Crate ruvector_pq_search 

Product Quantization with Asymmetric Distance Computation (PQ-ADC)

Decomposes D-dimensional vectors into M sub-vectors of d=D/M dimensions, trains K centroids per sub-space with Lloyd’s algorithm, and encodes each vector as M bytes (K=256 → 8 bits per sub-space).

Three search strategies:

• FlatPqIndex — linear ADC scan over all PQ codes (baseline)
• IvfPqIndex — coarse IVF + PQ per cell (alternative A)
• ResidualPqIndex — PQ + f32 residual correction for top-k (alternative B)

All implement the PqSearch trait; no external service dependency.

Re-exports

pub use codebook::PqCodebook;

pub use codebook::PqConfig;

pub use encoder::decode_vector;

pub use encoder::encode_vector;

pub use flat::FlatPqIndex;

pub use ivf_pq::IvfPqIndex;

pub use residual::ResidualPqIndex;

Modules

codebook

PQ codebook: M sub-spaces, K centroids each, trained with Lloyd’s k-means.

encoder

PQ encode/decode utilities.

flat

Flat PQ Index: linear ADC scan over all encoded vectors (baseline variant).

ivf_pq

IVF+PQ Index: coarse IVF clustering, then PQ within each inverted list.

residual

Residual-Corrected PQ Index: PQ scan + exact L2 re-score for top-k candidates.

Structs

ExactSearch

Brute-force exact L2 search; used for recall ground-truth only.

SearchResult

Single search result returned by all variants.

Traits

PqSearch

Unified trait for all PQ-based search backends.

Functions

recall_at_k

Compute recall@k: fraction of true top-k found in approximate top-k.