Installation

From PyPI

pip install superglm

The normal install includes pandas, and model-data entry points accept either a pandas.DataFrame or an eager Polars DataFrame. Polars is an optional input backend and is installed separately:

pip install polars

Pass eager frames to SuperGLM. A polars.LazyFrame must be materialized by the caller with LazyFrame.collect() before fitting or prediction. SuperGLM keeps the input backend native while compiling features; it does not convert a whole Polars frame to pandas. Outputs remain pandas DataFrames where the existing reporting, inference, plotting-data, and export APIs are table-oriented.

With optional dependencies

# Interactive Plotly charts
pip install "superglm[plotting]"

# Benchmarking (glum, statsmodels, pyarrow)
pip install "superglm[bench]"

The local model editor and its FastAPI/Uvicorn server are included in the normal installation.

Unreleased development version

pip install "superglm @ git+https://github.com/StrudelDoodleS/superglm.git"

Development install

git clone https://github.com/StrudelDoodleS/superglm.git
cd superglm
uv sync --extra dev --extra bench --extra plotting