What is the HNSW index in pgvector and how do you create and tune it?

-- Create an HNSW index (choose operator class to match your distance metric)

-- For L2 (Euclidean) distance <->:
CREATE INDEX ON items USING hnsw (embedding vector_l2_ops);

-- For cosine distance <=>:
CREATE INDEX ON items USING hnsw (embedding vector_cosine_ops);

-- For inner product <#>:
CREATE INDEX ON items USING hnsw (embedding vector_ip_ops);

-- For L1 distance <+>:
CREATE INDEX ON items USING hnsw (embedding vector_l1_ops);

-- HNSW with custom build parameters:
CREATE INDEX ON items USING hnsw (embedding vector_cosine_ops)
WITH (
    m = 16,             -- max connections per node per layer (default 16)
    ef_construction = 64  -- candidate list size during build (default 64)
);

-- Tune QUERY performance at query time:
SET hnsw.ef_search = 100;  -- larger = better recall, slower query (default 40)
SELECT id FROM items ORDER BY embedding <=> '[...]' LIMIT 5;
RESET hnsw.ef_search;

-- Speed up index BUILD with parallel workers:
SET max_parallel_maintenance_workers = 7;  -- match your CPU cores
SET maintenance_work_mem = '2GB';
CREATE INDEX ON items USING hnsw (embedding vector_cosine_ops);