The next step after AB-412 is to add the computed similarity metrics to Annoy indices. With building annoy indices, we should also add a cli to query the indices for similarity. In order to do this, we will need to take the following steps:

1. Develop a class to wrap the annoy index functionality in the specific context of using recordings as items.
2. Add build, save, and load functionality to the index class.
3. Add a utility to delete a saved index.
4. Add functionality to add items to the index, both by recording MBID and by the lowlevel.id in the database. Note that this will also require tracking empty rows in the database, and inserting arrays of the form [0, ..., 0] in their place in the index
5. Add functionality to the index class to query for items by lowlevel.id (or id tracked int the index) and also by MBID.
6. Add cli script to query index for similar recordings.