⚠️ Experimental Project: This is research code from personal trading experiments. Not production-ready, not profitable. Use at your own risk.

A modular AI-powered trading framework supporting multiple exchanges and strategies. The system uses LLM-based decision-making with risk management and position tracking.

ai-trading/
├── core/           # Core abstractions and interfaces
├── adapters/       # Exchange connectors (Binance, Tiger, yFinance)
├── engine/         # AI strategy execution engine
├── strategies/     # Strategy implementations
├── services/       # Support services (Telegram, charts, AI client)
├── execution/      # Order execution and guards
├── exit_engine/    # Exit rules and position management
└── filters/        # Pre-trade filters

• interfaces.py: Base contracts for adapters, formatters, and strategies
• context.py: Trading context and state management
• position.py: Position tracking and execution history
• defaults.py: Default implementations and utilities

Exchange-specific connectors implementing unified interface:

• binance_perp.py: Binance USDT-M perpetual futures
• binance_spot.py: Binance spot trading
• tiger_stocks.py: Tiger Brokers US stocks
• tiger_futures.py: Tiger Brokers futures
• yfinance_stocks.py: YFinance data adapter (read-only)

• strategy.py: Base AI strategy with LLM integration
• pipeline.py: Decision-making pipeline orchestration

• binance_perp.py: AI-driven perpetual futures strategy with technical analysis
• common/decision_utils.py: Decision payload normalization

• ai_client.py: LLM API client (OpenAI/Anthropic compatible)
• telegram.py: Telegram bot for monitoring and manual control
• chart.py: Chart generation with indicators
• position_manager.py: Multi-symbol position state management
• scheduler.py: Strategy execution scheduling
• state.py: Persistent state storage

• guards.py: Pre-execution risk checks and validations

• engine.py: Exit decision engine
• rules.py: Exit rule implementations (stop-loss, take-profit, trailing)
• models.py: Exit-related data models

• pre_filters.py: Pre-trade filtering (volatility, liquidity, etc.)

• 🤖 LLM-Based Decisions: Leverages Claude/GPT for market analysis
• 📊 Technical Analysis: Built-in indicators (EMA, RSI, MACD, ATR)
• 📈 Chart Generation: Automated chart creation for visual analysis
• 🎯 Risk Management: Position sizing, stop-loss, take-profit
• 🔔 Telegram Integration: Real-time alerts and manual control
• 💾 State Persistence: Position tracking across restarts
• 🔌 Multi-Exchange: Unified interface for different brokers

Set required environment variables:

# Exchange credentials
export BINANCE_USDM_API_KEY="your_key"
export BINANCE_USDM_API_SECRET="your_secret"

# Telegram (optional)
export TELEGRAM_BOT_TOKEN="your_token"
export TELEGRAM_CHAT_ID="your_chat_id"
export TELEGRAM_ADMIN_IDS="comma,separated,ids"

# AI provider (OpenAI or Anthropic)
export OPENAI_API_KEY="your_key"
# or
export ANTHROPIC_API_KEY="your_key"

from ai_trading.strategies.binance_perp import AIBinancePerpStrategy

config = {
    "exchange": "binance",
    "api_key_env": "BINANCE_USDM_API_KEY",
    "api_secret_env": "BINANCE_USDM_API_SECRET",
    "default_symbols": ["BTCUSDT", "ETHUSDT"],
    "ai": {
        "provider": "anthropic",
        "model": "claude-sonnet-4",
        "max_tokens": 4000
    },
    "constraints": {
        "long_only": False,
        "risk_preference": "medium"
    }
}

strategy = AIBinancePerpStrategy(config)
await strategy.run()

FOR EDUCATIONAL AND RESEARCH PURPOSES ONLY

• This code is experimental and unaudited
• Trading carries significant financial risk
• No guarantees of profitability
• Use at your own risk
• Author assumes no liability for losses

This is a personal research project. Feel free to fork and experiment, but no support is provided.