GlassBox

A transparent, white-box AutoML library built from scratch with NumPy.

GlassBox is a lightweight AutoML library designed for auditability and clarity. Every algorithm is implemented from first principles using only NumPy — no hidden black-box dependencies.

---

Features

• Frame

  ---

  A minimal Dataset container with CSV I/O, column selection, row slicing, and in-place mutations — all backed by NumPy arrays.

  Frame Guide

• Inspector

  ---

  Non-destructive Exploratory Data Analysis: statistical profiling, auto-typing, outlier detection, and Pearson / Cramér's V associations.

  Inspector Guide

• Cleaner

  ---

  A scikit-learn-style fit / transform pipeline for imputation, outlier capping, scaling, and encoding.

  Cleaner Guide

• Models

  ---

  Decision Trees, Random Forests, and K-Nearest Neighbors — for both classification and regression.

  Models Guide

---

Quick Install

pip install .          # runtime only (numpy)
pip install .[dev]     # with dev tools (mkdocs, ruff, black)

---

Philosophy

Principle	Meaning
Clarity over cleverness	Every line should be readable.
Explicit over implicit	No hidden transformations.
NumPy is the engine	Zero heavy dependencies.
Auditability first	Built for white-box inspection.