4 ms·

I guess I'm too simple to understand why this is useful? Just because it's been implemented in so few lines or?

• cluckindan 1y ago
  It makes it possible to approximate k-nearest neighbor vector search without having to calculate cosine similarity, dot product or euclidean distance for all stored vectors on the fly.