Summary

In this chapter, we learned how Redis Stack can store and index vectors to perform similarity searches and recommend similar documents. The power of VSS is that it can be used to find similar matches from incomplete or uncorrelated data. It can also leverage innovative AI/ML models, different search algorithms such as FLAT or HNSW, which focus on the precision or speed of the search, and different distances so that the most suitable option can be configured concerning the entity described by the vector embedding. VSS has been used for recommendation engines, but there are additional and relevant use cases that are trending right now, such as question answering, where we take advantage of generative models and their ability to take a set of prompts and perform text completion from the results of VSS so that we can provide complete answers out of them.

Data classification is another use case: by pre-training the database with a set of vectors modeling known objects (labeled...