Highlights
Deterministic quant pipeline: data, features, strategies, backtest, risk, reports
LightGBM direction model + walk-forward OOS evaluation
161 passing pytest cases (mock/synthetic in tests)
Agent layer: SupervisorAgent, ResearchAgent, optional LLM narration (no live orders)
Memory/RAG v2, optional LangGraph workflow
Text signal layer (M6 skeleton)
Verified research baseline
Walk-forward OOS sharpe 0.586 (EXP-20260602-008, server walk-forward)
Install
git clone https://github.com/ytq0198/Quant-MAS.git
cd Quant-MAS
python -m pip install -e .
python -m pytest -v
Not included
No live trading / broker integration
M7 RL / M8 MCP are roadmap only (not implemented)
MIT License · Research & education only — not financial advice.