Add vector-stores module: one dataset through VectorStore on pgvector, Redis, Qdrant and Elasticsearch

Ingest time, recall@10 against exact search, latency, metadata filtering and running cost, with the defaults that cost recall reproduced and every figure written to output/.
2026-10-09 05:50:01 +00:00
parent 527f4ba7ff
commit 74e6334704
26 changed files with 1065 additions and 0 deletions
@@ -0,0 +1,7 @@
+pgvector 0.8.7: hnsw.ef_search against the filtered query (category == 'c3' && year >= 2022); the filter is applied AFTER the index scan
+ef_search  iterative_scan  mean hits    p50 ms  filt recall
+40         off                   2.0       1.1        0.181
+200        off                   2.0       1.7        0.199
+1000       off                  10.0      26.3        1.000
+40         relaxed_order        10.0       7.9        0.621
+40         strict_order         10.0       9.4        0.300