This tree shows similarity of the latest container of each public collection, and unlike the live version in Singularity Hub, this version was generated using content hashes for a more robust metric. The other version (quick, generated on a nightly basis) has a metric that is based on the number of common folder paths between the containers, so containers with similar software will be shown closer together in the tree. We use hierarchical clustering from scipy to generate clusters, and d3.js to render the tree.