Approximate nearest neighbors search

For the fastest vector search

Relevance's vector search is default brute-force vector search, this is great for testing different configurations of vector search. Answering questions such as "Which vectors or combinations of vectors to use?", "Does traditional search improve the results?", etc.

For searching on billions of vectors though you'll need approximate vector search. We support many different vector databases as a backend to use approximate vector search.

Approximate nearest neighbors search is only available throught sales at the moment, contact us via https://relevance.ai to understand more about it.