- Python 3.10+
- Node.js 18+
# Backend
pip install -r backend/requirements.txt
# Frontend
cd frontend
npm installBackend (Trading Core):
python -m uvicorn backend.main:app --reloadFrontend (System Console):
cd frontend
npm run devAccess the System Console at http://localhost:3000.
- Click INITIALIZE to start the engine.
- View real-time decision logs in the terminal window.
- Check "Component Status" for health.
Run BTC-trained 7-year daily model research and evaluate on ETH + 10 other crypto pairs:
$env:PYTHONPATH="C:\Users\paras\Documents\antigravity files\TB 1AG"
python tests/run_crypto_autoresearch_7y.pyOutputs:
- JSON report:
data/reports/crypto_autoresearch_7y_*.json - CSV report:
data/reports/crypto_autoresearch_7y_*.csv - Daily verification folder:
data/reports/verification_*/
The verification output includes:
model_positionvsoracle_positionfor every daymodel_action, expected next-day edge, position size, and switching cost columns- daily return gap (
oracle_return - model_return) - monthly/yearly return breakdown in the JSON report
Walk-forward rule in pipeline:
- Training uses oracle action labels only from historical training windows
- Final evaluation oracle paths are computed after walk-forward scoring for verification and gap scoring
Run local-only iterative training/evaluation with daily verification and compounded equity columns:
$env:PYTHONPATH="C:\Users\paras\Documents\antigravity files\TB 1AG"
python tests/run_crypto_local_autoloop.py --years 1 --iterations 4 --month-stride 1 --initial-capital 1000Windows executable wrapper:
run_local_crypto_autoloop.bat --years 1 --iterations 4 --month-stride 1 --initial-capital 1000Artifacts:
- Loop summary CSV/JSON:
data/reports/local_crypto_autoloop_*/ - Per-run report JSON/CSV:
data/reports/crypto_autoresearch_*y_*.json|csv - Per-coin daily verification CSV (includes compounded equity):
data/reports/verification_*/