Automated system for trading on Polymarket prediction markets: data collection → feature calculation → ML models → backtesting → Execution Bot.

Python 3.11+

Docker (optional, for Railway or local PostgreSQL)

Service Description Web API FastAPI, Swagger UI (/docs), REST API (/api/v1/*). Full pipeline in background (hourly) collector Polymarket API data collection (Gamma markets, CLOB trades with optional L2 auth, orderbook). Filters to liquid markets news_collector RSS news collection (Google News, CoinDesk, CoinTelegraph, NY Times) with keyword filtering feature_store Features: MA, volatility, RSI, MACD, spread, volume. Prioritizes liquid markets ml_module Model training (Logistic Regression, Random Forest, XGBoost), walk-forward validation, signal generation. Multi-period target horizon backtester Simulation with fees (30 bps), slippage (10 bps), signals 1/0/-1 execution_bot Order placement via py-clob-client (dry-run by default)

See docs/RAILWAY.md:

FastAPI + Swagger, PostgreSQL

Full pipeline in background (hourly): cleanup → collector → news → features → ML → backtest

Optional: CLOB auth for real trade data

Configure .env for target environment

Migrations run automatically on startup

For real trading: POLYMARKET_DRY_RUN=false, set POLYMARKET_PRIVATE_KEY