房产搜索代理

# Create and activate a conda environment (or venv)
conda create -n rental-search python=3.12
conda activate rental-search

# Install Python dependencies
pip install -r requirements-dev.txt

# Install Node dependencies (Playwright MCP)
npm install

# Configure environment
cp .env.example .env
# Edit .env and fill in:
#   OPENAI_API_KEY       — required (LangGraph agent)
#   GOOGLE_MAPS_API_KEY  — optional (school-distance filtering)

uvicorn main:app --reload
# or via skill:
/dev-server

pytest
# or via skill:
/lint

.
├── main.py                  # FastAPI app, lifespan, routes
├── config.py                # Settings from .env via pydantic-settings
├── agent/
│   ├── __init__.py          # LangGraph ReAct property search agent
│   ├── chat_graph.py        # LangGraph chat workflow (async/sync tool routing)
│   ├── enrichment.py        # School-distance filtering via Google Maps API
│   ├── models.py            # PropertyListing Pydantic model
│   └── url_builder.py       # domain.com.au URL builder + listing parser
├── api/
│   ├── chat.py              # POST /chat endpoint
│   ├── search.py            # POST /search + SSE event translation
│   └── stream.py            # GET /stream/{run_id} SSE endpoint
├── mcp_subprocess/          # Playwright MCP subprocess lifecycle
├── templates/
│   └── index.html           # Single-page chat UI (noir concierge design)
├── tests/                   # pytest unit tests
├── requirements.txt         # Runtime dependencies
├── requirements-dev.txt     # Dev dependencies
└── .env.example             # Environment variable reference