Support for ANN Query Planning with HNSW Indexes #3848
Description
Now that HNSW-backed vector indexes are implemented and functional, we should add native support for planning Approximate Nearest Neighbor (ANN) queries in SQL, here is an example for performing efficient K-nearest neighbor searches through SQL:
SELECT docId, title, euclidean_distance(embedding, CAST([1.0, 0.0, 0.0] AS VECTOR(3, HALF))) AS distance
FROM documents
WHERE zone = 'zone1' AND bookshelf = 'fiction'
AND ROW_NUMBER() OVER (
PARTITION BY zone, bookshelf
ORDER BY euclidean_distance(embedding, CAST([1.0, 0.0, 0.0] AS VECTOR(3, HALF))) ASC
) <= 3