Featly is a lightweight, extensible feature development framework designed to streamline the feature engineering process for machine learning projects.
Featly aims to simplify feature development and management while maintaining flexibility and extensibility. Our goal is to provide a minimalistic yet powerful framework that helps data scientists and ML engineers focus on creating valuable features rather than dealing with infrastructure complexities.
- Simple Feature Definition: Use Pydantic models for type-safe feature definitions with built-in validation
- Flexible Storage: Support for various storage backends to accommodate different use cases
- Version Control: Built-in feature versioning to track changes and ensure reproducibility
- Type Safety: Strong typing support with runtime validation
- Extensible Design: Easy to extend and customize for specific needs
- Minimal Dependencies: Keep dependencies minimal to avoid complexity
pip install -e .This project is currently in early development. We are actively working on:
- [] Core feature definition framework
- [] Basic storage implementations
- [] Feature computation pipeline
- [] Feature validation and testing utilities
Why Featly? While there are several feature stores available (like Feast, Hopsworks), Featly focuses on:
- Simplicity: Minimalistic API design that's easy to understand and use
- Flexibility: Easy to extend and adapt to specific needs
- Developer Experience: Strong typing and validation out of the box
- Learning: Great for understanding feature store concepts and building custom solutions