Annoy is a simple package to find approximate nearest neighbors (ANN) that I just put on Github. I’m not trying to compete with existing packages, but Annoy has a couple of features that makes it pretty useful.

Erik Bernhardsson

Annoy is an open source library for finding approximate nearest neighbors (ANN) with a focus on low memory usage. It stores its index in a contiguous memory-mapped file, allowing multiple processes to share the same index without duplicating memory. Spotify uses it to find similar tracks among millions of songs in 40-dimensional vector space, where sharing memory across processes is critical given the large data sizes involved.