n=10,000: exact 0.067 ms/query (recall 1.000), HNSW build 0.7 s
   HNSW ef=10   recall@10 0.847  0.044 ms/query
   HNSW ef=40   recall@10 0.976  0.108 ms/query
   HNSW ef=160  recall@10 0.999  0.313 ms/query
n=50,000: exact 0.297 ms/query (recall 1.000), HNSW build 5.7 s
   HNSW ef=10   recall@10 0.833  0.062 ms/query
   HNSW ef=40   recall@10 0.962  0.157 ms/query
   HNSW ef=160  recall@10 0.996  0.477 ms/query
n=133,957: exact 0.805 ms/query (recall 1.000), HNSW build 19.1 s
   HNSW ef=10   recall@10 0.862  0.069 ms/query
   HNSW ef=40   recall@10 0.972  0.167 ms/query
   HNSW ef=160  recall@10 0.994  0.495 ms/query
   greedy walk, top-1 correct: ef=1: 0.460  ef=2: 0.630  ef=4: 0.754  ef=8: 0.880  ef=16: 0.944  ef=32: 0.972
Python 3.12.13 | numpy 2.5.4 | hnswlib 0.8.0 | sentence-transformers 6.1.0