多智能体开发系统

# 1. Clone repository
git clone <your-repo-url>
cd PM-Agents

# 2. Install Python dependencies
python -m venv venv
source venv/bin/activate  # Windows: venv\Scripts\activate
pip install -r requirements.txt

# 3. Install Node dependencies (for MCP servers)
npm install -g @modelcontextprotocol/server-filesystem
npm install -g @modelcontextprotocol/server-github
npm install -g @modelcontextprotocol/server-brave-search

# 4. Start Qdrant (vector database)
docker run -p 6333:6333 qdrant/qdrant

# 5. Set environment variables
export ANTHROPIC_API_KEY="your-key"
export GITHUB_TOKEN="your-token"
export BRAVE_API_KEY="your-key"

from pm_coordinator_agent import PMCoordinatorAgent

# Initialize coordinator
coordinator = PMCoordinatorAgent()

# Define your project
project = "Develop an AI-powered customer service chatbot for e-commerce platform"

# Execute project management
result = coordinator.manage_project(project)

# View results
print(result)

python pm_coordinator_agent.py

from pm_ollama_agents import OllamaCoordinator

# Initialize coordinator (Ollama must be running)
coordinator = OllamaCoordinator()

# Define your project
project = """
Develop an AI-powered project management assistant that helps teams 
automate task scheduling, resource allocation, and risk monitoring.
"""

# Execute project management
result = coordinator.manage_project(project)

# Results are automatically saved to pm_project_results.json

# Make sure Ollama is running first!
python pm_ollama_agents.py

User Request
     ↓
Coordinator Agent (Analyzes request, determines strategy)
     ↓
Planner Agent (Creates detailed plan, assigns agents)
     ↓
Supervisor Agent (Manages execution, coordinates specialists)
     ↓
┌────────────┬────────────┬────────────┬────────────┬────────────┐
│ Spec-Kit   │ Frontend   │ Python ML  │ R          │ Qdrant     │
│ Agent      │ Coder      │ Agent      │ Analytics  │ Vector     │
└────────────┴────────────┴────────────┴────────────┴────────────┘
     ↓            ↓            ↓            ↓            ↓
     └────────────┴────────────┴────────────┴────────────┘
                              ↓
                    TypeScript Validator
                    (Reviews code quality)
                              ↓
     ┌────────────┬────────────┬────────────┐
     │ Research   │ Browser    │ Reporter   │
     │ Agent      │ Agent      │ Agent      │
     └────────────┴────────────┴────────────┘
                              ↓
                    Final Deliverables
                              ↓
                    Back to Coordinator
                              ↓
                    User receives results

# Coordinator → Planner
analysis = coordinator.analyze_request(user_request)
plan = coordinator.delegate_to_planner(analysis)

# Planner → Supervisor
execution_plan = planner.create_detailed_plan(analysis)
results = planner.delegate_to_supervisor(execution_plan)

# Supervisor → Specialists
for task in execution_plan.tasks:
    specialist = supervisor.get_specialist(task.agent_type)
    result = specialist.process_task(task)
    supervisor.aggregate_result(result)

# Results flow back up the hierarchy

PM-Agents/
├── README.md                          # This file
├── AGENT_ARCHITECTURE.md              # Detailed agent specifications
├── claudcodesetup.md                  # Complete setup guide
├── TOOLS.md                           # All required tools and installation
├── architecture.drawio                # Architecture diagram (open in Draw.io)
├── requirements.txt                   # Python dependencies
├── package.json                       # Node.js dependencies (MCP servers)
│
├── agents/                            # Agent implementations
│   ├── base_agent.py                  # Base agent class
│   ├── coordinator/
│   │   └── coordinator_agent.py
│   ├── planner/
│   │   └── planner_agent.py
│   ├── supervisor/
│   │   └── supervisor_agent.py
│   └── specialists/
│       ├── speckit/                   # Project initialization
│       ├── qdrant/                    # Vector storage
│       ├── frontend/                  # React/Next.js development
│       ├── python_ml/                 # PyTorch ML/DL
│       ├── r_analytics/               # R statistical analysis
│       ├── ts_validator/              # TypeScript validation
│       ├── research/                  # Research tasks
│       ├── browser/                   # Web automation
│       └── reporter/                  # Documentation
│
├── configs/                           # Configuration files
│   ├── mcp_config.json                # MCP server configuration
│   └── agent_prompts.json             # Agent prompt templates
│
├── tools/                             # Utility scripts
│   ├── init_qdrant.py                 # Initialize Qdrant collections
│   └── index_codebase.py              # Index project code
│
├── templates/                         # Project templates
│   ├── nextjs-supabase/               # Frontend template
│   ├── pytorch-ml/                    # ML project template
│   └── r-analysis/                    # R analysis template
│
└── examples/                          # Example projects
    ├── web_app_example/
    ├── ml_pipeline_example/
    └── analytics_dashboard_example/