Skip to content

Latest commit

 

History

History
125 lines (92 loc) · 3.33 KB

LIBRARY_ORGANIZATION.md

File metadata and controls

125 lines (92 loc) · 3.33 KB

WinstonAI Library - Organization Summary

What Was Done

This update reorganizes the WinstonAI project into a proper Python library structure with clear separation of concerns and easy-to-use APIs.

Major Changes

  1. Created winston_ai/ Package Structure

    • Organized code into logical submodules
    • Added proper __init__.py files for each module
    • Implemented clean import system

  2. Modular Architecture

    winston_ai/
├── models/          # Neural network architectures
├── training/        # Training utilities and agents
├── trading/         # Live trading interface
├── indicators/      # Technical analysis
└── utils/          # Helper functions

  3. Example Scripts

    • examples/quickstart.py - Quick start guide
    • examples/train_model.py - Full training example
    • examples/use_model.py - Inference example

  4. Reorganized Files

    • Config files moved to data/configs/
    • Model checkpoints go to models/ (gitignored)
    • Legacy scripts remain in src/ for reference

How to Use the Library

Installation

pip install -e .

Quick Start

from winston_ai import Trainer, Config
import pandas as pd

# Load your data
data = pd.read_csv('market_data.csv')

# Train model
config = Config()
trainer = Trainer(data=data, config=config)
metrics = trainer.train(episodes=1000)

Using Trained Model

from winston_ai import LiveTrader

trader = LiveTrader('models/winston_ai_final.pth')
prediction = trader.predict(recent_data)
print(f"Action: {prediction['action_name']}")

Module Descriptions

winston_ai.models

  • WinstonAI - Standard DQN model with LSTM
  • AdvancedWinstonAI - Large GPU-optimized model with attention
  • MultiHeadAttention - Transformer attention mechanism

winston_ai.training

  • Trainer - High-level training orchestration
  • DQNAgent - Deep Q-Network agent with experience replay
  • BinaryOptionsEnvironment - Trading simulation environment

winston_ai.trading

  • LiveTrader - Interface for using trained models in production

winston_ai.indicators

  • TechnicalIndicators - 50+ technical indicators calculator

winston_ai.utils

  • Config - Configuration management
  • get_device(), setup_gpu() - GPU utilities
  • save_checkpoint(), load_checkpoint() - Model persistence

Benefits

  1. Better Organization - Clear separation of concerns
  2. Easy to Use - Simple, intuitive API
  3. Reusable - Import only what you need
  4. Maintainable - Modular structure makes updates easier
  5. Extensible - Easy to add new features
  6. Documented - Examples and documentation included

Backward Compatibility

  • Original scripts in src/ still work
  • Can be used alongside the new library
  • Gradual migration path available

Testing

Library has been tested and verified:

  • ✅ All imports work correctly
  • ✅ Configuration system functional
  • ✅ Training pipeline operational
  • ✅ Model architectures validated

Next Steps

  1. Train models using the new API
  2. Integrate with your trading platform
  3. Extend with custom strategies
  4. Add more technical indicators
  5. Implement backtesting framework

Support

  • See examples/README.md for detailed usage
  • Check README.md for updated documentation
  • Original scripts remain in src/ for reference