In order to implement annoy indices to query for similarity in AcousticBrainz, we first need to compute metrics that will measure similarity for each recording. We can do this with the following changes: 

• Make migrations to include similarity table, metrics table, and similarity_stats table - similarity table will hold columns with the computed metric for each id in the lowlevel table.  Metrics table will hold metadata about each metric, and similarity_stats table will include the mean and standard deviation of each computed metric on the entire set of recordings.
• Add a module containing the metric classes.
• Use a cli script to compute the metrics incrementally on the entire database.