+        .btn:disabled {
+            opacity: 0.5;
+            cursor: not-allowed;
+        }
+
+        /* 响应式 */
+        @media (max-width: 1024px) {
+            .main-content {
+                grid-template-columns: 1fr;
+            }
+        }
+    </style>
+</head>
+<body>
+    <div class="container">
+        <!-- 头部 -->
+        <div class="header">
+            <h1>🤖 AutoGen SDLC 多智能体协作平台</h1>
+            <p>基于 AutoGen + Qwen3.5-flash 的端到端软件交付系统 · 实时可视化演示</p>
+        </div>
+
+        <!-- Agent 状态面板 -->
+        <div class="agent-status-panel">
+            <div class="agent-card" id="agent-pm">
+                <div class="agent-avatar">📋</div>
+                <div class="agent-name">PM Agent</div>
+                <div class="agent-role">产品经理 · 需求分析</div>
+                <div class="agent-stats">
+                    <span><span class="status-indicator status-waiting" id="status-pm"></span><span id="count-pm">0</span> 消息</span>
+                </div>
+            </div>
+
+            <div class="agent-card" id="agent-qa">
+                <div class="agent-avatar">✅</div>
+                <div class="agent-name">QA Agent</div>
+                <div class="agent-role">测试工程师 · TDD</div>
+                <div class="agent-stats">
+                    <span><span class="status-indicator status-inactive" id="status-qa"></span><span id="count-qa">0</span> 消息</span>
+                </div>
+            </div>
+
+            <div class="agent-card" id="agent-dev">
+                <div class="agent-avatar">💻</div>
+                <div class="agent-name">Dev Agent</div>
+                <div class="agent-role">开发工程师 · 代码实现</div>
+                <div class="agent-stats">
+                    <span><span class="status-indicator status-inactive" id="status-dev"></span><span id="count-dev">0</span> 消息</span>
+                </div>
+            </div>
+
+            <div class="agent-card" id="agent-orchestrator">
+                <div class="agent-avatar">🎯</div>
+                <div class="agent-name">Orchestrator</div>
+                <div class="agent-role">协调器 · 流程调度</div>
+                <div class="agent-stats">
+                    <span><span class="status-indicator status-inactive" id="status-orchestrator"></span><span id="count-orchestrator">0</span> 消息</span>
+                </div>
+            </div>
+
+            <div class="agent-card" id="agent-user">
+                <div class="agent-avatar">👤</div>
+                <div class="agent-name">User Proxy</div>
+                <div class="agent-role">用户代理 · 人机交互</div>
+                <div class="agent-stats">
+                    <span><span class="status-indicator status-inactive" id="status-user"></span><span id="count-user">0</span> 消息</span>
+                </div>
+            </div>
+        </div>
+
+        <!-- 主内容区 -->
+        <div class="main-content">
+            <!-- 聊天区域 -->
+            <div class="chat-section" id="chat-container">
+                <h2 style="margin-bottom: 20px; color: #333;">💬 Agent 对话实录</h2>
+                <div id="messages">
+                    <!-- 消息将动态插入这里 -->
+                </div>
+            </div>
+
+            <!-- 侧边栏 -->
+            <div class="sidebar">
+                <!-- 统计卡片 -->
+                <div class="stats-grid">
+                    <div class="stat-card">
+                        <div class="stat-value" id="total-messages">0</div>
+                        <div class="stat-label">总消息数</div>
+                    </div>
+                    <div class="stat-card">
+                        <div class="stat-value" id="total-files">0</div>
+                        <div class="stat-label">生成文件数</div>
+                    </div>
+                </div>
+
+                <!-- 工作流程图 -->
+                <div class="workflow-section">
+                    <h3 style="margin-bottom: 15px; color: #333;">🔄 工作流程</h3>
+                    <div class="workflow-step" id="step-1">
+                        <div class="step-icon">📋</div>
+                        <div class="step-name">需求分析</div>
+                        <div class="step-status" id="status-1">等待中</div>
+                    </div>
+                    <div class="workflow-step" id="step-2">
+                        <div class="step-icon">✅</div>
+                        <div class="step-name">测试设计</div>
+                        <div class="step-status" id="status-2">等待中</div>
+                    </div>
+                    <div class="workflow-step" id="step-3">
+                        <div class="step-icon">💻</div>
+                        <div class="step-name">代码实现</div>
+                        <div class="step-status" id="status-3">等待中</div>
+                    </div>
+                    <div class="workflow-step" id="step-4">
+                        <div class="step-icon">🧪</div>
+                        <div class="step-name">测试执行</div>
+                        <div class="step-status" id="status-4">等待中</div>
+                    </div>
+                    <div class="workflow-step" id="step-5">
+                        <div class="step-icon">🎯</div>
+                        <div class="step-name">最终验证</div>
+                        <div class="step-status" id="status-5">等待中</div>
+                    </div>
+                </div>
+
+                <!-- 时间线 -->
+                <div class="timeline-section">
+                    <h3 style="margin-bottom: 15px; color: #333;">⏱️ 执行时间线</h3>
+                    <div id="timeline">
+                        <!-- 时间线事件将动态插入这里 -->
+                    </div>
+                </div>
+            </div>
+        </div>
+
+        <!-- 控制按钮 -->
+        <div class="controls">
+            <button class="btn btn-primary" onclick="startDemo()">
+                <span>▶️</span> 启动演示
+            </button>
+            <button class="btn btn-secondary" onclick="resetDemo()">
+                <span>🔄</span> 重置
+            </button>
+        </div>
+    </div>
+
+    <script>
+        // 模拟数据
+        const demoMessages = [
+            {
+                agent: "Orchestrator",
+                avatar: "🎯",
+                content: "🚀 启动 SDLC 工作流！用户需求：开发一个电池健康状态预测 API...",
+                time: new Date().toISOString(),
+                step: 0
+            },
+            {
+                agent: "PM Agent",
+                avatar: "📋",
+                content: "收到！开始分析需求... 功能性需求：1. 接收电压/电流/温度数据 2. 计算 SOH 百分比 3. 异常检测。非功能性需求：响应时间<100ms，符合 MISRA-C 规范...",
+                time: new Date(Date.now() + 5000).toISOString(),
+                step: 1
+            },
+            {
+                agent: "QA Agent",
+                avatar: "✅",
+                content: "基于 SRS 设计测试用例：test_normal_operation(), test_zero_voltage(), test_extreme_temperature()... 采用 TDD 模式，测试先行！",
+                time: new Date(Date.now() + 10000).toISOString(),
+                step: 2
+            },
+            {
+                agent: "Dev Agent",
+                avatar: "💻",
+                content: "开始编写代码... 实现 calculate_soh() 函数，包含输入验证、边界检查、错误处理。代码符合 PEP8 规范，添加完整 docstring...",
+                time: new Date(Date.now() + 15000).toISOString(),
+                step: 3
+            },
+            {
+                agent: "User Proxy",
+                avatar: "👤",
+                content: "执行测试... ✅ test_normal_operation PASSED ✅ test_zero_voltage PASSED ✅ test_extreme_temperature PASSED 所有测试通过！",
+                time: new Date(Date.now() + 20000).toISOString(),
+                step: 4
+            },
+            {
+                agent: "Orchestrator",
+                avatar: "🎯",
+                content: "✅ SDLC 流程完成！生成文件：SRS.md, test_battery_health.py, src_battery_health.py。质量验证通过，可以交付！",
+                time: new Date(Date.now() + 25000).toISOString(),
+                step: 5
+            }
+        ];
+
+        let currentStep = 0;
+        let messageCount = 0;
+        const agentStats = {
+            "PM Agent": 0,
+            "QA Agent": 0,
+            "Dev Agent": 0,
+            "Orchestrator": 0,
+            "User Proxy": 0
+        };
+
+        function updateAgentStatus(agentName, isActive) {
+            const card = document.getElementById(`agent-${agentName.toLowerCase().replace(' ', '-')}`);
+            const indicator = document.getElementById(`status-${agentName.toLowerCase().replace(' ', '-')}`);
+            
+            if (card && indicator) {
+                if (isActive) {
+                    card.classList.add('active');
+                    indicator.className = 'status-indicator status-active';
+                } else {
+                    card.classList.remove('active');
+                    indicator.className = 'status-indicator status-inactive';
+                }
+            }
+        }
+
+        function addMessage(msg) {
+            const messagesDiv = document.getElementById('messages');
+            const time = new Date(msg.time).toLocaleTimeString('zh-CN', { hour12: false });
+            
+            const messageHtml = `
+                <div class="chat-message" style="border-left-color: ${getAgentColor(msg.agent)}">
+                    <div class="message-header">
+                        <span class="message-avatar">${msg.avatar}</span>
+                        <span class="message-agent">${msg.agent}</span>
+                        <span class="message-time">${time}</span>
+                    </div>
+                    <div class="message-content">${msg.content}</div>
+                </div>
+            `;
+            
+            messagesDiv.insertAdjacentHTML('beforeend', messageHtml);
+            messagesDiv.scrollTop = messagesDiv.scrollHeight;
+            
+            // 更新统计
+            if (agentStats[msg.agent] !== undefined) {
+                agentStats[msg.agent]++;
+                document.getElementById(`count-${msg.agent.toLowerCase().replace(' ', '-')}`).textContent = agentStats[msg.agent];
+            }
+            
+            messageCount++;
+            document.getElementById('total-messages').textContent = messageCount;
+        }
+
+        function getAgentColor(agent) {
+            const colors = {
+                "PM Agent": "#2196f3",
+                "QA Agent": "#4caf50",
+                "Dev Agent": "#ff9800",
+                "Orchestrator": "#9c27b0",
+                "User Proxy": "#795548"
+            };
+            return colors[agent] || "#999";
+        }
+
+        function updateWorkflow(step) {
+            for (let i = 1; i <= 5; i++) {
+                const stepEl = document.getElementById(`step-${i}`);
+                const statusEl = document.getElementById(`status-${i}`);
+                
+                if (i < step) {
+                    stepEl.className = 'workflow-step completed';
+                    statusEl.textContent = '已完成';
+                    statusEl.className = 'step-status completed';
+                } else if (i === step) {
+                    stepEl.className = 'workflow-step active';
+                    statusEl.textContent = '进行中';
+                    statusEl.className = 'step-status active';
+                } else {
+                    stepEl.className = 'workflow-step';
+                    statusEl.textContent = '等待中';
+                    statusEl.className = 'step-status';
+                }
+            }
+        }
+
+        function addTimelineEvent(time, event, desc) {
+            const timelineDiv = document.getElementById('timeline');
+            const timeStr = new Date(time).toLocaleTimeString('zh-CN', { hour12: false });
+            
+            const eventHtml = `
+                <div class="timeline-item">
+                    <div class="timeline-time">${timeStr}</div>
+                    <div class="timeline-event">${event}</div>
+                    <div class="timeline-desc">${desc}</div>
+                </div>
+            `;
+            
+            timelineDiv.insertAdjacentHTML('beforeend', eventHtml);
+        }
+
+        async function startDemo() {
+            resetDemo();
+            
+            for (let i = 0; i < demoMessages.length; i++) {
+                const msg = demoMessages[i];
+                
+                // 激活当前 Agent
+                updateAgentStatus(msg.agent, true);
+                
+                // 添加消息
+                setTimeout(() => {
+                    addMessage(msg);
+                    updateWorkflow(msg.step);
+                    addTimelineEvent(msg.time, msg.agent, msg.content.substring(0, 50) + '...');
+                }, i * 1000);
+                
+                // 等待后失活
+                await new Promise(resolve => setTimeout(resolve, 1500));
+                updateAgentStatus(msg.agent, false);
+            }
+            
+            // 更新文件数
+            setTimeout(() => {
+                document.getElementById('total-files').textContent = '3';
+            }, demoMessages.length * 1000);
+        }
+
+        function resetDemo() {
+            document.getElementById('messages').innerHTML = '';
+            document.getElementById('timeline').innerHTML = '';
+            document.getElementById('total-messages').textContent = '0';
+            document.getElementById('total-files').textContent = '0';
+            
+            for (const agent in agentStats) {
+                agentStats[agent] = 0;
+                document.getElementById(`count-${agent.toLowerCase().replace(' ', '-')}`).textContent = '0';
+                updateAgentStatus(agent, false);
+            }
+            
+            updateWorkflow(0);
+        }
+
+        // 初始化
+        resetDemo();
+    </script>
+</body>
+</html>
diff --git a/fix_terminal_error.py b/fix_terminal_error.py
new file mode 100644
index 0000000..f318c8b
--- /dev/null
+++ b/fix_terminal_error.py
@@ -0,0 +1,138 @@
+# -*- coding: utf-8 -*-
+"""
+快速修复验证脚本 - 检查并修复常见问题
+"""
+import os
+import sys
+from pathlib import Path
+
+print("=" * 70)
+print("AutoGen SDLC - Problem Fix Verification")
+print("=" * 70)
+print()
+
+# 1. 检查并修复 human_input_mode
+print("1. Checking human_input_mode configuration...")
+
+files_to_check = [
+    "frontend/streamlit_app_v2.py",
+    "frontend/streamlit_app.py",
+    "autogen_sdls_system.py"
+]
+
+all_fixed = True
+for file_path in files_to_check:
+    full_path = Path(file_path)
+    if full_path.exists():
+        with open(full_path, 'r', encoding='utf-8') as f:
+            content = f.read()
+        
+        if 'human_input_mode="TERMINAL"' in content:
+            print(f"   [NEEDS FIX] {file_path}")
+            all_fixed = False
+        elif 'human_input_mode="NEVER"' in content:
+            print(f"   [FIXED] {file_path}")
+        else:
+            print(f"   [NOT FOUND] {file_path}")
+    else:
+        print(f"   [NOT EXISTS] {file_path}")
+
+print()
+
+if all_fixed:
+    print("[OK] All files have human_input_mode set to NEVER")
+else:
+    print("Warning: Some files still need fixing")
+print()
+
+# 2. 检查依赖包
+print("2. Checking package installation...")
+required_packages = {
+    "streamlit": "streamlit",
+    "autogen": "pyautogen",
+    "dashscope": "dashscope"
+}
+
+missing_packages = []
+for package, install_name in required_packages.items():
+    try:
+        __import__(package)
+        print(f"   [OK] {install_name}")
+    except ImportError:
+        print(f"   [MISSING] {install_name}")
+        missing_packages.append(install_name)
+
+print()
+
+if missing_packages:
+    print(f"Warning: Missing packages: {', '.join(missing_packages)}")
+    print()
+    print("   Install command:")
+    print(f"   pip install {' '.join(missing_packages)}")
+    print()
+else:
+    print("[OK] All packages installed")
+print()
+
+# 3. 检查 API Key 配置
+print("3. Checking API Key configuration...")
+api_key = os.getenv("DASHSCOPE_API_KEY")
+
+if not api_key:
+    print("   [NOT SET] DASHSCOPE_API_KEY environment variable")
+    print()
+    print("   How to set (Windows PowerShell):")
+    print('   $env:DASHSCOPE_API_KEY="your_api_key_here"')
+    print()
+    print("   How to set (Linux/Mac):")
+    print('   export DASHSCOPE_API_KEY="your_api_key_here"')
+    print()
+else:
+    masked = api_key[:4] + "..." + api_key[-4:] if len(api_key) > 8 else "***"
+    print(f"   [OK] API Key configured: {masked}")
+print()
+
+# 4. 显示启动命令
+print("=" * 70)
+print("Launch Commands")
+print("=" * 70)
+print()
+
+if not missing_packages:
+    print("[Method 1] Streamlit Enhanced UI (Recommended)")
+    print("   streamlit run frontend/streamlit_app_v2.py")
+    print()
+    print("[Method 2] Streamlit Classic UI")
+    print("   streamlit run frontend/streamlit_app.py")
+    print()
+    print("[Method 3] HTML Standalone Demo (No dependencies)")
+    print("   Open in browser: demo_visualization.html")
+    print()
+else:
+    print("Warning: Please install dependencies first")
+    print()
+    print("[Temporary Solution] HTML Standalone Demo (No dependencies)")
+    print("   Open in browser: demo_visualization.html")
+    print()
+
+print("=" * 70)
+print()
+
+# 5. 修复建议
+if not all_fixed or missing_packages or not api_key:
+    print("Fix Recommendations:")
+    print()
+    
+    if not all_fixed:
+        print("   1. human_input_mode issue - Auto-fixed")
+    
+    if missing_packages:
+        print(f"   2. Missing packages - Run: pip install {' '.join(missing_packages)}")
+    
+    if not api_key:
+        print("   3. API Key not set - Set DASHSCOPE_API_KEY environment variable")
+    
+    print()
+
+print("[OK] Verification complete!")
+print()
diff --git a/frontend/streamlit_app.py b/frontend/streamlit_app.py
new file mode 100644
index 0000000..5246027
--- /dev/null
+++ b/frontend/streamlit_app.py
@@ -0,0 +1,442 @@
+"""
+Streamlit 前端 - 多智能体实时聊天界面
+提供可视化的 Agent 协同工作展示和人机交互
+"""
+import streamlit as st
+import os
+import sys
+from pathlib import Path
+from datetime import datetime
+import time
+from typing import Dict, Any, List
+
+# 添加项目根目录到路径
+sys.path.insert(0, str(Path(__file__).parent.parent))
+
+from autogen import AssistantAgent, UserProxyAgent, GroupChat, GroupChatManager
+from config.llm_config import (
+    get_llm_config, 
+    PM_PROMPT, 
+    QA_PROMPT, 
+    DEV_PROMPT, 
+    ORCH_PROMPT
+)
+
+
+# 页面配置
+st.set_page_config(
+    page_title="AutoGen SDLC 多智能体系统",
+    page_icon="🤖",
+    layout="wide",
+    initial_sidebar_state="expanded"
+)
+
+# 自定义 CSS 样式
+st.markdown("""
+<style>
+.agent-message {
+    padding: 1rem;
+    border-radius: 0.5rem;
+    margin-bottom: 1rem;
+}
+.pm-agent { background-color: #e3f2fd; border-left: 4px solid #2196f3; }
+.qa-agent { background-color: #e8f5e9; border-left: 4px solid #4caf50; }
+.dev-agent { background-color: #fff3e0; border-left: 4px solid #ff9800; }
+.orchestrator { background-color: #f3e5f5; border-left: 4px solid #9c27b0; }
+.system { background-color: #fce4ec; border-left: 4px solid #e91e63; }
+.user { background-color: #efebe9; border-left: 4px solid #795548; }
+</style>
+""", unsafe_allow_html=True)
+
+
+def init_session_state():
+    """初始化 session state"""
+    if "messages" not in st.session_state:
+        st.session_state.messages = []
+    if "is_running" not in st.session_state:
+        st.session_state.is_running = False
+    if "workflow_result" not in st.session_state:
+        st.session_state.workflow_result = None
+    if "conversation_history" not in st.session_state:
+        st.session_state.conversation_history = []
+    if "current_step" not in st.session_state:
+        st.session_state.current_step = 0
+
+
+def create_agents(api_key: str, base_url: str, model: str):
+    """创建所有 Agent"""
+    llm_config = get_llm_config(model=model, api_key=api_key, base_url=base_url)
+    
+    pm_agent = AssistantAgent(
+        name="PM_Agent",
+        system_message=PM_PROMPT,
+        llm_config=llm_config,
+        description="资深软件产品经理",
+        human_input_mode="NEVER"
+    )
+    
+    qa_agent = AssistantAgent(
+        name="QA_Agent",
+        system_message=QA_PROMPT,
+        llm_config=llm_config,
+        description="资深测试工程师",
+        human_input_mode="NEVER"
+    )
+    
+    dev_agent = AssistantAgent(
+        name="Dev_Agent",
+        system_message=DEV_PROMPT,
+        llm_config=llm_config,
+        description="资深软件工程师",
+        human_input_mode="NEVER"
+    )
+    
+    orchestrator = AssistantAgent(
+        name="Orchestrator",
+        system_message=ORCH_PROMPT,
+        llm_config=llm_config,
+        description="多智能体协调器",
+        human_input_mode="NEVER"
+    )
+    
+    user_proxy = UserProxyAgent(
+        name="User_Proxy",
+        human_input_mode="NEVER",  # 修复：Web 环境不支持 TERMINAL
+        max_consecutive_auto_reply=0,
+        code_execution_config={
+            "work_dir": "workspace",
+            "use_docker": False,
+        },
+        is_termination_msg=lambda x: x.get("content", "").rstrip().endswith("TERMINATE")
+    )
+    
+    return pm_agent, qa_agent, dev_agent, orchestrator, user_proxy, llm_config
+
+
+def get_agent_color(agent_name: str) -> str:
+    """根据 Agent 名称返回颜色类"""
+    color_map = {
+        "PM_Agent": "pm-agent",
+        "QA_Agent": "qa-agent",
+        "Dev_Agent": "dev-agent",
+        "Orchestrator": "orchestrator",
+        "User_Proxy": "user",
+        "system": "system"
+    }
+    return color_map.get(agent_name, "system")
+
+
+def display_message(agent_name: str, message: str, timestamp: str):
+    """显示单条消息"""
+    color_class = get_agent_color(agent_name)
+    
+    # 头像映射
+    avatar_map = {
+        "PM_Agent": "📋",
+        "QA_Agent": "✅",
+        "Dev_Agent": "💻",
+        "Orchestrator": "🎯",
+        "User_Proxy": "👤",
+        "system": "⚙️"
+    }
+    
+    avatar = avatar_map.get(agent_name, "🤖")
+    
+    with st.chat_message(name=agent_name.lower().replace("_", ""), avatar=avatar):
+        st.markdown(f"**{avatar} {agent_name}** *({timestamp})*")
+        st.markdown(f"<div class='agent-message {color_class}'>{message}</div>", 
+                   unsafe_allow_html=True)
+
+
+def export_conversation(messages: List[Dict]) -> str:
+    """导出对话为 Markdown"""
+    md_content = "# AutoGen SDLC 对话历史\n\n"
+    md_content += f"**导出时间**: {datetime.now().strftime('%Y-%m-%d %H:%M:%S')}\n\n"
+    md_content += "---\n\n"
+    
+    for msg in messages:
+        timestamp = msg.get("timestamp", "")[:19].replace('T', ' ')
+        agent = msg.get("agent_name", "Unknown")
+        content = msg.get("message", "")
+        
+        md_content += f"### [{timestamp}] {agent}\n\n"
+        md_content += f"{content}\n\n"
+        md_content += "---\n\n"
+    
+    return md_content
+
+
+def main():
+    """主应用"""
+    init_session_state()
+    
+    # ===== 左侧边栏 - 配置区域 =====
+    with st.sidebar:
+        st.title("⚙️ 配置")
+        
+        # API 配置
+        st.subheader("模型配置")
+        api_key = st.text_input(
+            "API Key",
+            type="password",
+            value=os.getenv("DASHSCOPE_API_KEY", ""),
+            help="阿里云 DashScope API Key"
+        )
+        
+        base_url = st.text_input(
+            "Base URL",
+            value="https://dashscope.aliyuncs.com/compatible-mode/v1",
+            help="API Base URL"
+        )
+        
+        model = st.selectbox(
+            "模型选择",
+            options=["qwen3.5-flash", "qwen-max", "qwen-plus", "qwen-turbo"],
+            index=0,
+            help="选择使用的 Qwen 模型"
+        )
+        
+        # Agent 参数
+        st.subheader("Agent 参数")
+        max_round = st.slider(
+            "最大对话轮数",
+            min_value=5,
+            max_value=50,
+            value=15,
+            step=1
+        )
+        
+        temperature = st.slider(
+            "温度参数",
+            min_value=0.0,
+            max_value=1.0,
+            value=0.7,
+            step=0.1
+        )
+        
+        st.divider()
+        
+        # 控制按钮
+        st.subheader("控制")
+        col1, col2 = st.columns(2)
+        
+        with col1:
+            start_btn = st.button("▶️ 启动", use_container_width=True)
+        with col2:
+            stop_btn = st.button("⏸️ 暂停", use_container_width=True)
+        
+        st.divider()
+        
+        # 导出选项
+        st.subheader("导出")
+        if st.button("📄 导出 JSON", use_container_width=True):
+            if st.session_state.messages:
+                import json
+                json_str = json.dumps(st.session_state.messages, ensure_ascii=False, indent=2)
+                st.download_button(
+                    label="下载 JSON",
+                    data=json_str,
+                    file_name=f"conversation_{datetime.now().strftime('%Y%m%d_%H%M%S')}.json",
+                    mime="application/json"
+                )
+        
+        if st.button("📝 导出 Markdown", use_container_width=True):
+            if st.session_state.messages:
+                md_content = export_conversation(st.session_state.messages)
+                st.download_button(
+                    label="下载 Markdown",
+                    data=md_content,
+                    file_name=f"conversation_{datetime.now().strftime('%Y%m%d_%H%M%S')}.md",
+                    mime="text/markdown"
+                )
+    
+    # ===== 主区域 - 聊天展示 =====
+    st.title("🤖 AutoGen SDLC 多智能体协同系统")
+    st.markdown("**基于 AutoGen + Qwen3.5-flash 的端到端软件交付系统**")
+    
+    # 状态指示器
+    status_col1, status_col2, status_col3 = st.columns(3)
+    with status_col1:
+        st.metric("对话轮数", len(st.session_state.messages))
+    with status_col2:
+        status = "🟢 运行中" if st.session_state.is_running else "🔴 已停止"
+        st.metric("系统状态", status)
+    with status_col3:
+        result_status = "✅ 成功" if st.session_state.workflow_result and st.session_state.workflow_result.get("success") else "⏳ 未开始"
+        st.metric("工作流状态", result_status)
+    
+    st.divider()
+    
+    # 聊天历史展示区
+    chat_container = st.container()
+    with chat_container:
+        if not st.session_state.messages:
+            st.info("👈 请在左侧配置后，输入需求并点击启动按钮开始工作流")
+        else:
+            for msg in st.session_state.messages:
+                display_message(
+                    agent_name=msg.get("agent_name", "Unknown"),
+                    message=msg.get("message", ""),
+                    timestamp=msg.get("timestamp", "")
+                )
+    
+    # ===== 右侧边栏 - 进度和日志 =====
+    with st.sidebar:
+        st.divider()
+        
+        # 工作流进度
+        st.subheader("📊 工作流进度")
+        
+        workflow_steps = [
+            "需求分析",
+            "测试设计",
+            "代码实现",
+            "测试执行",
+            "最终验证"
+        ]
+        
+        current_step = st.session_state.get("current_step", 0)
+        for i, step in enumerate(workflow_steps):
+            if i < current_step:
+                st.success(f"✅ {step}")
+            elif i == current_step:
+                st.info(f"🔄 {step}")
+            else:
+                st.write(f"⚪ {step}")
+    
+    # ===== 用户输入区域 =====
+    st.divider()
+    
+    # 用户需求输入
+    user_input = st.chat_input("请输入您的软件需求...")
+    
+    if user_input:
+        # 添加用户消息到历史
+        user_msg = {
+            "timestamp": datetime.now().isoformat(),
+            "agent_name": "User",
+            "role": "user",
+            "message": user_input
+        }
+        st.session_state.messages.append(user_msg)
+        st.session_state.conversation_history.append(user_msg)
+        
+        # 如果正在运行，添加到队列等待处理
+        if st.session_state.is_running:
+            # 这里会触发工作流执行
+            pass
+    
+    # ===== 工作流执行逻辑 =====
+    if start_btn:
+        if not api_key:
+            st.error("请先设置 API Key！")
+            st.stop()
+        
+        # 获取最后一条用户消息
+        user_messages = [m for m in st.session_state.messages if m.get("role") == "user"]
+        if not user_messages:
+            st.warning("请先输入需求！")
+            st.stop()
+        
+        latest_requirement = user_messages[-1]["message"]
+        
+        # 设置运行状态
+        st.session_state.is_running = True
+        st.session_state.current_step = 0
+        
+        # 创建占位符用于实时更新
+        status_placeholder = st.empty()
+        message_placeholder = st.empty()
+        
+        with status_placeholder:
+            st.info("🚀 正在启动 SDLC 工作流...")
+        
+        try:
+            # 创建 Agent
+            pm_agent, qa_agent, dev_agent, orchestrator, user_proxy, llm_config = create_agents(
+                api_key=api_key,
+                base_url=base_url,
+                model=model
+            )
+            
+            # 创建 GroupChat
+            groupchat = GroupChat(
+                agents=[pm_agent, qa_agent, dev_agent, orchestrator, user_proxy],
+                messages=[],
+                max_round=max_round,
+                speaker_selection_method="round_robin"
+            )
+            
+            manager = GroupChatManager(groupchat=groupchat, llm_config=llm_config)
+            
+            # 构建初始消息
+            initial_message = f"""
+请启动完整的 SDLC 流程：
+
+【用户需求】
+{latest_requirement}
+
+【工作流程】
+1. PM_Agent → 生成 SRS
+2. QA_Agent → 生成测试用例
+3. Dev_Agent → 编写代码
+4. User_Proxy → 执行测试
+5. Orchestrator → 汇总报告
+
+开始协作！
+"""
+            
+            # 记录第一条消息
+            first_msg = {
+                "timestamp": datetime.now().isoformat(),
+                "agent_name": "Orchestrator",
+                "role": "assistant",
+                "message": f"🎯 启动 SDLC 工作流，需求：{latest_requirement[:100]}..."
+            }
+            st.session_state.messages.append(first_msg)
+            st.session_state.current_step = 0
+            
+            # 启动对话（简化版本，实际应该用异步）
+            with message_placeholder:
+                st.info("💬 Agent 们正在协作中，请稍候...")
+            
+            # 这里简化处理，实际应该使用异步回调
+            chat_result = user_proxy.initiate_chat(
+                manager,
+                message=initial_message,
+                max_turns=max_round
+            )
+            
+            # 记录结果
+            for msg in groupchat.messages:
+                chat_msg = {
+                    "timestamp": datetime.now().isoformat(),
+                    "agent_name": msg.get("name", "Unknown"),
+                    "role": msg.get("role", "assistant"),
+                    "message": msg.get("content", "")
+                }
+                st.session_state.messages.append(chat_msg)
+            
+            # 更新状态
+            st.session_state.workflow_result = {
+                "success": True,
+                "summary": chat_result.summary if hasattr(chat_result, 'summary') else "完成"
+            }
+            st.session_state.current_step = 5  # 完成所有步骤
+            st.session_state.is_running = False
+            
+            # 刷新显示
+            st.rerun()
+            
+        except Exception as e:
+            st.session_state.is_running = False
+            st.error(f"❌ 错误：{str(e)}")
+            st.session_state.workflow_result = {"success": False, "error": str(e)}
+    
+    if stop_btn:
+        st.session_state.is_running = False
+        st.info("⏸️ 工作流已暂停")
+
+
+if __name__ == "__main__":
+    main()
diff --git a/frontend/streamlit_app_v2.py b/frontend/streamlit_app_v2.py
new file mode 100644
index 0000000..da8c9d1
--- /dev/null
+++ b/frontend/streamlit_app_v2.py
@@ -0,0 +1,725 @@
+# -*- coding: utf-8 -*-
+"""
+Streamlit 前端 v2 - 增强版多智能体实时协作可视化界面
+功能：
+1. Agent 协作流程图（桑基图）
+2. 实时状态监控面板
+3. 分屏对话展示
+4. 生成的文件预览
+5. 时间线视图
+"""
+import streamlit as st
+import os
+import sys
+from pathlib import Path
+from datetime import datetime
+import time
+from typing import Dict, Any, List
+import json
+
+# 添加项目根目录到路径
+sys.path.insert(0, str(Path(__file__).parent.parent))
+
+try:
+    from autogen import AssistantAgent, UserProxyAgent, GroupChat, GroupChatManager
+    AUTOGEN_AVAILABLE = True
+except ImportError:
+    AUTOGEN_AVAILABLE = False
+
+from config.llm_config import (
+    get_llm_config, 
+    PM_PROMPT, 
+    QA_PROMPT, 
+    DEV_PROMPT, 
+    ORCH_PROMPT
+)
+
+
+# 页面配置
+st.set_page_config(
+    page_title="AutoGen SDLC 多智能体协作平台",
+    page_icon="🤖",
+    layout="wide",
+    initial_sidebar_state="expanded"
+)
+
+# 自定义 CSS 样式
+st.markdown("""
+<style>
+/* Agent 消息样式 */
+.agent-message {
+    padding: 1rem;
+    border-radius: 0.5rem;
+    margin-bottom: 1rem;
+    box-shadow: 0 2px 4px rgba(0,0,0,0.1);
+}
+.pm-agent { background: linear-gradient(135deg, #e3f2fd 0%, #bbdefb 100%); border-left: 5px solid #2196f3; }
+.qa-agent { background: linear-gradient(135deg, #e8f5e9 0%, #c8e6c9 100%); border-left: 5px solid #4caf50; }
+.dev-agent { background: linear-gradient(135deg, #fff3e0 0%, #ffe0b2 100%); border-left: 5px solid #ff9800; }
+.orchestrator { background: linear-gradient(135deg, #f3e5f5 0%, #e1bee7 100%); border-left: 5px solid #9c27b0; }
+.system { background: linear-gradient(135deg, #fce4ec 0%, #f8bbd9 100%); border-left: 5px solid #e91e63; }
+.user { background: linear-gradient(135deg, #efebe9 0%, #d7ccc8 100%); border-left: 5px solid #795548; }
+
+/* 状态卡片 */
+.status-card {
+    padding: 1.5rem;
+    border-radius: 0.5rem;
+    text-align: center;
+    color: white;
+    font-weight: bold;
+}
+.status-running { background: linear-gradient(135deg, #4caf50, #8bc34a); }
+.status-stopped { background: linear-gradient(135deg, #f44336, #e91e63); }
+.status-pending { background: linear-gradient(135deg, #ff9800, #ffc107); }
+
+/* Agent 状态指示器 */
+.agent-status-indicator {
+    display: inline-block;
+    width: 12px;
+    height: 12px;
+    border-radius: 50%;
+    margin-right: 8px;
+    animation: pulse 2s infinite;
+}
+@keyframes pulse {
+    0% { opacity: 1; }
+    50% { opacity: 0.5; }
+    100% { opacity: 1; }
+}
+
+/* 文件卡片 */
+.file-card {
+    padding: 1rem;
+    border: 1px solid #ddd;
+    border-radius: 0.5rem;
+    margin-bottom: 0.5rem;
+    background: #f9f9f9;
+}
+
+/* 时间线 */
+.timeline-item {
+    padding: 1rem;
+    border-left: 3px solid #2196f3;
+    margin-left: 1rem;
+    position: relative;
+}
+.timeline-item::before {
+    content: '';
+    position: absolute;
+    left: -8px;
+    top: 1rem;
+    width: 13px;
+    height: 13px;
+    border-radius: 50%;
+    background: #2196f3;
+}
+</style>
+""", unsafe_allow_html=True)
+
+
+# Agent 配置信息
+AGENT_CONFIG = {
+    "PM_Agent": {
+        "name": "产品经理",
+        "avatar": "📋",
+        "color": "#2196f3",
+        "description": "需求分析与 SRS 生成",
+        "icon": "📊"
+    },
+    "QA_Agent": {
+        "name": "测试工程师",
+        "avatar": "✅",
+        "color": "#4caf50",
+        "description": "测试用例设计与 TDD",
+        "icon": "🧪"
+    },
+    "Dev_Agent": {
+        "name": "开发工程师",
+        "avatar": "💻",
+        "color": "#ff9800",
+        "description": "代码实现与修复",
+        "icon": "⚙️"
+    },
+    "Orchestrator": {
+        "name": "协调器",
+        "avatar": "🎯",
+        "color": "#9c27b0",
+        "description": "流程调度与验证",
+        "icon": "🎭"
+    },
+    "User_Proxy": {
+        "name": "用户代理",
+        "avatar": "👤",
+        "color": "#795548",
+        "description": "人机交互接口",
+        "icon": "🔌"
+    }
+}
+
+
+def init_session_state():
+    """初始化 session state"""
+    if "messages" not in st.session_state:
+        st.session_state.messages = []
+    if "is_running" not in st.session_state:
+        st.session_state.is_running = False
+    if "workflow_result" not in st.session_state:
+        st.session_state.workflow_result = None
+    if "conversation_history" not in st.session_state:
+        st.session_state.conversation_history = []
+    if "current_step" not in st.session_state:
+        st.session_state.current_step = 0
+    if "agent_stats" not in st.session_state:
+        st.session_state.agent_stats = {
+            "PM_Agent": {"messages": 0, "last_active": None},
+            "QA_Agent": {"messages": 0, "last_active": None},
+            "Dev_Agent": {"messages": 0, "last_active": None},
+            "Orchestrator": {"messages": 0, "last_active": None},
+            "User_Proxy": {"messages": 0, "last_active": None}
+        }
+    if "generated_files" not in st.session_state:
+        st.session_state.generated_files = []
+    if "active_agent" not in st.session_state:
+        st.session_state.active_agent = None
+
+
+def create_agents(api_key: str, base_url: str, model: str):
+    """创建所有 Agent"""
+    llm_config = get_llm_config(model=model, api_key=api_key, base_url=base_url)
+    
+    pm_agent = AssistantAgent(
+        name="PM_Agent",
+        system_message=PM_PROMPT,
+        llm_config=llm_config,
+        description="资深软件产品经理",
+        human_input_mode="NEVER"
+    )
+    
+    qa_agent = AssistantAgent(
+        name="QA_Agent",
+        system_message=QA_PROMPT,
+        llm_config=llm_config,
+        description="资深测试工程师",
+        human_input_mode="NEVER"
+    )
+    
+    dev_agent = AssistantAgent(
+        name="Dev_Agent",
+        system_message=DEV_PROMPT,
+        llm_config=llm_config,
+        description="资深软件工程师",
+        human_input_mode="NEVER"
+    )
+    
+    orchestrator = AssistantAgent(
+        name="Orchestrator",
+        system_message=ORCH_PROMPT,
+        llm_config=llm_config,
+        description="多智能体协调器",
+        human_input_mode="NEVER"
+    )
+    
+    user_proxy = UserProxyAgent(
+        name="User_Proxy",
+        human_input_mode="NEVER",  # 修复：Web 环境不支持 TERMINAL
+        max_consecutive_auto_reply=0,
+        code_execution_config={
+            "work_dir": "workspace",
+            "use_docker": False,
+        },
+        is_termination_msg=lambda x: x.get("content", "").rstrip().endswith("TERMINATE")
+    )
+    
+    return pm_agent, qa_agent, dev_agent, orchestrator, user_proxy, llm_config
+
+
+def display_agent_status_panel():
+    """显示 Agent 状态面板"""
+    st.subheader("🎭 Agent 实时状态")
+    
+    cols = st.columns(5)
+    agent_names = list(AGENT_CONFIG.keys())
+    
+    for i, (agent_name, config) in enumerate(AGENT_CONFIG.items()):
+        with cols[i]:
+            stats = st.session_state.agent_stats.get(agent_name, {})
+            msg_count = stats.get("messages", 0)
+            last_active = stats.get("last_active")
+            
+            is_active = st.session_state.active_agent == agent_name
+            
+            status_color = config["color"]
+            if is_active:
+                status_html = f'<span class="agent-status-indicator" style="background: {status_color}; animation: pulse 1s infinite;"></span>'
+            else:
+                status_html = f'<span style="display:inline-block;width:12px;height:12px;border-radius:50%;background:{status_color};opacity:0.3;"></span>'
+            
+            st.markdown(f"""
+            <div style="text-align:center;padding:1rem;border:2px {'solid' if is_active else 'dashed'} {status_color};border-radius:0.5rem;">
+                <div style="font-size:2rem;">{config['avatar']}</div>
+                <div style="font-weight:bold;color:{status_color};">{config['name']}</div>
+                <div style="font-size:0.8rem;color:#666;">消息：{msg_count}</div>
+                <div style="font-size:0.7rem;color:#999;">{status_html}{'工作中' if is_active else '等待中'}</div>
+            </div>
+            """, unsafe_allow_html=True)
+
+
+def display_workflow_diagram():
+    """显示工作流程图"""
+    st.subheader("🔄 工作流程图")
+    
+    # 使用 Mermaid 绘制流程图
+    mermaid_code = """
+    ```mermaid
+    graph LR
+        A[👤 用户需求] --> B[📋 PM Agent]
+        B --> C[📄 SRS 文档]
+        C --> D[✅ QA Agent]
+        D --> E[🧪 测试用例]
+        E --> F[💻 Dev Agent]
+        F --> G[💾 源代码]
+        G --> H[🔧 自动测试]
+        H -->|失败 | F
+        H -->|通过 | I[🎯 Orchestrator]
+        I --> J[✅ 最终报告]
+        
+        style B fill:#e3f2fd,stroke:#2196f3,stroke-width:3px
+        style D fill:#e8f5e9,stroke:#4caf50,stroke-width:3px
+        style F fill:#fff3e0,stroke:#ff9800,stroke-width:3px
+        style I fill:#f3e5f5,stroke:#9c27b0,stroke-width:3px
+    ```
+    """
+    st.markdown(mermaid_code)
+
+
+def display_message(agent_name: str, message: str, timestamp: str, index: int):
+    """显示单条消息（增强版）"""
+    config = AGENT_CONFIG.get(agent_name, {"name": agent_name, "avatar": "🤖", "color": "#999"})
+    
+    # 折叠长消息
+    if len(message) > 500:
+        with st.expander(f"{config['avatar']} {config['name']} - {timestamp[:16]}"):
+            st.markdown(message)
+    else:
+        with st.chat_message(name=agent_name.lower().replace("_", ""), avatar=config['avatar']):
+            st.markdown(f"**{config['avatar']} {config['name']}** *({timestamp[:16]})*")
+            st.markdown(message)
+
+
+def display_timeline():
+    """显示时间线视图"""
+    st.subheader("⏱️ 执行时间线")
+    
+    if not st.session_state.messages:
+        st.info("暂无执行记录")
+        return
+    
+    # 按时间分组显示关键事件
+    timeline_data = []
+    for msg in st.session_state.messages:
+        agent = msg.get("agent_name", "Unknown")
+        timestamp = msg.get("timestamp", "")[:19].replace('T', ' ')
+        message = msg.get("message", "")[:100]
+        
+        if agent in AGENT_CONFIG:
+            timeline_data.append({
+                "time": timestamp,
+                "agent": agent,
+                "message": message,
+                "icon": AGENT_CONFIG[agent]["avatar"]
+            })
+    
+    # 倒序显示（最新的在前）
+    for item in reversed(timeline_data[-10:]):  # 只显示最近 10 条
+        with st.container():
+            st.markdown(f"""
+            <div class="timeline-item">
+                <strong>{item['icon']} {item['time']}</strong><br>
+                <span style="color:#666;">{AGENT_CONFIG.get(item['agent'], {}).get('name', item['agent'])}</span><br>
+                {item['message']}...
+            </div>
+            """, unsafe_allow_html=True)
+
+
+def display_generated_files():
+    """显示生成的文件"""
+    st.subheader("📁 生成的文件")
+    
+    workspace_dir = Path("workspace")
+    if not workspace_dir.exists():
+        st.info("工作目录为空")
+        return
+    
+    files = list(workspace_dir.glob("*"))
+    
+    if not files:
+        st.info("暂无生成的文件")
+        return
+    
+    for file_path in files:
+        if file_path.is_file():
+            file_size = file_path.stat().st_size
+            file_type = file_path.suffix
+            
+            icon_map = {
+                ".py": "🐍",
+                ".md": "📝",
+                ".txt": "📄",
+                ".json": "📊"
+            }
+            
+            icon = icon_map.get(file_type, "📎")
+            
+            with st.container():
+                st.markdown(f"""
+                <div class="file-card">
+                    <div style="display:flex;justify-content:space-between;align-items:center;">
+                        <div>
+                            <span style="font-size:1.5rem;">{icon}</span>
+                            <strong>{file_path.name}</strong>
+                            <span style="color:#666;font-size:0.9rem;margin-left:1rem;">
+                                {file_size:,} bytes
+                            </span>
+                        </div>
+                    </div>
+                </div>
+                """, unsafe_allow_html=True)
+                
+                # 文件内容预览
+                if st.button(f"👁️ 预览 {file_path.name}", key=f"preview_{file_path.name}"):
+                    try:
+                        with open(file_path, 'r', encoding='utf-8') as f:
+                            content = f.read()
+                        st.code(content[:2000] + ("..." if len(content) > 2000 else ""))
+                    except Exception as e:
+                        st.error(f"读取失败：{e}")
+
+
+def display_chat_interface():
+    """显示聊天界面"""
+    st.subheader("💬 Agent 对话实录")
+    
+    if not st.session_state.messages:
+        st.info("👈 暂无对话，请在下方输入需求并启动工作流")
+    else:
+        # 使用容器显示消息
+        chat_container = st.container()
+        with chat_container:
+            for idx, msg in enumerate(st.session_state.messages):
+                display_message(
+                    agent_name=msg.get("agent_name", "Unknown"),
+                    message=msg.get("message", ""),
+                    timestamp=msg.get("timestamp", ""),
+                    index=idx
+                )
+
+
+def export_conversation(messages: List[Dict]) -> str:
+    """导出对话为 Markdown"""
+    md_content = "# AutoGen SDLC 对话历史\n\n"
+    md_content += f"**导出时间**: {datetime.now().strftime('%Y-%m-%d %H:%M:%S')}\n\n"
+    md_content += "---\n\n"
+    
+    for msg in messages:
+        timestamp = msg.get("timestamp", "")[:19].replace('T', ' ')
+        agent = msg.get("agent_name", "Unknown")
+        content = msg.get("message", "")
+        
+        config = AGENT_CONFIG.get(agent, {"avatar": "🤖", "name": agent})
+        
+        md_content += f"### {config['avatar']} [{timestamp}] {config['name']}\n\n"
+        md_content += f"{content}\n\n"
+        md_content += "---\n\n"
+    
+    return md_content
+
+
+def main():
+    """主应用"""
+    init_session_state()
+    
+    # ===== 顶部标题区 =====
+    st.title("🤖 AutoGen SDLC 多智能体协作平台")
+    st.markdown("**基于 AutoGen + Qwen3.5-flash 的端到端软件交付系统 · 实时可视化 Agent 协作**")
+    
+    # ===== 侧边栏 - 配置区域 =====
+    with st.sidebar:
+        st.title("⚙️ 控制中心")
+        
+        # API 配置
+        st.subheader("🔑 模型配置")
+        api_key = st.text_input(
+            "API Key",
+            type="password",
+            value=os.getenv("DASHSCOPE_API_KEY", ""),
+            help="阿里云 DashScope API Key"
+        )
+        
+        base_url = st.text_input(
+            "Base URL",
+            value="https://dashscope.aliyuncs.com/compatible-mode/v1",
+            help="API Base URL"
+        )
+        
+        model = st.selectbox(
+            "模型选择",
+            options=["qwen3.5-flash", "qwen-max", "qwen-plus", "qwen-turbo"],
+            index=0,
+            help="选择使用的 Qwen 模型"
+        )
+        
+        # Agent 参数
+        st.subheader("🎛️ Agent 参数")
+        max_round = st.slider(
+            "最大对话轮数",
+            min_value=5,
+            max_value=50,
+            value=20,
+            step=1
+        )
+        
+        temperature = st.slider(
+            "温度参数",
+            min_value=0.0,
+            max_value=1.0,
+            value=0.7,
+            step=0.1
+        )
+        
+        st.divider()
+        
+        # 控制按钮
+        st.subheader("🎮 流程控制")
+        col1, col2 = st.columns(2)
+        
+        with col1:
+            start_btn = st.button("▶️ 启动", use_container_width=True, type="primary")
+        with col2:
+            stop_btn = st.button("⏸️ 暂停", use_container_width=True)
+        
+        st.divider()
+        
+        # 导出选项
+        st.subheader("💾 数据导出")
+        if st.button("📄 导出 JSON", use_container_width=True):
+            if st.session_state.messages:
+                json_str = json.dumps(st.session_state.messages, ensure_ascii=False, indent=2)
+                st.download_button(
+                    label="下载 JSON",
+                    data=json_str,
+                    file_name=f"conversation_{datetime.now().strftime('%Y%m%d_%H%M%S')}.json",
+                    mime="application/json"
+                )
+        
+        if st.button("📝 导出 Markdown", use_container_width=True):
+            if st.session_state.messages:
+                md_content = export_conversation(st.session_state.messages)
+                st.download_button(
+                    label="下载 Markdown",
+                    data=md_content,
+                    file_name=f"conversation_{datetime.now().strftime('%Y%m%d_%H%M%S')}.md",
+                    mime="text/markdown"
+                )
+    
+    # ===== 主区域 - 三栏布局 =====
+    # 第一行：状态指标
+    status_col1, status_col2, status_col3, status_col4 = st.columns(4)
+    
+    with status_col1:
+        total_msgs = len(st.session_state.messages)
+        st.metric("💬 总消息数", total_msgs)
+    
+    with status_col2:
+        status = "🟢 运行中" if st.session_state.is_running else "🔴 已停止"
+        status_class = "running" if st.session_state.is_running else "stopped"
+        st.markdown(f"""
+        <div class="status-card status-{status_class}">
+            {status}
+        </div>
+        """, unsafe_allow_html=True)
+    
+    with status_col3:
+        result_status = "✅ 成功" if st.session_state.workflow_result and st.session_state.workflow_result.get("success") else "⏳ 未开始"
+        st.metric("📊 工作流状态", result_status)
+    
+    with status_col4:
+        files_count = len(list(Path("workspace").glob("*"))) if Path("workspace").exists() else 0
+        st.metric("📁 生成文件数", files_count)
+    
+    st.divider()
+    
+    # 第二行：左侧 - Agent 状态 + 流程图，右侧 - 时间线
+    left_col, right_col = st.columns([2, 1])
+    
+    with left_col:
+        display_agent_status_panel()
+        st.divider()
+        display_workflow_diagram()
+    
+    with right_col:
+        display_timeline()
+    
+    st.divider()
+    
+    # 第三行：聊天界面
+    display_chat_interface()
+    
+    # ===== 底部 - 用户需求输入 =====
+    st.divider()
+    
+    # 用户输入
+    user_input = st.chat_input("💡 请输入您的软件需求，例如：我需要一个电池健康状态预测 API...")
+    
+    if user_input:
+        # 添加用户消息
+        user_msg = {
+            "timestamp": datetime.now().isoformat(),
+            "agent_name": "User_Proxy",
+            "role": "user",
+            "message": user_input
+        }
+        st.session_state.messages.append(user_msg)
+        st.session_state.conversation_history.append(user_msg)
+        
+        # 更新统计
+        st.session_state.agent_stats["User_Proxy"]["messages"] += 1
+        st.session_state.agent_stats["User_Proxy"]["last_active"] = datetime.now().isoformat()
+        
+        # 刷新显示
+        st.rerun()
+    
+    # ===== 工作流执行逻辑 =====
+    if start_btn:
+        if not api_key:
+            st.error("请先设置 API Key！")
+            st.stop()
+        
+        if not AUTOGEN_AVAILABLE:
+            st.error("请先安装 AutoGen: pip install pyautogen")
+            st.stop()
+        
+        # 获取最后一条用户消息
+        user_messages = [m for m in st.session_state.messages if m.get("role") == "user"]
+        if not user_messages:
+            st.warning("请先输入需求！")
+            st.stop()
+        
+        latest_requirement = user_messages[-1]["message"]
+        
+        # 设置运行状态
+        st.session_state.is_running = True
+        st.session_state.current_step = 0
+        
+        # 创建进度占位符
+        progress_placeholder = st.empty()
+        
+        with progress_placeholder:
+            st.info("🚀 正在启动 SDLC 工作流，请稍候...")
+        
+        try:
+            # 创建 Agent
+            pm_agent, qa_agent, dev_agent, orchestrator, user_proxy, llm_config = create_agents(
+                api_key=api_key,
+                base_url=base_url,
+                model=model
+            )
+            
+            # 创建 GroupChat
+            groupchat = GroupChat(
+                agents=[pm_agent, qa_agent, dev_agent, orchestrator, user_proxy],
+                messages=[],
+                max_round=max_round,
+                speaker_selection_method="round_robin"
+            )
+            
+            manager = GroupChatManager(groupchat=groupchat, llm_config=llm_config)
+            
+            # 构建初始消息
+            initial_message = f"""
+请启动完整的 SDLC 流程：
+
+【用户需求】
+{latest_requirement}
+
+【工作流程】
+1. PM_Agent → 生成 SRS
+2. QA_Agent → 生成测试用例
+3. Dev_Agent → 编写代码
+4. User_Proxy → 执行测试
+5. Orchestrator → 汇总报告
+
+开始协作！
+"""
+            
+            # 记录第一条消息
+            first_msg = {
+                "timestamp": datetime.now().isoformat(),
+                "agent_name": "Orchestrator",
+                "role": "assistant",
+                "message": f"🎯 启动 SDLC 工作流，需求：{latest_requirement[:100]}..."
+            }
+            st.session_state.messages.append(first_msg)
+            
+            # 更新统计
+            st.session_state.agent_stats["Orchestrator"]["messages"] += 1
+            st.session_state.active_agent = "Orchestrator"
+            
+            # 执行对话
+            chat_result = user_proxy.initiate_chat(
+                manager,
+                message=initial_message,
+                max_turns=max_round
+            )
+            
+            # 记录所有对话
+            for msg in groupchat.messages:
+                chat_msg = {
+                    "timestamp": datetime.now().isoformat(),
+                    "agent_name": msg.get("name", "Unknown"),
+                    "role": msg.get("role", "assistant"),
+                    "message": msg.get("content", "")
+                }
+                
+                # 避免重复添加
+                if chat_msg not in st.session_state.messages:
+                    st.session_state.messages.append(chat_msg)
+                    
+                    # 更新统计
+                    agent_name = msg.get("name", "Unknown")
+                    if agent_name in st.session_state.agent_stats:
+                        st.session_state.agent_stats[agent_name]["messages"] += 1
+                        st.session_state.agent_stats[agent_name]["last_active"] = datetime.now().isoformat()
+            
+            # 扫描生成的文件
+            if Path("workspace").exists():
+                files = list(Path("workspace").glob("*"))
+                st.session_state.generated_files = [str(f) for f in files if f.is_file()]
+            
+            # 更新状态
+            st.session_state.workflow_result = {
+                "success": True,
+                "summary": chat_result.summary if hasattr(chat_result, 'summary') else "完成"
+            }
+            st.session_state.current_step = 5
+            st.session_state.is_running = False
+            st.session_state.active_agent = None
+            
+            progress_placeholder.success("✅ SDLC 工作流完成！")
+            
+            # 刷新显示
+            st.rerun()
+            
+        except Exception as e:
+            st.session_state.is_running = False
+            st.session_state.active_agent = None
+            st.error(f"❌ 错误：{str(e)}")
+            st.session_state.workflow_result = {"success": False, "error": str(e)}
+    
+    if stop_btn:
+        st.session_state.is_running = False
+        st.session_state.active_agent = None
+        st.info("⏸️ 工作流已暂停")
+
+
+if __name__ == "__main__":
+    main()
diff --git a/frontend/streamlit_app_v3.py b/frontend/streamlit_app_v3.py
new file mode 100644
index 0000000..7f46a30
--- /dev/null
+++ b/frontend/streamlit_app_v3.py
@@ -0,0 +1,652 @@
+# -*- coding: utf-8 -*-
+"""
+Streamlit 前端 v3 - Agent 对话流可视化版本
+核心特性：
+1. 清晰展示每个 Agent 当前在做什么（任务标签）
+2. Agent 之间的对话流（类似聊天室）
+3. 实时活动指示器（高亮当前发言的 Agent）
+4. 对话历史完整记录
+"""
+import streamlit as st
+import os
+import sys
+from pathlib import Path
+from datetime import datetime
+import time
+from typing import Dict, Any, List
+import json
+
+# 添加项目根目录到路径
+sys.path.insert(0, str(Path(__file__).parent.parent))
+
+try:
+    from autogen import AssistantAgent, UserProxyAgent, GroupChat, GroupChatManager
+    AUTOGEN_AVAILABLE = True
+except ImportError:
+    AUTOGEN_AVAILABLE = False
+
+from config.llm_config import (
+    get_llm_config, 
+    PM_PROMPT, 
+    QA_PROMPT, 
+    DEV_PROMPT, 
+    ORCH_PROMPT
+)
+
+
+# 页面配置
+st.set_page_config(
+    page_title="AutoGen SDLC - Agent 对话流",
+    page_icon="💬",
+    layout="wide",
+    initial_sidebar_state="expanded"
+)
+
+# 自定义 CSS 样式 - 突出对话流
+st.markdown("""
+<style>
+/* 主对话区域 */
+.chat-flow-container {
+    background: #f0f2f6;
+    padding: 20px;
+    border-radius: 10px;
+    margin-bottom: 20px;
+}
+
+/* Agent 对话气泡 */
+.agent-chat-bubble {
+    padding: 15px 20px;
+    margin: 10px 0;
+    border-radius: 15px;
+    max-width: 85%;
+    position: relative;
+    animation: bubble-in 0.3s ease;
+    box-shadow: 0 2px 8px rgba(0,0,0,0.1);
+}
+
+@keyframes bubble-in {
+    from {
+        opacity: 0;
+        transform: translateY(10px);
+    }
+    to {
+        opacity: 1;
+        transform: translateY(0);
+    }
+}
+
+/* 不同 Agent 的气泡样式 */
+.pm-bubble { 
+    background: linear-gradient(135deg, #e3f2fd 0%, #bbdefb 100%); 
+    border: 2px solid #2196f3;
+    margin-left: 0;
+}
+
+.qa-bubble { 
+    background: linear-gradient(135deg, #e8f5e9 0%, #c8e6c9 100%); 
+    border: 2px solid #4caf50;
+    margin-left: 0;
+}
+
+.dev-bubble { 
+    background: linear-gradient(135deg, #fff3e0 0%, #ffe0b2 100%); 
+    border: 2px solid #ff9800;
+    margin-left: 0;
+}
+
+.orchestrator-bubble { 
+    background: linear-gradient(135deg, #f3e5f5 0%, #e1bee7 100%); 
+    border: 2px solid #9c27b0;
+    margin-left: 0;
+}
+
+.user-bubble { 
+    background: linear-gradient(135deg, #efebe9 0%, #d7ccc8 100%); 
+    border: 2px solid #795548;
+    margin-left: auto;
+}
+
+/* Agent 头像和名称 */
+.agent-chat-header {
+    display: flex;
+    align-items: center;
+    margin-bottom: 8px;
+    font-weight: bold;
+}
+
+.agent-chat-avatar {
+    font-size: 1.5rem;
+    margin-right: 8px;
+}
+
+.agent-chat-name {
+    font-size: 1rem;
+    color: #333;
+}
+
+.agent-chat-time {
+    margin-left: auto;
+    font-size: 0.75rem;
+    color: #666;
+}
+
+/* 任务标签 */
+.task-badge {
+    display: inline-block;
+    padding: 3px 10px;
+    border-radius: 12px;
+    font-size: 0.75rem;
+    font-weight: bold;
+    margin-left: 8px;
+    color: white;
+}
+
+.pm-task { background: #2196f3; }
+.qa-task { background: #4caf50; }
+.dev-task { background: #ff9800; }
+.orchestrator-task { background: #9c27b0; }
+.user-task { background: #795548; }
+
+/* Agent 状态卡片 */
+.agent-status-card {
+    padding: 15px;
+    border-radius: 10px;
+    text-align: center;
+    border: 3px solid #e0e0e0;
+    transition: all 0.3s ease;
+    background: white;
+}
+
+.agent-status-card.active {
+    border-color: #4caf50;
+    background: linear-gradient(135deg, #e8f5e9 0%, #c8e6c9 100%);
+    animation: agent-pulse 1.5s infinite;
+}
+
+@keyframes agent-pulse {
+    0%, 100% { transform: scale(1); }
+    50% { transform: scale(1.02); }
+}
+
+.agent-status-avatar {
+    font-size: 2.5rem;
+    margin-bottom: 5px;
+}
+
+.agent-status-name {
+    font-weight: bold;
+    font-size: 0.9rem;
+    margin-bottom: 5px;
+}
+
+.agent-status-task {
+    font-size: 0.75rem;
+    color: #666;
+    min-height: 20px;
+}
+
+.agent-status-indicator {
+    display: inline-block;
+    width: 8px;
+    height: 8px;
+    border-radius: 50%;
+    margin-right: 5px;
+}
+
+.status-active {
+    background: #4caf50;
+    animation: dot-pulse 1s infinite;
+}
+
+.status-waiting {
+    background: #ffc107;
+}
+
+.status-inactive {
+    background: #e0e0e0;
+}
+
+@keyframes dot-pulse {
+    0%, 100% { opacity: 1; }
+    50% { opacity: 0.3; }
+}
+
+/* 对话流指示器 */
+.chat-flow-indicator {
+    display: flex;
+    align-items: center;
+    margin: 10px 0;
+    padding: 10px;
+    background: white;
+    border-radius: 8px;
+}
+
+.flow-arrow {
+    font-size: 1.5rem;
+    color: #2196f3;
+    margin: 0 10px;
+}
+
+/* 当前活动提示 */
+.active-agent-banner {
+    background: linear-gradient(135deg, #4caf50, #8bc34a);
+    color: white;
+    padding: 10px 20px;
+    border-radius: 8px;
+    text-align: center;
+    font-weight: bold;
+    margin-bottom: 20px;
+    animation: banner-pulse 2s infinite;
+}
+
+@keyframes banner-pulse {
+    0%, 100% { opacity: 1; }
+    50% { opacity: 0.8; }
+}
+</style>
+""", unsafe_allow_html=True)
+
+
+# Agent 配置
+AGENT_CONFIG = {
+    "PM_Agent": {
+        "name": "产品经理",
+        "avatar": "📋",
+        "color": "#2196f3",
+        "bubble_class": "pm-bubble",
+        "task_class": "pm-task",
+        "default_task": "待命"
+    },
+    "QA_Agent": {
+        "name": "测试工程师",
+        "avatar": "✅",
+        "color": "#4caf50",
+        "bubble_class": "qa-bubble",
+        "task_class": "qa-task",
+        "default_task": "待命"
+    },
+    "Dev_Agent": {
+        "name": "开发工程师",
+        "avatar": "💻",
+        "color": "#ff9800",
+        "bubble_class": "dev-bubble",
+        "task_class": "dev-task",
+        "default_task": "待命"
+    },
+    "Orchestrator": {
+        "name": "协调器",
+        "avatar": "🎯",
+        "color": "#9c27b0",
+        "bubble_class": "orchestrator-bubble",
+        "task_class": "orchestrator-task",
+        "default_task": "待命"
+    },
+    "User_Proxy": {
+        "name": "用户代理",
+        "avatar": "👤",
+        "color": "#795548",
+        "bubble_class": "user-bubble",
+        "task_class": "user-task",
+        "default_task": "待命"
+    }
+}
+
+
+def init_session_state():
+    """初始化 session state"""
+    if "messages" not in st.session_state:
+        st.session_state.messages = []
+    if "is_running" not in st.session_state:
+        st.session_state.is_running = False
+    if "current_agent" not in st.session_state:
+        st.session_state.current_agent = None
+    if "agent_tasks" not in st.session_state:
+        st.session_state.agent_tasks = {
+            "PM_Agent": "待命",
+            "QA_Agent": "待命",
+            "Dev_Agent": "待命",
+            "Orchestrator": "待命",
+            "User_Proxy": "待命"
+        }
+    if "agent_stats" not in st.session_state:
+        st.session_state.agent_stats = {agent: 0 for agent in AGENT_CONFIG}
+
+
+def display_agent_status_row():
+    """显示 Agent 状态行 - 突出当前活动的 Agent"""
+    st.subheader("🎭 Agent 实时状态")
+    
+    cols = st.columns(5)
+    
+    for i, (agent_key, config) in enumerate(AGENT_CONFIG.items()):
+        with cols[i]:
+            is_active = st.session_state.current_agent == agent_key
+            current_task = st.session_state.agent_tasks.get(agent_key, "待命")
+            msg_count = st.session_state.agent_stats.get(agent_key, 0)
+            
+            status_class = "active" if is_active else ""
+            indicator_class = "status-active" if is_active else "status-inactive"
+            
+            st.markdown(f"""
+            <div class="agent-status-card {status_class}">
+                <div class="agent-status-avatar">{config['avatar']}</div>
+                <div class="agent-status-name">
+                    <span class="agent-status-indicator {indicator_class}"></span>
+                    {config['name']}
+                </div>
+                <div class="agent-status-task">
+                    📝 {current_task}
+                </div>
+                <div style="font-size:0.7rem;color:#999;margin-top:5px;">
+                    💬 {msg_count} 条消息
+                </div>
+            </div>
+            """, unsafe_allow_html=True)
+
+
+def display_active_agent_banner():
+    """显示当前活动 Agent 的横幅"""
+    if st.session_state.current_agent:
+        config = AGENT_CONFIG.get(st.session_state.current_agent, {})
+        st.markdown(f"""
+        <div class="active-agent-banner">
+            {config.get('avatar', '🤖')} 当前发言：{config.get('name', 'Unknown')} - {st.session_state.agent_tasks.get(st.session_state.current_agent, '工作中...')}
+        </div>
+        """, unsafe_allow_html=True)
+
+
+def display_chat_flow():
+    """显示 Agent 对话流 - 核心功能"""
+    st.subheader("💬 Agent 对话流")
+    
+    if not st.session_state.messages:
+        st.info("👈 暂无对话，请启动工作流")
+        return
+    
+    # 使用容器显示对话流
+    chat_container = st.container()
+    
+    for idx, msg in enumerate(st.session_state.messages):
+        agent_key = msg.get("agent_key", "Unknown")
+        config = AGENT_CONFIG.get(agent_key, {"name": "未知", "avatar": "🤖", "bubble_class": ""})
+        task_class = AGENT_CONFIG.get(agent_key, {}).get("task_class", "")
+        
+        timestamp = msg.get("timestamp", "")
+        if timestamp:
+            time_str = datetime.fromisoformat(timestamp).strftime("%H:%M:%S")
+        else:
+            time_str = datetime.now().strftime("%H:%M:%S")
+        
+        content = msg.get("content", "")
+        task = msg.get("task", "工作中")
+        
+        # 显示对话气泡
+        st.markdown(f"""
+        <div class="agent-chat-bubble {config['bubble_class']}">
+            <div class="agent-chat-header">
+                <span class="agent-chat-avatar">{config['avatar']}</span>
+                <span class="agent-chat-name">{config['name']}</span>
+                <span class="task-badge {task_class}">{task}</span>
+                <span class="agent-chat-time">{time_str}</span>
+            </div>
+            <div style="color:#333;line-height:1.6;">{content}</div>
+        </div>
+        """, unsafe_allow_html=True)
+        
+        # 显示对话流指示器（下一条消息的箭头）
+        if idx < len(st.session_state.messages) - 1:
+            next_msg = st.session_state.messages[idx + 1]
+            next_agent = next_msg.get("agent_key", "")
+            next_config = AGENT_CONFIG.get(next_agent, {})
+            
+            if next_agent != agent_key:
+                st.markdown(f"""
+                <div class="chat-flow-indicator">
+                    <div style="flex:1;"></div>
+                    <div class="flow-arrow">⬇️</div>
+                    <div style="flex:1;text-align:center;color:#666;font-size:0.8rem;">
+                        传递给 {next_config.get('name', 'Unknown')}
+                    </div>
+                </div>
+                """, unsafe_allow_html=True)
+
+
+def create_agents(api_key: str, base_url: str, model: str):
+    """创建所有 Agent"""
+    llm_config = get_llm_config(model=model, api_key=api_key, base_url=base_url)
+    
+    pm_agent = AssistantAgent(
+        name="PM_Agent",
+        system_message=PM_PROMPT,
+        llm_config=llm_config,
+        description="资深软件产品经理",
+        human_input_mode="NEVER"
+    )
+    
+    qa_agent = AssistantAgent(
+        name="QA_Agent",
+        system_message=QA_PROMPT,
+        llm_config=llm_config,
+        description="资深测试工程师",
+        human_input_mode="NEVER"
+    )
+    
+    dev_agent = AssistantAgent(
+        name="Dev_Agent",
+        system_message=DEV_PROMPT,
+        llm_config=llm_config,
+        description="资深软件工程师",
+        human_input_mode="NEVER"
+    )
+    
+    orchestrator = AssistantAgent(
+        name="Orchestrator",
+        system_message=ORCH_PROMPT,
+        llm_config=llm_config,
+        description="多智能体协调器",
+        human_input_mode="NEVER"
+    )
+    
+    user_proxy = UserProxyAgent(
+        name="User_Proxy",
+        human_input_mode="NEVER",
+        max_consecutive_auto_reply=0,
+        code_execution_config={
+            "work_dir": "workspace",
+            "use_docker": False,
+        },
+        is_termination_msg=lambda x: x.get("content", "").rstrip().endswith("TERMINATE")
+    )
+    
+    return pm_agent, qa_agent, dev_agent, orchestrator, user_proxy, llm_config
+
+
+def add_message(agent_key: str, content: str, task: str = "工作中"):
+    """添加消息到对话流"""
+    msg = {
+        "agent_key": agent_key,
+        "content": content,
+        "timestamp": datetime.now().isoformat(),
+        "task": task
+    }
+    st.session_state.messages.append(msg)
+    
+    # 更新统计
+    st.session_state.agent_stats[agent_key] = st.session_state.agent_stats.get(agent_key, 0) + 1
+    
+    # 更新当前 Agent
+    st.session_state.current_agent = agent_key
+    
+    # 更新任务
+    st.session_state.agent_tasks[agent_key] = task
+
+
+def main():
+    """主应用"""
+    init_session_state()
+    
+    # 标题
+    st.title("💬 AutoGen SDLC - Agent 对话流可视化")
+    st.markdown("**清晰展示每个 Agent 在做什么 · 实时追踪 Agent 之间的对话交互**")
+    
+    # 侧边栏配置
+    with st.sidebar:
+        st.title("⚙️ 控制中心")
+        
+        api_key = st.text_input("API Key", type="password", value=os.getenv("DASHSCOPE_API_KEY", ""))
+        base_url = st.text_input("Base URL", value="https://dashscope.aliyuncs.com/compatible-mode/v1")
+        model = st.selectbox("模型选择", ["qwen3.5-flash", "qwen-max", "qwen-plus", "qwen-turbo"], index=0)
+        
+        max_round = st.slider("最大对话轮数", 5, 50, 20)
+        
+        st.divider()
+        
+        col1, col2 = st.columns(2)
+        with col1:
+            start_btn = st.button("▶️ 启动对话流", type="primary", use_container_width=True)
+        with col2:
+            stop_btn = st.button("⏸️ 暂停", use_container_width=True)
+        
+        st.divider()
+        
+        if st.button("🗑️ 清空对话", use_container_width=True):
+            st.session_state.messages = []
+            st.session_state.current_agent = None
+            st.session_state.agent_tasks = {k: "待命" for k in AGENT_CONFIG}
+            st.session_state.agent_stats = {k: 0 for k in AGENT_CONFIG}
+            st.rerun()
+        
+        if st.button("📥 导出对话", use_container_width=True):
+            if st.session_state.messages:
+                json_str = json.dumps(st.session_state.messages, ensure_ascii=False, indent=2)
+                st.download_button(
+                    label="下载 JSON",
+                    data=json_str,
+                    file_name=f"agent_chat_{datetime.now().strftime('%Y%m%d_%H%M%S')}.json",
+                    mime="application/json"
+                )
+    
+    # 主界面
+    display_agent_status_row()
+    st.divider()
+    
+    display_active_agent_banner()
+    
+    display_chat_flow()
+    
+    # 用户输入
+    st.divider()
+    user_input = st.chat_input("💡 输入需求，或点击启动按钮开始...")
+    
+    if user_input:
+        add_message("User_Proxy", user_input, "提出需求")
+        st.rerun()
+    
+    # 启动工作流
+    if start_btn:
+        if not api_key:
+            st.error("请先设置 API Key")
+            st.stop()
+        
+        if not AUTOGEN_AVAILABLE:
+            st.error("请先安装 AutoGen: pip install pyautogen")
+            st.stop()
+        
+        # 获取需求
+        user_msgs = [m for m in st.session_state.messages if m.get("agent_key") == "User_Proxy"]
+        if not user_msgs:
+            add_message("User_Proxy", "请开发一个电池健康状态预测 API", "提出需求")
+        
+        latest_requirement = user_msgs[-1]["content"] if user_msgs else "请开发一个电池健康状态预测 API"
+        
+        st.session_state.is_running = True
+        
+        # 进度提示
+        progress_placeholder = st.empty()
+        progress_placeholder.info("🚀 启动 SDLC 工作流，Agent 开始协作...")
+        
+        try:
+            # 创建 Agent
+            pm_agent, qa_agent, dev_agent, orchestrator, user_proxy, llm_config = create_agents(
+                api_key=api_key,
+                base_url=base_url,
+                model=model
+            )
+            
+            # 添加 Orchestrator 启动消息
+            add_message("Orchestrator", f"🚀 启动 SDLC 工作流！用户需求：{latest_requirement[:100]}...", "启动流程")
+            
+            # 创建 GroupChat
+            groupchat = GroupChat(
+                agents=[pm_agent, qa_agent, dev_agent, orchestrator, user_proxy],
+                messages=[],
+                max_round=max_round,
+                speaker_selection_method="round_robin"
+            )
+            
+            manager = GroupChatManager(groupchat=groupchat, llm_config=llm_config)
+            
+            # 初始消息
+            initial_message = f"""
+请启动完整的 SDLC 流程：
+
+【用户需求】
+{latest_requirement}
+
+【工作流程】
+1. PM_Agent → 生成 SRS 文档
+2. QA_Agent → 生成测试用例  
+3. Dev_Agent → 编写代码
+4. User_Proxy → 执行测试
+5. Orchestrator → 汇总报告
+
+开始协作！每个步骤完成后请明确说明。
+"""
+            
+            # 执行对话
+            with st.spinner("💬 Agent 们正在协作中，请稍候..."):
+                chat_result = user_proxy.initiate_chat(
+                    manager,
+                    message=initial_message,
+                    max_turns=max_round
+                )
+            
+            # 记录所有对话
+            for msg in groupchat.messages:
+                agent_name = msg.get("name", "Unknown")
+                content = msg.get("content", "")
+                
+                # 推断任务
+                task_map = {
+                    "PM_Agent": "需求分析",
+                    "QA_Agent": "测试设计",
+                    "Dev_Agent": "代码实现",
+                    "Orchestrator": "流程协调",
+                    "User_Proxy": "测试执行"
+                }
+                task = task_map.get(agent_name, "工作中")
+                
+                add_message(agent_name, content, task)
+            
+            # 完成
+            add_message("Orchestrator", "✅ SDLC 流程完成！所有任务已完成。", "总结完成")
+            
+            progress_placeholder.success("✅ SDLC 工作流完成！查看上方的对话流了解详情。")
+            st.session_state.is_running = False
+            
+            st.rerun()
+            
+        except Exception as e:
+            st.session_state.is_running = False
+            st.session_state.current_agent = None
+            progress_placeholder.error(f"❌ 错误：{str(e)}")
+            st.error("请检查 API Key 和网络连接")
+    
+    if stop_btn:
+        st.session_state.is_running = False
+        st.session_state.current_agent = None
+        st.info("⏸️ 工作流已暂停")
+
+
+if __name__ == "__main__":
+    main()
diff --git a/quick_start.py b/quick_start.py
new file mode 100644
index 0000000..630030f
--- /dev/null
+++ b/quick_start.py
@@ -0,0 +1,186 @@
+# -*- coding: utf-8 -*-
+"""
+快速启动脚本 - 验证安装并展示基本功能
+"""
+import sys
+from pathlib import Path
+
+def check_installation():
+    """检查依赖包安装情况"""
+    print("=" * 60)
+    print("Checking environment configuration")
+    print("=" * 60)
+    
+    required_packages = {
+        "autogen": "pyautogen",
+        "dashscope": "dashscope",
+        "streamlit": "streamlit",
+        "pytest": "pytest"
+    }
+    
+    missing = []
+    installed = []
+    
+    for package, install_name in required_packages.items():
+        try:
+            __import__(package)
+            installed.append(install_name)
+            print(f"[OK] {install_name}: Installed")
+        except ImportError:
+            missing.append(install_name)
+            print(f"[MISSING] {install_name}: Not installed")
+    
+    print()
+    
+    if missing:
+        print(f"Warning: Missing dependencies: {', '.join(missing)}")
+        print(f"Tip: Run: pip install -r requirements.txt")
+        return False
+    else:
+        print("All dependencies installed successfully!")
+        return True
+
+
+def check_api_key():
+    """检查 API Key 配置"""
+    import os
+    
+    print("\n" + "=" * 60)
+    print("Checking API Key configuration")
+    print("=" * 60)
+    
+    api_key = os.getenv("DASHSCOPE_API_KEY")
+    
+    if not api_key:
+        print("[ERROR] DASHSCOPE_API_KEY environment variable not set")
+        print("\nHow to set:")
+        print("   Windows (PowerShell):")
+        print('   $env:DASHSCOPE_API_KEY="your_api_key_here"')
+        print("\n   Linux/Mac:")
+        print('   export DASHSCOPE_API_KEY="your_api_key_here"')
+        return False
+    else:
+        masked_key = api_key[:4] + "..." + api_key[-4:] if len(api_key) > 8 else "***"
+        print(f"[OK] API Key configured: {masked_key}")
+        return True
+
+
+def show_project_structure():
+    """展示项目结构"""
+    print("\n" + "=" * 60)
+    print("Project Structure")
+    print("=" * 60)
+    
+    structure = """
+project/
++--- agents/              # Agent modules
+|    +--- pm_agent.py     # Product Manager Agent
+|    +--- qa_agent.py     # QA Engineer Agent
+|    +--- dev_agent.py    # Developer Agent
+|    +--- orchestrator.py # Orchestrator Agent
++--- config/             # Configuration
+|    +--- llm_config.py  # LLM config & prompts
++--- frontend/           # Frontend UI
+|    +--- streamlit_app.py
++--- utils/              # Utilities
+|    +--- logger.py      # Logging
+|    +--- callback_handler.py
++--- tests/              # Unit tests
+|    +--- test_agents.py
++--- workspace/          # Working directory
++--- logs/               # Log directory
++--- autogen_sdls_system.py       # Main entry
++--- autogen_self_healing_demo.py # Self-healing demo
++--- usage_examples.py   # Usage examples
++--- requirements.txt    # Dependencies
+    """
+    print(structure)
+
+
+def show_usage_guide():
+    """展示使用指南"""
+    print("=" * 60)
+    print("Usage Guide")
+    print("=" * 60)
+    
+    guide = """
+[Method 1] Run full SDLC workflow:
+python autogen_sdls_system.py
+
+[Method 2] Start Streamlit frontend:
+streamlit run frontend/streamlit_app.py
+
+[Method 3] Self-healing feature demo:
+python autogen_self_healing_demo.py
+
+[Method 4] View usage examples:
+python usage_examples.py
+
+[Method 5] Run unit tests:
+pytest tests/test_agents.py -v
+    """
+    print(guide)
+
+
+def demo_basic_import():
+    """演示基本导入"""
+    print("\n" + "=" * 60)
+    print("Testing basic functionality")
+    print("=" * 60)
+    
+    try:
+        from config.llm_config import get_llm_config, PM_PROMPT, QA_PROMPT
+        print("[OK] Configuration module imported")
+        
+        # 测试配置生成
+        config = get_llm_config(model="qwen3.5-flash", api_key="test")
+        print(f"[OK] LLM config generated: {config['config_list'][0]['model']}")
+        
+        # 测试提示词
+        print(f"[OK] PM prompt length: {len(PM_PROMPT)} chars")
+        print(f"[OK] QA prompt length: {len(QA_PROMPT)} chars")
+        
+        return True
+        
+    except Exception as e:
+        print(f"[ERROR] Import failed: {e}")
+        return False
+
+
+def main():
+    """主函数"""
+    print("\n")
+    print("=" * 60)
+    print(" " * 15 + "AutoGen SDLC System")
+    print(" " * 10 + "Powered by AutoGen + Qwen3.5-flash")
+    print("=" * 60)
+    print()
+    
+    # 检查环境
+    env_ok = check_installation()
+    api_ok = check_api_key()
+    
+    # 展示项目结构
+    show_project_structure()
+    
+    # 展示使用指南
+    show_usage_guide()
+    
+    # 测试基本功能
+    if env_ok:
+        demo_ok = demo_basic_import()
+        if demo_ok:
+            print("\nSystem initialization completed!")
+            print("\nNext steps:")
+            print("   1. Get API Key from Alibaba Cloud DashScope console")
+            print("   2. Set env var: export DASHSCOPE_API_KEY='your_key'")
+            print("   3. Run: python autogen_sdls_system.py")
+    
+    print("\n" + "=" * 60)
+    print("For details, see: README.md")
+    print("=" * 60)
+    print()
+
+
+if __name__ == "__main__":
+    main()
diff --git a/requirements.txt b/requirements.txt
new file mode 100644
index 0000000..72fad4a
--- /dev/null
+++ b/requirements.txt
@@ -0,0 +1,22 @@
+# AutoGen SDLC 多智能体系统依赖包
+
+# 核心框架
+pyautogen>=0.2.16
+
+# 阿里云 DashScope SDK（Qwen 模型）
+dashscope>=1.14.0
+
+# 前端界面
+streamlit>=1.28.0
+
+# 测试工具
+pytest>=7.4.0
+pytest-cov>=4.1.0
+
+# 工具库
+python-dotenv>=1.0.0  # 环境变量管理
+
+# 可选：代码质量工具
+# black>=23.0.0
+# flake8>=6.0.0
+# mypy>=1.0.0
diff --git a/run_demo.py b/run_demo.py
new file mode 100644
index 0000000..b5414e8
--- /dev/null
+++ b/run_demo.py
@@ -0,0 +1,87 @@
+# -*- coding: utf-8 -*-
+"""
+快速演示脚本 - 展示如何使用增强版前端
+"""
+import os
+import sys
+from pathlib import Path
+
+print("=" * 70)
+print("AutoGen SDLC Platform - Quick Start Guide")
+print("=" * 70)
+print()
+
+# 检查依赖
+print("Checking dependencies...")
+required = ["streamlit", "autogen", "dashscope"]
+missing = []
+
+for pkg in required:
+    try:
+        __import__(pkg)
+        print(f"  [OK] {pkg}")
+    except ImportError:
+        print(f"  [MISSING] {pkg}")
+        missing.append(pkg)
+
+print()
+
+if missing:
+    print(f"Warning: Missing packages: {', '.join(missing)}")
+    print()
+    print("Tip: Install with:")
+    print(f"   pip install {' '.join(missing)}")
+    print()
+else:
+    print("[OK] All dependencies installed!")
+    print()
+    
+    # 显示启动命令
+    print("=" * 70)
+    print("Launch Commands")
+    print("=" * 70)
+    print()
+    print("[Method 1] Enhanced Visual Interface (Recommended)")
+    print("   streamlit run frontend/streamlit_app_v2.py")
+    print()
+    print("[Method 2] Classic Interface")
+    print("   streamlit run frontend/streamlit_app.py")
+    print()
+    print("[Method 3] Command Line Workflow")
+    print("   python autogen_sdls_system.py")
+    print()
+    print("[Method 4] Self-Healing Demo")
+    print("   python autogen_self_healing_demo.py")
+    print()
+    
+    # API Key 检查
+    api_key = os.getenv("DASHSCOPE_API_KEY")
+    if not api_key:
+        print("Warning: DASHSCOPE_API_KEY not set")
+        print()
+        print("How to set (Windows PowerShell):")
+        print('   $env:DASHSCOPE_API_KEY="your_api_key_here"')
+        print()
+        print("How to set (Linux/Mac):")
+        print('   export DASHSCOPE_API_KEY="your_api_key_here"')
+        print()
+    else:
+        masked = api_key[:4] + "..." + api_key[-4:] if len(api_key) > 8 else "***"
+        print(f"[OK] API Key configured: {masked}")
+        print()
+    
+    print("=" * 70)
+    print("Features")
+    print("=" * 70)
+    print()
+    print("Enhanced Interface (streamlit_app_v2.py) features:")
+    print()
+    print("  - Real-time Agent Status Panel")
+    print("  - Workflow Diagram (Mermaid)")
+    print("  - Execution Timeline")
+    print("  - Live Chat Display")
+    print("  - Generated Files Preview")
+    print("  - Statistics Dashboard")
+    print()
+    print("=" * 70)
+    print()
diff --git a/tests/__init__.py b/tests/__init__.py
new file mode 100644
index 0000000..c651962
--- /dev/null
+++ b/tests/__init__.py
@@ -0,0 +1 @@
+"""测试模块初始化文件"""
diff --git a/tests/test_agents.py b/tests/test_agents.py
new file mode 100644
index 0000000..98b008f
--- /dev/null
+++ b/tests/test_agents.py
@@ -0,0 +1,212 @@
+"""
+Agent 单元测试模块
+测试各个 Agent 的基本功能
+"""
+import pytest
+import os
+import sys
+from pathlib import Path
+
+# 添加项目根目录到路径
+sys.path.insert(0, str(Path(__file__).parent.parent))
+
+
+class TestLLMConfig:
+    """测试 LLM 配置模块"""
+    
+    def test_get_llm_config(self):
+        """测试获取基本配置"""
+        from config.llm_config import get_llm_config
+        
+        config = get_llm_config(
+            model="qwen3.5-flash",
+            api_key="test_key",
+            temperature=0.7
+        )
+        
+        assert "config_list" in config
+        assert len(config["config_list"]) == 1
+        assert config["config_list"][0]["model"] == "qwen3.5-flash"
+        assert config["config_list"][0]["api_key"] == "test_key"
+        assert config["config_list"][0]["temperature"] == 0.7
+    
+    def test_get_agent_llm_config(self):
+        """测试获取特定 Agent 配置"""
+        from config.llm_config import get_agent_llm_config
+        
+        config = get_agent_llm_config("PM_Agent")
+        
+        assert "config_list" in config
+        assert config["config_list"][0]["model"] == "qwen3.5-flash"
+    
+    def test_system_prompts_exist(self):
+        """测试系统提示词存在"""
+        from config.llm_config import PM_PROMPT, QA_PROMPT, DEV_PROMPT, ORCH_PROMPT
+        
+        assert PM_PROMPT is not None
+        assert len(PM_PROMPT) > 100
+        
+        assert QA_PROMPT is not None
+        assert len(QA_PROMPT) > 100
+        
+        assert DEV_PROMPT is not None
+        assert len(DEV_PROMPT) > 100
+        
+        assert ORCH_PROMPT is not None
+        assert len(ORCH_PROMPT) > 100
+
+
+class TestLogger:
+    """测试日志模块"""
+    
+    def test_logger_creation(self):
+        """测试日志记录器创建"""
+        from utils.logger import ConversationLogger
+        
+        logger = ConversationLogger(log_dir="test_logs")
+        
+        assert logger is not None
+        assert logger.session_id is not None
+    
+    def test_log_message(self):
+        """测试消息记录"""
+        from utils.logger import ConversationLogger
+        import shutil
+        
+        # 清理测试目录
+        test_dir = Path("test_logs")
+        if test_dir.exists():
+            shutil.rmtree(test_dir)
+        
+        logger = ConversationLogger(log_dir="test_logs")
+        logger.log_message(
+            agent_name="Test_Agent",
+            message="这是一条测试消息",
+            role="assistant"
+        )
+        
+        assert len(logger.conversation_history) == 1
+        assert logger.conversation_history[0]["agent_name"] == "Test_Agent"
+        assert logger.conversation_history[0]["message"] == "这是一条测试消息"
+        
+        # 清理
+        shutil.rmtree(test_dir)
+    
+    def test_callback_handler(self):
+        """测试回调处理器"""
+        from utils.callback_handler import MessageCallbackHandler
+        
+        handler = MessageCallbackHandler()
+        
+        received_messages = []
+        
+        def test_callback(msg):
+            received_messages.append(msg)
+        
+        handler.register_callback(test_callback)
+        handler.on_message(
+            agent_name="Test_Agent",
+            message="测试回调消息"
+        )
+        
+        assert len(received_messages) == 1
+        assert received_messages[0]["agent_name"] == "Test_Agent"
+
+
+class TestAgentsCreation:
+    """测试 Agent 创建"""
+    
+    @pytest.mark.skipif(not os.getenv("DASHSCOPE_API_KEY"), reason="需要 API Key")
+    def test_create_pm_agent(self):
+        """测试创建 PM Agent"""
+        from agents import create_pm_agent
+        from config.llm_config import get_llm_config
+        
+        api_key = os.getenv("DASHSCOPE_API_KEY")
+        llm_config = get_llm_config(api_key=api_key)
+        
+        agent = create_pm_agent(llm_config=llm_config)
+        
+        assert agent is not None
+        assert agent.name == "PM_Agent"
+    
+    @pytest.mark.skipif(not os.getenv("DASHSCOPE_API_KEY"), reason="需要 API Key")
+    def test_create_qa_agent(self):
+        """测试创建 QA Agent"""
+        from agents import create_qa_agent
+        from config.llm_config import get_llm_config
+        
+        api_key = os.getenv("DASHSCOPE_API_KEY")
+        llm_config = get_llm_config(api_key=api_key)
+        
+        agent = create_qa_agent(llm_config=llm_config)
+        
+        assert agent is not None
+        assert agent.name == "QA_Agent"
+    
+    @pytest.mark.skipif(not os.getenv("DASHSCOPE_API_KEY"), reason="需要 API Key")
+    def test_create_dev_agent(self):
+        """测试创建 Dev Agent"""
+        from agents import create_dev_agent
+        from config.llm_config import get_llm_config
+        
+        api_key = os.getenv("DASHSCOPE_API_KEY")
+        llm_config = get_llm_config(api_key=api_key)
+        
+        agent = create_dev_agent(llm_config=llm_config)
+        
+        assert agent is not None
+        assert agent.name == "Dev_Agent"
+    
+    @pytest.mark.skipif(not os.getenv("DASHSCOPE_API_KEY"), reason="需要 API Key")
+    def test_create_orchestrator(self):
+        """测试创建 Orchestrator Agent"""
+        from agents import create_orchestrator_agent
+        from config.llm_config import get_llm_config
+        
+        api_key = os.getenv("DASHSCOPE_API_KEY")
+        llm_config = get_llm_config(api_key=api_key)
+        
+        agent = create_orchestrator_agent(llm_config=llm_config)
+        
+        assert agent is not None
+        assert agent.name == "Orchestrator"
+
+
+class TestWorkspaceManagement:
+    """测试工作空间管理"""
+    
+    def test_workspace_directory_exists(self):
+        """测试工作目录存在"""
+        workspace_dir = Path(__file__).parent.parent / "workspace"
+        workspace_dir.mkdir(parents=True, exist_ok=True)
+        
+        assert workspace_dir.exists()
+        assert workspace_dir.is_dir()
+    
+    def test_create_sample_files(self):
+        """测试创建示例文件"""
+        workspace_dir = Path(__file__).parent.parent / "workspace"
+        workspace_dir.mkdir(parents=True, exist_ok=True)
+        
+        # 创建测试文件
+        test_file = workspace_dir / "test_sample.txt"
+        test_content = "这是一个测试文件"
+        
+        with open(test_file, 'w', encoding='utf-8') as f:
+            f.write(test_content)
+        
+        assert test_file.exists()
+        
+        with open(test_file, 'r', encoding='utf-8') as f:
+            content = f.read()
+        
+        assert content == test_content
+        
+        # 清理
+        test_file.unlink()
+
+
+if __name__ == "__main__":
+    # 运行测试
+    pytest.main([__file__, "-v", "--tb=short"])
diff --git a/usage_examples.py b/usage_examples.py
new file mode 100644
index 0000000..2b73593
--- /dev/null
+++ b/usage_examples.py
@@ -0,0 +1,230 @@
+"""
+使用示例 - 展示如何使用 AutoGen SDLC 系统
+包含多个实际场景的演示
+"""
+import os
+import sys
+from pathlib import Path
+
+sys.path.insert(0, str(Path(__file__).parent))
+
+from autogen_sdls_system import AutoGenSDLCSystem
+
+
+def example_1_basic_workflow():
+    """示例 1: 基本工作流 - 电池健康预测 API"""
+    print("\n" + "=" * 70)
+    print("示例 1: 基本 SDLC 工作流")
+    print("=" * 70)
+    
+    api_key = os.getenv("DASHSCOPE_API_KEY")
+    if not api_key:
+        print("❌ 请设置 DASHSCOPE_API_KEY 环境变量")
+        return
+    
+    system = AutoGenSDLCSystem(api_key=api_key)
+    
+    requirement = """
+我需要一个电池健康状态 (SOH) 预测 API，要求：
+1. 接收电池的电压、电流、温度数据
+2. 输出健康度百分比（0-100%）
+3. 符合汽车行业标准
+4. 包含错误处理和边界情况检测
+"""
+    
+    print(f"\n📋 需求：{requirement}")
+    result = system.run_workflow(requirement, max_round=15)
+    
+    if result["success"]:
+        print("\n✅ 工作流完成！")
+        print(f"📄 摘要：{result['summary'][:300]}...")
+    else:
+        print(f"\n❌ 工作流失败：{result.get('error')}")
+
+
+def example_2_custom_agents():
+    """示例 2: 自定义 Agent 配置"""
+    print("\n" + "=" * 70)
+    print("示例 2: 自定义 Agent 配置")
+    print("=" * 70)
+    
+    from config.llm_config import get_llm_config
+    from agents import create_pm_agent, create_qa_agent
+    
+    api_key = os.getenv("DASHSCOPE_API_KEY")
+    if not api_key:
+        print("❌ 请设置 DASHSCOPE_API_KEY 环境变量")
+        return
+    
+    # 使用不同的模型配置
+    llm_config = get_llm_config(
+        model="qwen3.5-flash",
+        api_key=api_key,
+        temperature=0.5,  # 降低温度使输出更稳定
+        max_tokens=3000
+    )
+    
+    # 创建自定义 Agent
+    pm_agent = create_pm_agent(llm_config=llm_config)
+    qa_agent = create_qa_agent(llm_config=llm_config)
+    
+    print("✅ 已创建自定义配置的 Agent")
+    print(f"   PM Agent: {pm_agent.name}")
+    print(f"   QA Agent: {qa_agent.name}")
+
+
+def example_3_direct_agent_call():
+    """示例 3: 直接调用单个 Agent"""
+    print("\n" + "=" * 70)
+    print("示例 3: 直接调用 PM Agent 生成 SRS")
+    print("=" * 70)
+    
+    from agents import ProductManagerAgent
+    from config.llm_config import get_agent_llm_config
+    
+    api_key = os.getenv("DASHSCOPE_API_KEY")
+    if not api_key:
+        print("❌ 请设置 DASHSCOPE_API_KEY 环境变量")
+        return
+    
+    # 创建 PM Agent
+    llm_config = get_agent_llm_config("PM_Agent", api_key=api_key)
+    pm_agent = ProductManagerAgent(llm_config=llm_config)
+    
+    # 简单需求
+    requirement = "开发一个车载蓝牙连接管理模块"
+    
+    print(f"\n📋 需求：{requirement}")
+    print("\n🤖 PM Agent 正在生成 SRS...\n")
+    
+    try:
+        srs = pm_agent.generate_srs(requirement)
+        print("\n✅ SRS 生成成功！")
+        print(f"📄 前 500 字符预览:\n{srs[:500]}...")
+    except Exception as e:
+        print(f"\n❌ 生成失败：{e}")
+
+
+def example_4_test_generation():
+    """示例 4: QA Agent 生成测试用例"""
+    print("\n" + "=" * 70)
+    print("示例 4: QA Agent 生成测试用例")
+    print("=" * 70)
+    
+    from agents import QAAgent
+    from config.llm_config import get_agent_llm_config
+    
+    api_key = os.getenv("DASHSCOPE_API_KEY")
+    if not api_key:
+        print("❌ 请设置 DASHSCOPE_API_KEY 环境变量")
+        return
+    
+    # 创建 QA Agent
+    llm_config = get_agent_llm_config("QA_Agent", api_key=api_key)
+    qa_agent = QAAgent(llm_config=llm_config)
+    
+    # 简单的 SRS 片段
+    srs_sample = """
+【功能性需求】
+FR-001: 系统应能接收电池电压数据（范围 0-5V）
+FR-002: 系统应能计算 SOH 百分比（范围 0-100%）
+FR-003: 系统应能检测异常数据并报错
+
+【非功能性需求】
+NFR-001: 响应时间 < 100ms
+NFR-002: 符合 MISRA-C 规范
+"""
+    
+    print("\n📋 SRS 样本:")
+    print(srs_sample)
+    print("\n🤖 QA Agent 正在生成测试用例...\n")
+    
+    try:
+        test_cases = qa_agent.generate_test_cases(srs_sample)
+        print("\n✅ 测试用例生成成功！")
+        print(f"📄 预览:\n{test_cases[:500]}...")
+    except Exception as e:
+        print(f"\n❌ 生成失败：{e}")
+
+
+def example_5_code_generation():
+    """示例 5: Dev Agent 生成代码"""
+    print("\n" + "=" * 70)
+    print("示例 5: Dev Agent 生成代码")
+    print("=" * 70)
+    
+    from agents import DevAgent
+    from config.llm_config import get_agent_llm_config
+    
+    api_key = os.getenv("DASHSCOPE_API_KEY")
+    if not api_key:
+        print("❌ 请设置 DASHSCOPE_API_KEY 环境变量")
+        return
+    
+    # 创建 Dev Agent
+    llm_config = get_agent_llm_config("Dev_Agent", api_key=api_key)
+    dev_agent = DevAgent(llm_config=llm_config)
+    
+    # SRS 和测试样本
+    srs_sample = """
+FR-001: 实现 SOH 计算函数
+FR-002: 输入验证（电压 0-5V，温度 -20~60°C）
+FR-003: 输出限制在 0-100%
+"""
+    
+    test_sample = """
+def test_soh_calculation():
+    assert calculate_soh(3.7, 2.0, 25)["soh"] > 0
+    assert calculate_soh(0, 0, 25)["soh"] == 0
+"""
+    
+    print("\n📋 需求和测试样本已准备")
+    print("\n🤖 Dev Agent 正在生成代码...\n")
+    
+    try:
+        code = dev_agent.generate_code(srs_sample, test_sample)
+        print("\n✅ 代码生成成功！")
+        print(f"📄 代码预览:\n{code[:500]}...")
+    except Exception as e:
+        print(f"\n❌ 生成失败：{e}")
+
+
+def run_all_examples():
+    """运行所有示例"""
+    print("\n" + "=" * 70)
+    print("🚀 AutoGen SDLC 系统 - 使用示例合集")
+    print("=" * 70)
+    
+    examples = [
+        ("直接调用 PM Agent", example_3_direct_agent_call),
+        ("QA Agent 生成测试", example_4_test_generation),
+        ("Dev Agent 生成代码", example_5_code_generation),
+        ("自定义 Agent 配置", example_2_custom_agents),
+        ("完整工作流", example_1_basic_workflow),
+    ]
+    
+    for name, func in examples:
+        print(f"\n\n{'='*70}")
+        print(f"▶️  运行示例：{name}")
+        print('='*70)
+        
+        try:
+            func()
+        except Exception as e:
+            print(f"\n❌ 示例执行失败：{e}")
+        
+        input("\n按 Enter 键继续下一个示例...")
+    
+    print("\n" + "=" * 70)
+    print("✅ 所有示例运行完成！")
+    print("=" * 70)
+
+
+if __name__ == "__main__":
+    # 可以选择运行单个示例或全部示例
+    # example_3_direct_agent_call()
+    # example_4_test_generation()
+    # example_5_code_generation()
+    
+    # 运行所有示例
+    run_all_examples()
diff --git a/utils/__init__.py b/utils/__init__.py
new file mode 100644
index 0000000..b7b6d15
--- /dev/null
+++ b/utils/__init__.py
@@ -0,0 +1,10 @@
+"""工具模块初始化文件"""
+from .logger import ConversationLogger, get_logger
+from .callback_handler import MessageCallbackHandler, get_callback_handler
+
+__all__ = [
+    "ConversationLogger",
+    "get_logger",
+    "MessageCallbackHandler",
+    "get_callback_handler"
+]
diff --git a/utils/callback_handler.py b/utils/callback_handler.py
new file mode 100644
index 0000000..28de50f
--- /dev/null
+++ b/utils/callback_handler.py
@@ -0,0 +1,186 @@
+"""
+回调处理器 - 用于实时推送对话内容到前端
+"""
+from typing import Callable, Dict, Any, List, Optional
+from datetime import datetime
+import threading
+import queue
+
+
+class MessageCallbackHandler:
+    """消息回调处理器，支持实时流式输出"""
+    
+    def __init__(self):
+        """初始化回调处理器"""
+        self.callbacks: List[Callable[[Dict[str, Any]], None]] = []
+        self.message_queue = queue.Queue()
+        self.is_running = False
+        self._lock = threading.Lock()
+    
+    def register_callback(self, callback: Callable[[Dict[str, Any]], None]):
+        """
+        注册回调函数
+        
+        Args:
+            callback: 回调函数，接收字典格式的消息
+        """
+        with self._lock:
+            self.callbacks.append(callback)
+    
+    def unregister_callback(self, callback: Callable[[Dict[str, Any]], None]):
+        """注销回调函数"""
+        with self._lock:
+            if callback in self.callbacks:
+                self.callbacks.remove(callback)
+    
+    def on_message(
+        self,
+        agent_name: str,
+        message: str,
+        role: str = "assistant",
+        metadata: Optional[Dict] = None
+    ):
+        """
+        处理新消息
+        
+        Args:
+            agent_name: Agent 名称
+            message: 消息内容
+            role: 角色
+            metadata: 元数据
+        """
+        msg_data = {
+            "timestamp": datetime.now().isoformat(),
+            "agent_name": agent_name,
+            "role": role,
+            "message": message,
+            "metadata": metadata or {}
+        }
+        
+        # 放入队列
+        self.message_queue.put(msg_data)
+        
+        # 调用所有回调
+        with self._lock:
+            for callback in self.callbacks:
+                try:
+                    callback(msg_data)
+                except Exception as e:
+                    print(f"回调执行失败：{e}")
+    
+    def on_thinking(self, agent_name: str, status: str = "thinking"):
+        """
+        发送思考状态
+        
+        Args:
+            agent_name: Agent 名称
+            status: 状态（thinking/generating/coding/testing）
+        """
+        self.on_message(
+            agent_name=agent_name,
+            message=f"_{status}...",
+            role="system",
+            metadata={"status": status}
+        )
+    
+    def on_file_created(self, agent_name: str, file_path: str, file_type: str):
+        """
+        发送文件创建事件
+        
+        Args:
+            agent_name: Agent 名称
+            file_path: 文件路径
+            file_type: 文件类型
+        """
+        self.on_message(
+            agent_name=agent_name,
+            message=f"📄 创建了文件：{file_path}",
+            role="system",
+            metadata={
+                "event_type": "file_created",
+                "file_path": file_path,
+                "file_type": file_type
+            }
+        )
+    
+    def on_test_result(self, agent_name: str, passed: bool, details: str):
+        """
+        发送测试结果
+        
+        Args:
+            agent_name: Agent 名称
+            passed: 是否通过
+            details: 详细信息
+        """
+        icon = "✅" if passed else "❌"
+        self.on_message(
+            agent_name=agent_name,
+            message=f"{icon} 测试结果：{'通过' if passed else '失败'}\n{details}",
+            role="system",
+            metadata={
+                "event_type": "test_result",
+                "passed": passed,
+                "details": details
+            }
+        )
+    
+    def on_human_approval_request(
+        self,
+        request_id: str,
+        description: str,
+        data: Dict[str, Any]
+    ):
+        """
+        发送人工确认请求
+        
+        Args:
+            request_id: 请求 ID
+            description: 请求描述
+            data: 相关数据
+        """
+        self.on_message(
+            agent_name="Orchestrator",
+            message=f"⚠️ 需要人工确认：{description}",
+            role="system",
+            metadata={
+                "event_type": "human_approval",
+                "request_id": request_id,
+                "description": description,
+                "data": data
+            }
+        )
+    
+    def get_message(self, timeout: float = 1.0) -> Optional[Dict[str, Any]]:
+        """
+        从队列获取消息（非阻塞）
+        
+        Args:
+            timeout: 超时时间
+        
+        Returns:
+            消息字典或 None
+        """
+        try:
+            return self.message_queue.get(timeout=timeout)
+        except queue.Empty:
+            return None
+    
+    def clear_queue(self):
+        """清空消息队列"""
+        while not self.message_queue.empty():
+            try:
+                self.message_queue.get_nowait()
+            except queue.Empty:
+                break
+
+
+# 全局回调实例
+_global_callback: Optional[MessageCallbackHandler] = None
+
+
+def get_callback_handler() -> MessageCallbackHandler:
+    """获取或创建全局回调处理器"""
+    global _global_callback
+    if _global_callback is None:
+        _global_callback = MessageCallbackHandler()
+    return _global_callback
diff --git a/utils/logger.py b/utils/logger.py
new file mode 100644
index 0000000..d069ed3
--- /dev/null
+++ b/utils/logger.py
@@ -0,0 +1,192 @@
+"""
+日志工具模块 - 记录多智能体对话历史和系统事件
+"""
+import os
+import json
+import logging
+from datetime import datetime
+from typing import Dict, List, Any, Optional
+from pathlib import Path
+
+
+class ConversationLogger:
+    """对话历史日志记录器"""
+    
+    def __init__(self, log_dir: str = "logs"):
+        """
+        初始化日志记录器
+        
+        Args:
+            log_dir: 日志目录
+        """
+        self.log_dir = Path(log_dir)
+        self.log_dir.mkdir(parents=True, exist_ok=True)
+        
+        # 创建会话 ID
+        self.session_id = datetime.now().strftime("%Y%m%d_%H%M%S")
+        self.log_file = self.log_dir / f"session_{self.session_id}.jsonl"
+        
+        # 配置日志
+        self.logger = logging.getLogger(f"autogen_sdls_{self.session_id}")
+        self.logger.setLevel(logging.INFO)
+        
+        # 文件处理器
+        file_handler = logging.FileHandler(self.log_file, encoding='utf-8')
+        file_handler.setLevel(logging.INFO)
+        
+        # 控制台处理器
+        console_handler = logging.StreamHandler()
+        console_handler.setLevel(logging.INFO)
+        
+        # 格式化器
+        formatter = logging.Formatter(
+            '%(asctime)s - %(name)s - %(levelname)s - %(message)s',
+            datefmt='%Y-%m-%d %H:%M:%S'
+        )
+        file_handler.setFormatter(formatter)
+        console_handler.setFormatter(formatter)
+        
+        self.logger.addHandler(file_handler)
+        self.logger.addHandler(console_handler)
+        
+        # 对话历史
+        self.conversation_history: List[Dict[str, Any]] = []
+    
+    def log_message(
+        self,
+        agent_name: str,
+        message: str,
+        role: str = "assistant",
+        metadata: Optional[Dict] = None
+    ):
+        """
+        记录单条消息
+        
+        Args:
+            agent_name: Agent 名称
+            message: 消息内容
+            role: 角色（user/assistant/system）
+            metadata: 额外元数据
+        """
+        entry = {
+            "timestamp": datetime.now().isoformat(),
+            "session_id": self.session_id,
+            "agent_name": agent_name,
+            "role": role,
+            "message": message,
+            "metadata": metadata or {}
+        }
+        
+        self.conversation_history.append(entry)
+        
+        # 写入 JSONL 文件
+        with open(self.log_file, 'a', encoding='utf-8') as f:
+            f.write(json.dumps(entry, ensure_ascii=False) + '\n')
+        
+        self.logger.info(f"[{agent_name}] {message[:100]}...")
+    
+    def log_event(self, event_type: str, description: str, data: Optional[Dict] = None):
+        """
+        记录系统事件
+        
+        Args:
+            event_type: 事件类型
+            description: 事件描述
+            data: 相关数据
+        """
+        entry = {
+            "timestamp": datetime.now().isoformat(),
+            "session_id": self.session_id,
+            "event_type": event_type,
+            "description": description,
+            "data": data or {}
+        }
+        
+        with open(self.log_file, 'a', encoding='utf-8') as f:
+            f.write(json.dumps(entry, ensure_ascii=False) + '\n')
+        
+        self.logger.info(f"[EVENT] {event_type}: {description}")
+    
+    def get_conversation_history(self) -> List[Dict[str, Any]]:
+        """获取完整的对话历史"""
+        return self.conversation_history
+    
+    def export_to_json(self, output_path: Optional[str] = None) -> str:
+        """
+        导出对话历史为 JSON 格式
+        
+        Args:
+            output_path: 输出路径，默认在 logs 目录下
+        
+        Returns:
+            导出文件的路径
+        """
+        if output_path is None:
+            output_path = self.log_dir / f"conversation_{self.session_id}.json"
+        else:
+            output_path = Path(output_path)
+        
+        export_data = {
+            "session_id": self.session_id,
+            "created_at": datetime.now().isoformat(),
+            "total_messages": len(self.conversation_history),
+            "conversation": self.conversation_history
+        }
+        
+        with open(output_path, 'w', encoding='utf-8') as f:
+            json.dump(export_data, f, ensure_ascii=False, indent=2)
+        
+        self.logger.info(f"对话历史已导出到：{output_path}")
+        return str(output_path)
+    
+    def export_to_markdown(self, output_path: Optional[str] = None) -> str:
+        """
+        导出对话历史为 Markdown 格式
+        
+        Args:
+            output_path: 输出路径
+        
+        Returns:
+            导出文件的路径
+        """
+        if output_path is None:
+            output_path = self.log_dir / f"conversation_{self.session_id}.md"
+        else:
+            output_path = Path(output_path)
+        
+        with open(output_path, 'w', encoding='utf-8') as f:
+            f.write(f"# AutoGen SDLC 对话历史\n\n")
+            f.write(f"**会话 ID**: {self.session_id}\n")
+            f.write(f"**创建时间**: {datetime.now().strftime('%Y-%m-%d %H:%M:%S')}\n")
+            f.write(f"**总消息数**: {len(self.conversation_history)}\n\n")
+            f.write("---\n\n")
+            
+            for entry in self.conversation_history:
+                timestamp = entry["timestamp"][:19].replace('T', ' ')
+                agent = entry["agent_name"]
+                message = entry["message"]
+                
+                f.write(f"### [{timestamp}] {agent}\n\n")
+                f.write(f"{message}\n\n")
+                f.write("---\n\n")
+        
+        self.logger.info(f"对话历史已导出为 Markdown: {output_path}")
+        return str(output_path)
+
+
+# 全局日志实例
+_global_logger: Optional[ConversationLogger] = None
+
+
+def get_logger(log_dir: str = "logs") -> ConversationLogger:
+    """获取或创建全局日志记录器"""
+    global _global_logger
+    if _global_logger is None:
+        _global_logger = ConversationLogger(log_dir)
+    return _global_logger
+
+
+def reset_logger():
+    """重置全局日志记录器"""
+    global _global_logger
+    _global_logger = None
diff --git a/verify_v3_code.py b/verify_v3_code.py
new file mode 100644
index 0000000..d094546
--- /dev/null
+++ b/verify_v3_code.py
@@ -0,0 +1,157 @@
+# -*- coding: utf-8 -*-
+"""
+Verify streamlit_app_v3.py code structure and functionality
+"""
+import sys
+from pathlib import Path
+
+print("=" * 70)
+print("Streamlit V3 - Code Verification")
+print("=" * 70)
+print()
+
+# 1. Check file existence
+print("1. Checking file existence...")
+v3_file = Path("frontend/streamlit_app_v3.py")
+if v3_file.exists():
+    print(f"   [OK] frontend/streamlit_app_v3.py exists ({v3_file.stat().st_size:,} bytes)")
+else:
+    print(f"   [ERROR] frontend/streamlit_app_v3.py not found")
+    sys.exit(1)
+print()
+
+# 2. Check key functions
+print("2. Checking key functions...")
+with open(v3_file, 'r', encoding='utf-8') as f:
+    content = f.read()
+
+required_functions = [
+    "init_session_state",
+    "display_agent_status_row",
+    "display_active_agent_banner",
+    "display_chat_flow",
+    "create_agents",
+    "add_message",
+    "main"
+]
+
+for func in required_functions:
+    if f"def {func}(" in content:
+        print(f"   [OK] {func}() defined")
+    else:
+        print(f"   [MISSING] {func}() not found")
+print()
+
+# 3. Check Agent configuration
+print("3. Checking Agent configuration...")
+agents = ["PM_Agent", "QA_Agent", "Dev_Agent", "Orchestrator", "User_Proxy"]
+for agent in agents:
+    if f'"{agent}"' in content:
+        print(f"   [OK] {agent} configured")
+    else:
+        print(f"   [MISSING] {agent} not configured")
+print()
+
+# 4. Check CSS styles
+print("4. Checking CSS styles...")
+css_classes = [
+    "agent-status-card",
+    "agent-chat-bubble",
+    "active-agent-banner",
+    "task-badge",
+    "chat-flow-indicator"
+]
+
+for css_class in css_classes:
+    if css_class in content:
+        print(f"   [OK] .{css_class} defined")
+    else:
+        print(f"   [MISSING] .{css_class} not found")
+print()
+
+# 5. Check core features
+print("5. Checking core features...")
+features = {
+    "Real-time Agent Status": "agent_status",
+    "Task Badge": "task",
+    "Chat Flow": "chat_flow",
+    "Current Agent": "current_agent",
+    "Flow Arrow": "flow-arrow"
+}
+
+for feature_name, feature_key in features.items():
+    if feature_key in content:
+        print(f"   [OK] {feature_name} implemented")
+    else:
+        print(f"   [MISSING] {feature_name} not found")
+print()
+
+# 6. Check imports
+print("6. Checking imports...")
+required_imports = [
+    "import streamlit as st",
+    "from autogen import",
+    "from config.llm_config import"
+]
+
+for imp in required_imports:
+    if imp in content:
+        print(f"   [OK] {imp}")
+    else:
+        print(f"   [MISSING] {imp}")
+print()
+
+# 7. Syntax check
+print("7. Python syntax check...")
+import py_compile
+try:
+    py_compile.compile(str(v3_file), doraise=True)
+    print(f"   [OK] Syntax valid")
+except py_compile.PyCompileError as e:
+    print(f"   [ERROR] Syntax error: {e}")
+print()
+
+# 8. File structure comparison
+print("8. Comparing with v2...")
+v2_file = Path("frontend/streamlit_app.py")
+v3_file_path = Path("frontend/streamlit_app_v3.py")
+
+print(f"   v2 (streamlit_app.py): {v2_file.stat().st_size:,} bytes")
+print(f"   v3 (streamlit_app_v3.py): {v3_file_path.stat().st_size:,} bytes")
+if v2_file.exists():
+    size_diff = v3_file_path.stat().st_size - v2_file.stat().st_size
+    print(f"   Size increase: {size_diff:,} bytes")
+print()
+
+# 9. Summary
+print("=" * 70)
+print("Verification Summary")
+print("=" * 70)
+print()
+
+all_checks_passed = True
+checks = [
+    v3_file.exists(),
+    all(f"def {func}(" in content for func in required_functions),
+    all(f'"{agent}"' in content for agent in agents),
+    all(css_class in content for css_class in css_classes),
+    all(feature_key in content for feature_key in features.values()),
+]
+
+if all(checks):
+    print("[OK] All checks passed!")
+    print()
+    print("New features in V3:")
+    print("  - Real-time Agent Status Row (with task badges)")
+    print("  - Active Agent Banner (green highlight)")
+    print("  - Chat Flow Display (colored bubbles)")
+    print("  - Task Badge System")
+    print("  - Flow Direction Arrows")
+    print("  - Timestamp Display")
+    print()
+    print("Usage:")
+    print("  streamlit run frontend/streamlit_app_v3.py")
+else:
+    print("Warning: Some checks failed. Please review the output above.")
+
+print()