354 lines
12 KiB
Python
354 lines
12 KiB
Python
"""
|
||
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
|
||
import re
|
||
from datetime import datetime
|
||
|
||
# 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
|
||
|
||
|
||
def extract_code(content):
|
||
"""从 Markdown 代码块中提取纯代码"""
|
||
if "```python" in content:
|
||
parts = content.split("```python")
|
||
if len(parts) > 1:
|
||
code = parts[1].split("```")[0].strip()
|
||
return code
|
||
elif "```" in content:
|
||
parts = content.split("```")
|
||
if len(parts) > 1:
|
||
code = parts[1].strip()
|
||
return code
|
||
return content
|
||
|
||
|
||
def save_file_from_content(content: str, agent_name: str, workspace_dir: Path) -> List[str]:
|
||
"""
|
||
根据 Agent 的内容实时保存文件
|
||
|
||
Args:
|
||
content: Agent 的输出内容
|
||
agent_name: Agent 名称
|
||
workspace_dir: 工作目录
|
||
|
||
Returns:
|
||
保存的文件路径列表
|
||
"""
|
||
saved_files = []
|
||
|
||
# PM Agent 生成 SRS
|
||
if agent_name == "PM_Agent" and ("需求" in content or "SRS" in content or "软件需求规格说明书" in content):
|
||
file = workspace_dir / "SRS.md"
|
||
with open(file, "w", encoding="utf-8") as f:
|
||
f.write(f"# 软件需求规格说明书\n\n生成时间：{datetime.now().strftime('%Y-%m-%d %H:%M:%S')}\n\n")
|
||
f.write(content)
|
||
saved_files.append(str(file))
|
||
print(f"💾 [实时保存] {file}")
|
||
|
||
# QA Agent 生成测试
|
||
elif agent_name == "QA_Agent" and ("test" in content.lower() or "测试" in content or "def test_" in content):
|
||
file = workspace_dir / "test_sample.py"
|
||
code = extract_code(content)
|
||
if code: # 确保有代码内容才保存
|
||
with open(file, "w", encoding="utf-8") as f:
|
||
f.write(f"# 测试用例\n\n{code}")
|
||
saved_files.append(str(file))
|
||
print(f"💾 [实时保存] {file}")
|
||
|
||
# Dev Agent 生成代码 - 支持多个文件
|
||
elif agent_name == "Dev_Agent" and ("def " in content or "class " in content or "import " in content):
|
||
# 尝试提取带文件名的代码块
|
||
code_blocks = re.findall(r'```python\s*\n#(?:\s*File:|\s*filename:)?\s*([^\n]+)\n(.*?)```', content, re.DOTALL)
|
||
|
||
if code_blocks:
|
||
# 保存多个文件
|
||
for filename, code_content in code_blocks:
|
||
filename = filename.strip()
|
||
file_path = workspace_dir / filename
|
||
with open(file_path, "w", encoding="utf-8") as f:
|
||
f.write(f"# 生成时间：{datetime.now().strftime('%Y-%m-%d %H:%M:%S')}\n\n{code_content}")
|
||
saved_files.append(str(file_path))
|
||
print(f"💾 [实时保存] {file_path}")
|
||
else:
|
||
# 没有文件名标记，保存为 src_sample.py
|
||
file = workspace_dir / "src_sample.py"
|
||
code = extract_code(content)
|
||
if code: # 确保有代码内容才保存
|
||
with open(file, "w", encoding="utf-8") as f:
|
||
f.write(f"# 源代码\n\n{code}")
|
||
saved_files.append(str(file))
|
||
print(f"💾 [实时保存] {file}")
|
||
|
||
return saved_files
|
||
|
||
|
||
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
|
||
|
||
# 记录已保存的文件
|
||
self.saved_files = []
|
||
|
||
def _on_new_message(self, sender, message, **kwargs):
|
||
"""
|
||
新消息回调函数 - 实时保存文件
|
||
|
||
Args:
|
||
sender: 发送者
|
||
message: 消息内容
|
||
**kwargs: 其他参数
|
||
"""
|
||
try:
|
||
agent_name = getattr(sender, 'name', 'Unknown')
|
||
content = message.get('content', '') if isinstance(message, dict) else str(message)
|
||
|
||
# 实时保存文件
|
||
saved_files = save_file_from_content(content, agent_name, self.workspace_dir)
|
||
|
||
# 如果有文件保存，记录并通知回调
|
||
if saved_files:
|
||
self.saved_files.extend(saved_files)
|
||
for file_path in saved_files:
|
||
# 确定文件类型
|
||
if file_path.endswith('.md'):
|
||
file_type = 'markdown'
|
||
elif file_path.endswith('.py'):
|
||
file_type = 'python'
|
||
else:
|
||
file_type = 'other'
|
||
|
||
# 触发回调
|
||
self.callback_handler.on_file_created(agent_name, file_path, file_type)
|
||
except Exception as e:
|
||
print(f"⚠️ 实时保存文件时出错：{e}")
|
||
|
||
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=10, # 允许连续自动回复，避免提前终止
|
||
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)
|
||
|
||
# 重置已保存文件列表
|
||
self.saved_files = []
|
||
|
||
# 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 - 使用回调函数实时处理消息
|
||
print("\n🚀 开始对话...\n")
|
||
|
||
# 为每个 Agent 注册实时文件保存回调
|
||
for agent in [self.pm_agent, self.qa_agent, self.dev_agent, self.orchestrator]:
|
||
agent.register_reply(
|
||
self._on_new_message,
|
||
reply_func_total_count=1,
|
||
position=0 # 在最前面执行，确保实时保存
|
||
)
|
||
|
||
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),
|
||
"saved_files": self.saved_files
|
||
}
|
||
|
||
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 []
|
||
} |