Project Eagle

A highly versatile, adaptable, trainable, efficient, and intelligent AI-powered trading bot built in Python.

Architecture

eagle/
├── core/           Event bus, configuration, engine orchestration
├── data/           Async multi-provider data fetching and caching
├── indicators/     150+ vectorized technical indicators
├── features/       Feature engineering pipeline (100+ features)
├── models/         ML models (LSTM, Transformer, XGBoost, Ensemble)
├── strategies/     Plugin-based trading strategies
├── risk/           Risk management, position sizing, VaR
├── execution/      Order management and broker adapters
├── backtest/       Walk-forward backtesting engine
├── dashboard/      Real-time web monitoring (coming soon)
├── utils/          Logging, alerts, and helpers
└── cli.py          Command line interface

Key Features

Intelligence

• Bidirectional LSTM with attention mechanism
• Transformer encoder with positional encoding
• XGBoost gradient boosted trees
• Weighted voting, stacking, and majority vote ensembles
• Automatic retraining with walk-forward optimization

Strategies

• Momentum / trend-following
• Mean reversion with z-score detection
• ML-driven signal generation
• Plugin architecture for custom strategies

Risk Management

• ATR-based dynamic position sizing
• Value at Risk (VaR) and Conditional VaR
• Maximum drawdown limits with auto-halt
• Daily loss limits with cooldown
• Correlation-based exposure controls
• Trailing stop losses

Execution

• Paper trading with slippage/commission simulation
• Alpaca Markets integration (paper and live)
• Binance crypto exchange support
• Order type support (market, limit, stop)

Data

• Async multi-provider data fetching (Yahoo, Binance, Alpaca, Alpha Vantage)
• In-memory LRU cache with TTL
• Automatic failover between providers

Quick Start

Installation

# Clone the repository
git clone https://github.com/your-org/eagle.git
cd eagle

# Install in development mode
pip install -e ".[dev]"

Configuration

# Copy default config
cp config/default.yaml config/config.yaml

# Edit with your API keys and preferences
# You can also set environment variables:
# EAGLE__DATA__PROVIDERS__ALPACA__API_KEY=your_key

Usage

# Run in paper trading mode
eagle run

# Run backtesting
eagle backtest --start 2023-01-01 --end 2024-12-31

# Train ML models
eagle train --model all

# Train specific model
eagle train --model lstm

# Run with custom config
eagle run --config path/to/config.yaml --mode paper

Python API

import asyncio
from eagle.core.engine import EagleEngine
from eagle.core.config import load_config

async def main():
    config = load_config("config/config.yaml")
    engine = EagleEngine(config=config)
    await engine.run()

asyncio.run(main())

Configuration

All parameters are configurable via config/default.yaml. Key sections:

Section	Description
app	Mode (paper/live/backtest), logging level
data	Data providers, caching, history length
universe	Asset watchlist and timeframes
features	Technical indicator parameters
models	ML model hyperparameters
strategies	Active strategies and their parameters
risk	Risk limits, position sizing, VaR
execution	Broker config, order types, slippage
backtest	Date range, walk-forward settings

Environment variables override YAML config using the format:

EAGLE__SECTION__KEY=value

Models

Model	Type	Best For
LSTM	Deep Learning	Sequential patterns, regime detection
Transformer	Deep Learning	Long-range dependencies, attention patterns
XGBoost	Gradient Boosting	Tabular features, fast inference
Ensemble	Meta-learner	Combining all models for robust signals

Project Structure

Eagle/
├── config/
│   └── default.yaml            Default configuration
├── eagle/
│   ├── __init__.py
│   ├── cli.py                  Command line interface
│   ├── core/
│   │   ├── config.py           Pydantic configuration
│   │   ├── engine.py           Main engine orchestrator
│   │   └── events.py           Async event bus
│   ├── data/
│   │   └── fetcher.py          Multi-provider data fetching
│   ├── indicators/
│   │   └── technical.py        Technical indicator library
│   ├── features/
│   │   └── engineer.py         Feature engineering pipeline
│   ├── models/
│   │   ├── base_model.py       Base model interface
│   │   ├── lstm.py             BiLSTM with attention
│   │   ├── transformer.py      Transformer encoder
│   │   ├── ensemble.py         XGBoost and ensemble
│   │   └── manager.py          Model lifecycle manager
│   ├── strategies/
│   │   └── manager.py          Strategy framework
│   ├── risk/
│   │   ├── manager.py          Risk management engine
│   │   └── portfolio.py        Portfolio tracker
│   ├── execution/
│   │   └── engine.py           Order execution engine
│   ├── backtest/
│   │   └── engine.py           Backtesting engine
│   └── utils/
│       └── logger.py           Structured logging
├── pyproject.toml
└── README.md

License

MIT