BuildML

BuildML is a Python machine-learning library. One object holds your data, the train / validation / test split, preprocessing, the model, and the history of what you ran. That object is buildml.Session. The core path is classification and regression. The same object also runs forecasting, AutoML, fairness, recommenders, RAG, graphs, NLP, Torch, and the other domains in this repo. If you try to prepare or fit before a split, it stops you.

A Session is a unified, stateful ML lifecycle. Enforced leakage safeguards, fold-local preprocessing, contextual teaching, workflow guidance, checkpointing, and auditable export live on that same object.

Install with pip install buildml (Python 3.10 through 3.13). This repo is BuildML 2.6.1, the current stable Session line. PyPI serves 2.6.1.

Start here

  1. Install and check the import: Installation

  2. Run a first Session: A first Session

  3. Learn the few ideas that matter: Concept guide

  4. Use the order as a decision path: Workflow guide

After that, open the Guide index when you need a domain tutorial, or Current capabilities when you need a map of what is shipped. The Markdown files under guides/ are the source for those tutorials. Sphinx includes them so this site and GitHub stay on the same text.

Legacy 1.x remains available under pip install "buildml==1.0.9" and as buildml/_legacy/ for reference. New work uses Session. See BuildML 1.x archive.

Guides and reference

Indices and tables