There's a few different ideas that I'd like to explore to utilize compairr more in our AIRR Knowledge project. Some of these ideas would utilize compairr in a long-running program scenario, versus discrete execution at the command line.
One example use case is to load up a large data set as a "reference" then perform "queries" against that with single or small sets of rearrangements. Those queries might return rearrangements in the reference based upon similarity and/or assign them to a cluster.