Expand description
Vector index trait and types (zero dependencies). Vector index trait for compressed approximate nearest neighbor search.
Defines the abstract interface that concrete implementations (e.g.,
llm-kernel-vector-index with TurboQuant) must satisfy. This module has
zero external dependencies — implementations live in separate crates.
use llm_kernel::embedding::vector_index::SearchHit;
let hit = SearchHit { id: 42, score: 0.95 };
assert_eq!(hit.id, 42);Structs§
- Search
Hit - A single search hit from vector index lookup.
Enums§
- Fusion
- How to combine several ranked
SearchHitlists into one — the join point for hybrid retrieval (e.g. a dense branch plus a sparse/lexical branch).
Traits§
- Vector
Index - Trait for compressed vector indexes.