This project provides an AI-driven automated trading system for Binance Futures, designed for developers, researchers, and power users interested in exploring quantitative trading strategies. It enables multi-AI model competition (DeepSeek vs. Qwen), features a self-learning mechanism for strategy optimization, and offers a real-time, Binance-style dashboard for monitoring performance and decisions. The system aims to offer a professional platform for experimenting with AI in trading, with a strong emphasis on detailed logging and performance analysis.

How It Works

The system employs a dual-architecture approach with a Go backend handling trading logic, market analysis, AI communication, and API interactions, while a React frontend provides a professional, Binance-style user interface. Core functionality revolves around its multi-AI competition mode, allowing different AI models (DeepSeek, Qwen) to trade independently and their performance to be compared in real-time. A key differentiator is the AI Self-Learning Mechanism, which analyzes historical trading feedback (win rates, P/L ratios, best/worst performing coins) to dynamically adjust trading strategies and avoid repeating mistakes. Market analysis incorporates technical indicators (EMA, MACD, RSI, ATR) across different timeframes and open interest data, coupled with robust risk controls like per-coin position limits and fixed leverage.

Quick Start & Requirements

• Primary Install/Run:
  • Backend: go mod download (dependencies), go build -o nofx, ./nofx
  • Frontend: cd web, npm install, npm run dev
  • Access dashboard at http://localhost:3000.
• Prerequisites:
  • Go 1.21+
  • Node.js 18+
  • TA-Lib library (installation instructions provided for macOS/Ubuntu).
  • Binance Futures API keys (with Futures permission enabled).
  • AI API keys (DeepSeek: https://platform.deepseek.com, Qwen: https://dashscope.aliyuncs.com).
• Setup: Requires obtaining API keys and configuring config.json. Estimated setup time involves dependency installation and key configuration.

Highlighted Details

• Multi-AI Competition Mode: Enables live trading battles between different AI models (e.g., Qwen vs. DeepSeek) with independent account management and real-time performance comparison charts.
• AI Self-Learning Mechanism: Analyzes historical trading performance (last 20 cycles) to optimize decisions, identify successful patterns, and dynamically adjust trading style.
• Accurate PnL & Duration Tracking (v2.0.2): Calculates actual USDT profit (Position Value × Price Change % × Leverage) and tracks position holding duration for improved AI decision-making.
• Professional UI: Features a Binance-style interface with real-time data refresh (5 seconds), equity curve charts, multi-AI ROI comparison, and detailed decision logging with AI's Chain of Thought (CoT).

Maintenance & Community

• Community: A Telegram developer community is available for discussion and support.
• Contact: Reachable via Twitter/X at @Web3Tinkle.
• Contribution: The project welcomes issues and pull requests.

Licensing & Compatibility

• License: MIT License.
• Compatibility: The MIT license generally permits commercial use and integration with closed-source projects. However, the project is explicitly marked as experimental and carries significant risks.

Limitations & Caveats

The system is experimental and carries significant risks associated with AI auto-trading; it is strongly recommended for learning/research or testing with small amounts only. While initially mentioning multiple exchanges, the detailed setup and architecture primarily focus on Binance Futures. Users must manage API keys for both Binance and AI providers, and the setup involves several technical dependencies (Go, Node.js, TA-Lib).