{"repo":"7thDrag0n/AutoEvolve","free":true,"listed":false,"github":"https://github.com/7thDrag0n/AutoEvolve","clone":"git clone https://github.com/7thDrag0n/AutoEvolve.git","description":"AutoEvolve is a self-improving trading bot system built on top of FreqTrade. It uses a large language model (LLM) to continuously analyse trading performance, rewrite its own strategy code, deploy it, and roll back if results worsen — all without human intervention.","language":"Python","stars":10,"topics":["auto","bot","framework","freqtrade","llm"],"license":"GPL-3.0","category":"ai-agents","readme_excerpt":"🤖 AutoEvolve AutoEvolve is a self-improving trading bot system built on top of FreqTrade. It uses a large language model (LLM) to continuously analyse trading performance, rewrite its own strategy code, deploy it, and roll back if results worsen — all without human intervention. ⚠️ This is experimental software. Trade only what you can afford to lose. Always test in dry-run mode first. --- How It Works AutoEvolve runs alongside FreqTrade and watches its performance in a continuous loop: 1. Watch — Every 60 seconds, reads the FreqTrade SQLite database and computes metrics (win rate, profit factor, Sharpe, drawdown, consecutive losses, profit drawdown from peak) 2. Trigger — When a trigger condition fires (too many losses, profit eroding, idle strategy, etc.), an evolution cycle begins 3. Analyse — The LLM receives the current strategy code, all performance metrics, recent trade history, and the full evolution log 4. Rewrite — The LLM produces an improved strategy in two phases: ANALYSE (diagnose what's wrong, plain English) then CODE (implement the fix) 5. Deploy — The new strategy is validated, sanitised, and deployed to FreqTrade via hot-reload (no restart needed) 6. Monitor — The next N trades are observed. If performance drops below thresholds, the previous generation is automatically restored 7. Fix errors — If FreqTrade reports a strategy runtime error, AutoEvolve calls the LLM to fix the code and redeploys the same generation (no generation number burned) --- Features ","default_branch":null,"files":null,"tree":[],"storefront":"/r/7thDrag0n","claimed":false,"request_supported":{"post":"https://gitbuyer.com/r/7thDrag0n/AutoEvolve/request-supported","requests":0},"note":"indexed from public GitHub; nothing is for sale on this page. Clone it from GitHub. Paid listings live at /search."}