A self-evolving AI trading agent that learns from every trade it makes.
Galeon is not another signal bot or rule-based trading tool. It is an autonomous trading cognition system that perceives multi-dimensional market data, makes decisions, executes trades, and — most importantly — automatically learns from outcomes to continuously improve its own strategy.
Over 2,000+ live trades executed, achieving 58% win rate and 250%+ cumulative profit, with performance still improving as the system evolves.
Most trading bots follow a fixed playbook: when rules work, they profit; when the market shifts, they bleed — until a human manually re-tunes them. Galeon closes this loop autonomously.
Perceive ──► Cognize ──► Decide ──► Execute
▲ │
│ ▼
Evolve ◄── Backtest ◄── Analyze ◄── Review
Every trade Galeon completes feeds back into the system:
- Perceive — Ingest real-time multi-dimensional market data
- Cognize — Identify token lifecycle stage, market regime, and signal alignment
- Decide — Multi-dimensional voting produces direction and confidence score
- Execute — Enter positions with dynamic sizing and risk controls
- Review — On exit, auto-verify what went right and wrong
- Analyze — Attribute outcomes to specific dimensions and rules
- Backtest — Validate proposed parameter changes against historical data
- Evolve — Apply validated adjustments, with overfitting protection
The rules that govern Galeon today were not written by humans — they were discovered by the system itself from 2,000+ trades.
Galeon's edge comes from synthesizing signals that no single-dimension bot can capture:
| Layer | Dimensions | Purpose |
|---|---|---|
| On-Chain | Smart Money flow, Buy/Sell Ratio, holder distribution | Detect "smart money" intent before price moves |
| Derivatives | Funding rate, Open Interest stages, Taker ratio, Top Trader positions | Read contract market microstructure |
| Technical | Multi-timeframe momentum, EMA channels, K-line patterns, RSI | Identify trend and entry timing |
| Macro | BTC/ETH correlation, market regime (bull/bear/transition), sentiment cycles | Context-aware strategy weighting |
| LLM Cognition | ChatGPT-powered reasoning for complex multi-signal scenarios | Handle ambiguity that rules can't |
On-chain data is sourced via Bitget Agent Skill, providing real-time multi-dimensional chain analytics.
Risk management is not a single stop-loss — it's a multi-layer defense system that also learns:
- Red Line Layer — Price anomaly detection (crash > 50% auto-blocked), honeypot/rug scoring, extreme BTC drawdown halt
- Self-Learned Blocking — System automatically identifies and blocks stage + direction combinations with historically low win rates. These rules emerge from data, not human intuition
- Time-Decay Stop Loss — The longer a position is held at a loss, the tighter the stop becomes. Prevents "hold and hope" behavior
- Loss Cooldown — After a losing exit, cooldown period scales with loss severity. Prevents revenge trading on the same token
┌──────────────────────────────────────────────────────────────┐
│ Frontend (React + TypeScript) │
│ Dashboard · Trade Monitor · Learning Reports · P&L Analytics│
└──────────────────────────┬───────────────────────────────────┘
│ REST API + WebSocket
┌──────────────────────────┴───────────────────────────────────┐
│ Galeon Brain Engine │
│ │
│ ┌─────────────┐ ┌──────────────┐ ┌─────────────────────┐ │
│ │ Perception │ │ Cognition │ │ Control System │ │
│ │ │ │ │ │ │ │
│ │ • On-Chain │─►│ • Stage ID │─►│ • Confidence Gate │ │
│ │ • Derivs │ │ • Rules Eng │ │ • Red Line Checks │ │
│ │ • Technical │ │ • LLM Reason │ │ • Position Sizing │ │
│ │ • Macro │ │ • Voting │ │ • Dynamic Params │ │
│ └─────────────┘ └──────────────┘ └──────────┬──────────┘ │
│ │ │
│ ┌──────────────────────────────────────────────┴──────────┐ │
│ │ Execution Layer │ │
│ │ Entry · Partial Exit · Staged TP · Time-Decay SL │ │
│ └──────────────────────────┬──────────────────────────────┘ │
│ │ │
│ ┌──────────────────────────┴──────────────────────────────┐ │
│ │ Evolution Layer │ │
│ │ Auto-Verify · Attribution · RuleEvolver · Backtest │ │
│ │ Learning Reports · Overfitting Protection │ │
│ └─────────────────────────────────────────────────────────┘ │
└──────────────────────────────────────────────────────────────┘
│
┌────────────┴────────────┐
│ Data Sources │
│ • Bitget Agent Skill │
│ • Derivatives API │
│ • WebSocket Streams │
│ • LLM (ChatGPT) │
└─────────────────────────┘
| Metric | Value |
|---|---|
| Total Trades | 2,000+ |
| Win Rate | 58% |
| Cumulative Profit | 250%+ |
| Self-Learned Rules | Auto-generated from trade data |
| Auto Parameter Adjustments | 500+ (with overfitting protection) |
| Uptime | Continuous 24/7 operation |
| Component | Technology |
|---|---|
| Backend | Node.js, Express |
| Frontend | React, TypeScript, Ant Design |
| Database | MySQL |
| LLM | ChatGPT (complex scenario reasoning) |
| On-Chain Data | Bitget Agent Skill |
| Real-time | WebSocket price streams |
| Evolution | Custom RuleEvolver with bounded optimization |
- Node.js >= 16.0.0
- MySQL 5.7+
git clone https://github.com/Gameland0/Galeon.git
cd Galeon
# Backend
cd server && npm install
cp .env.example .env # Configure your API keys
npm start
# Frontend
cd ../dapp && npm install
npm start- Multi-dimensional perception (on-chain + derivatives + technical + macro)
- Rules Engine + LLM dual-track cognition
- Self-evolution loop (verify → attribute → adjust → backtest)
- Multi-layer risk control with self-learned blocking rules
- Cross-token correlation analysis for sector rotation prediction
- Regime-adaptive strategy switching (auto-select optimal params per market condition)
- Multi-strategy parallel execution with automatic best-strategy selection
MIT License - see LICENSE file for details