Chemfp 4.0 was recently released, with support for several diversity selection algorithms, and an improved API for interactive use in a notebook environment.

Chemfp is an analytics package for cheminformatics fingerprints. It contains command-line tools and an extensive Python library for fingerprint generation, high-performance similarity search, diversity selection, and exploratory research.

The new diversity selection algorithms are MaxMin, sphere exclusion (both random and directed), and HeapSweep.

People who live in the Jupyter notebook will likely enjoy the new chemfp user experience. Most long-term actions support progress bars, chemfp's Python objects have more informative repr()s, search results added Pandas integration, and there are new high-level APIs that let you express a lot of functionality compactly.

The Base License covers most in-house use of chemfp, though a few features are either limited or disabled and require a license key to unlock. For alternative licenses, including source code and no-cost academic licensing, see https://chemfp.com/license/ -- or try one of the re-formatted ChEMBL datasets at https://chemfp.com/datasets/ which include an embedded authorization key.