Installation

Prerequisites

• Python 3.12 or higher

Installation Profiles

Pick an installation profile based on how you want to get started.

1. Core (minimal dependencies)

pipuv

pip install nonconform

uv add nonconform

This includes NumPy, SciPy, and scikit-learn.

2. Anomaly-ready (recommended for most users)

pipuv

pip install "nonconform[pyod,data]"

uv add "nonconform[pyod,data]"

This adds: - PyOD detector zoo ([pyod]) - oddball benchmark datasets ([data])

3. Full installation

pipuv

pip install "nonconform[all]"

uv add "nonconform[all]"

Optional Dependencies

nonconform offers optional extras for specific use cases:

Extra	What it adds	Install when you need
[pyod]	PyOD library	Access to 40+ anomaly detection algorithms (Isolation Forest, LOF, KNN, etc.)
[data]	oddball + PyArrow	Benchmark datasets for experimentation and testing
[fdr]	online-fdr	Streaming/online FDR control for real-time applications
[probabilistic]	KDEpy + Optuna	KDE-based probabilistic p-values and optional hyperparameter tuning
[all]	All of the above	Full functionality

Which Extras Do You Need?

• Add [pyod] if you want a larger set of anomaly detectors.
• Add [data] if you want oddball benchmark datasets.
• Add [probabilistic] if you use Probabilistic() estimation and KDE tuning.

• Add [fdr] if you need:

• Real-time anomaly detection with streaming FDR control

• Sequential testing over time

Verify Installation

import nonconform
print(nonconform.__version__)

Next Steps

Head to the Quick Start to see nonconform in action.