scikit-lite 0.0.1

Lightweight machine learning library with scikit-learn compatible API

scikit-lite

⚠️ PRE-ALPHA - This package is in very early development. APIs will change.

A simple machine learning library built for educational purposes.

Installation

For users:

pip install scikit-lite

Development Installation

Prerequisites

• Python >= 3.11
• Rust and Cargo (install from https://rustup.rs)
• uv (install with pip install uv)

Setup

1. Clone the repository:

   git clone https://github.com/kowanietz/scikit-lite.git
   cd scikit-lite
   

2. Install dependencies and build Rust extensions:

   uv sync --extra dev
   uv run maturin develop
   

3. Verify installation:

   python -c "import sklite; print(sklite.rust_health_check())"
   

Development Workflow

Rebuilding After Changes

When you modify Rust code in src/:

# Quick rebuild (debug mode)
uv run maturin develop

# Optimized rebuild (release mode, slower build but faster runtime)
uv run maturin develop --release

Code Quality

# Install pre-commit hooks (one-time setup)
uv run pre-commit install

# Run all checks manually
uv run pre-commit run --all-files

Publishing to PyPI

Prerequisites

1. Ensure versions match in both:

   • pyproject.toml: version = "x.y.z"
   • Cargo.toml: version = "x.y.z"

2. Set PyPI token:

   export MATURIN_PYPI_TOKEN="your-pypi-token"
   

Publish

# Build and publish in one command
uv run maturin publish --release

Maturin will automatically:

• Build optimized wheels for your platform
• Upload to PyPI

TODO: Migrate release workflow to GitHub Actions for multiplatform builds.

Contributing

This project is in early development. Contributions are welcome but please note the API is unstable.

License

MIT License - See LICENSE file for details.

Acknowledgments

Inspired by scikit-learn's excellent API design and educational resources.