A monorepo for quantitative trading research and production: market data infrastructure, feature engineering, regime detection, simulation, XGBoost-based alpha modeling, and V7 policy-driven execution.
847 tests passing. Python 3.12+.
v7-engine/
├── lib/ ← shared primitives (Binance client, indicators, costs, time)
├── simulation/ ← economic simulation truth authority (engine, Monte Carlo, contracts)
├── alphaforge/ ← AlphaForge research: labels, features, XGBoost, calibration, policy
├── v7/ ← V7 pipeline: ModelRegistry, inference, AnalysisResult builder
├── policycritic/ ← Policy Critic: IQL-based advisory offline-RL component
├── runtime/ ← Python backend (FastAPI, scan loop, analyzer, learning, DB)
├── interface/ ← React + TypeScript + Vite operator UI
├── contracts/ ← cross-domain schemas, registry, field mappings
├── integration/ ← adapter stubs, cross-domain tests, boundary enforcement
├── v6/ ← stub for legacy V6 config compatibility
├── train_pipeline.py ← end-to-end AlphaForge training pipeline
├── Dockerfile ← containerized test environment
├── compose.yaml ← Docker Compose for development
├── pyproject.toml ← pip install -e . support
└── requirements.txt ← pinned dependencies
# Install
pip install -e .
# Run all tests (847)
make test-all
# Or manually
PYTHONPATH=alphaforge/src:v7/src python -m pytest lib/tests/ simulation/tests/ \
integration/tests/ runtime/tests/ alphaforge/tests/ v7/tests/ policycritic/tests/ -q
# Docker
docker compose up --build| Suite | Count |
|---|---|
lib/tests/ |
182 |
simulation/tests/ |
73 |
integration/tests/ |
96 |
runtime/tests/ |
374 |
alphaforge/tests/ |
112 |
v7/tests/ |
6 |
policycritic/tests/ |
7 |
| Total | 847 |
| Domain | Imports | Forbidden Imports |
|---|---|---|
lib/ |
nothing | v7, alphaforge, simulation |
simulation/ |
lib | v7, alphaforge |
alphaforge/ |
lib | simulation, v7 |
v7/ |
lib | simulation (lazy importlib only) |
runtime/ |
lib, v6 | alphaforge (via services) |
integration/ |
nothing | simulation, alphaforge, v7 |
AlphaForge P1–P9 complete (label builder → evaluation/monitoring).
End-to-end training: python train_pipeline.py.