1) Collector Service
A persistent non-LLM collector runs on schedule, fetches BTCUSDT candles from Binance, and writes idempotent rows to Postgres.
Reliability: retries, endpoint fallback, and rate-limit awareness.
First Next.js App
Welcome to Crypto Lab
A living research system for BTC market behavior — collecting fresh data every 4 hours, classifying volatility + regimes, backtesting ideas, simulating risk, and emailing a daily analyst-style brief.
2) Postgres Data Core
All core data lands in one place: raw bars, engineered features, run summaries, and trade logs.
Tables include ohlcv_bars, feature_bars, backtest_runs, and backtest_trades.
3) Daily Analysis Email
A scheduled analysis worker generates a daily report, appends an LLM analysis section, and sends it by email.
Result: concise insights, risks, and next-step experiments every day.
How It Thinks About Trading
- Classifies volatility into low / medium / high regimes.
- Switches strategy behavior between trend and mean-reversion conditions.
- Backtests rule sets with fees and position transitions.
- Runs Monte Carlo simulations for drawdown and tail-risk estimates.