I should’ve said that we were already using it for actual search where embeddings similarity is just one component of the overall score. For pure vector stuff a

I should’ve said that we were already using it for actual search where embeddings similarity is just one component of the overall score. For pure vector stuff a dedicated solution will be faster.