Code for the blog series on benchmarking nearest-neighbour libraries:
The target are libraries that expose a Python interface and allow (exact or approximate) k-NN search in high dimensional vector spaces.
Run run_all.sh
to download Wikipedia, index it and run all experiments.