Installation

From PyPI

pip install pippen

Optional extras install only what you need:

Extra	Installs	For
sources	nba_api, pbpstats	Downloading raw data yourself
fit	NumPyro, LightGBM, Optuna, MLflow, SHAP	Fitting models rather than using published ones
api	FastAPI, Uvicorn	Running the inference service
dashboard	Streamlit, Plotly	Running the dashboard locally

pip install "pippen[sources,fit]"

From source

This project uses uv, which manages the virtual environment and the lockfile together.

git clone https://github.com/AlphaNerdFx/pippen
cd pippen
uv sync --extra dev
uv run pippen --help

Why uv rather than pip and venv

uv sync reads uv.lock, which pins every transitive dependency to an exact version. Two people running it on different machines get byte-identical environments. That is what makes a result reproducible.

If you would rather not use uv, requirements.txt is generated from the same lockfile:

python -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt
pip install -e .

Supported versions

Python 3.10 through 3.13, tested on every push. Linux and macOS are tested; Windows should work but is not covered by continuous integration.