It's not so much for search (although it can obviously be used that way) as it is for just computing various kinds of similarity/dependency for arbitrary vectors of equal length.
To the extent you would want to use it for vector/semantic search applications, I would advise doing a first pass with something like FAISS to narrow down the list of potential matches from the entire universe of stored vectors, and THEN trying to compute these other measures of similarity/dependency only on the "most likely" vectors using my library. Hopefully that makes sense.
To the extent you would want to use it for vector/semantic search applications, I would advise doing a first pass with something like FAISS to narrow down the list of potential matches from the entire universe of stored vectors, and THEN trying to compute these other measures of similarity/dependency only on the "most likely" vectors using my library. Hopefully that makes sense.