Module search 

Hybrid search with semantic and lexical retrieval.

Combines vector similarity search with FTS5 BM25 using Reciprocal Rank Fusion (RRF).

Features

• Semantic Search: Vector similarity using embeddings
• BM25 Search: Full-text search using SQLite FTS5
• Hybrid Search: Combines both using Reciprocal Rank Fusion
• HNSW Index: Optional scalable approximate nearest neighbor search (requires usearch-hnsw feature)

Re-exports

pub use hnsw::HnswConfig;

pub use hnsw::HnswIndex;

pub use hnsw::HnswResult;

Modules

hnsw

HNSW (Hierarchical Navigable Small World) vector index.

Structs

EmbeddingModelInfo

Information about embedding model versions for a buffer.

IncrementalEmbedResult

Result of an incremental embedding operation.

RrfConfig

Configuration for RRF algorithm.

SearchConfig

Configuration for hybrid search.

SearchResult

Search result with chunk ID and combined score.

Constants

DEFAULT_PREVIEW_LEN

Default preview length in characters.

DEFAULT_SIMILARITY_THRESHOLD

Default similarity threshold for semantic search.

DEFAULT_TOP_K

Default number of results to return.

Functions

buffer_fully_embedded

Checks if a buffer has all chunks embedded.

check_model_mismatch

Checks if existing embeddings were created with a different model.

embed_buffer_chunks

Generates and stores embeddings for all chunks in a buffer.

embed_buffer_chunks_incremental

Incrementally embeds chunks in a buffer.

get_embedding_model_info

Gets embedding model information for a buffer.

hybrid_search

Performs hybrid search combining semantic and BM25 results.

populate_previews

Populates content previews for search results.

reciprocal_rank_fusion

Performs Reciprocal Rank Fusion on multiple ranked lists.

search_bm25

Performs BM25-only search.

search_semantic

Performs semantic-only search.

weighted_rrf

Performs weighted RRF where each list has a weight.