Usage & Hybrid Search

Index management, Top-K querying, prefiltering, hybrid search with vchord, and Block-WeakAnd internals.

Learn how to build, tune, and query vchord_bm25 indexes in production:

Create bm25 indexes, tune segment sizes and BM25 parameters (k1, b), monitor parallel builds, and inspect pages.

Run Top-K Block-WeakAnd queries, understand negative BM25 scores, and tune bm25_limit.

Combine WHERE predicates with BM25 ranking using post-filtering, pre-filtering, and partial indexes.

Fuse sparse BM25 lexical search with VectorChord dense vector search via Reciprocal Rank Fusion (RRF) and ColBERT.

Deep dive into the Okapi BM25 formula, 1-byte fieldnorm quantization, delta-bitpacking, and Block-Max WAND.