Struct HybridIndex 

pub struct HybridIndex {
    pub semantic: SearchIndex,
    /* private fields */
}

Combined semantic + keyword search index with RRF fusion.

Build once from chunks and pre-computed embeddings; query repeatedly via search.

Fields

semantic: SearchIndex

Semantic (dense vector) search index.

Implementations

impl HybridIndex

pub fn new( chunks: Vec<CodeChunk>, embeddings: &[Vec<f32>], cascade_dim: Option<usize>, ) -> Result<Self>

Build a HybridIndex from raw chunks and their pre-computed embeddings.

Constructs both the SearchIndex and Bm25Index in one call. cascade_dim is forwarded to SearchIndex::new for optional MRL cascade pre-filtering.

Errors

Returns an error if the BM25 index cannot be built (e.g., tantivy schema or writer failure).

pub fn from_parts(semantic: SearchIndex, bm25: Bm25Index) -> Self

Assemble a HybridIndex from pre-built components.

Useful when the caller has already constructed the sub-indices separately (e.g., loaded from a cache).

pub fn search( &self, query_embedding: &[f32], query_text: &str, top_k: usize, threshold: f32, mode: SearchMode, ) -> Vec<(usize, f32)>

Search the index and return (chunk_index, score) pairs.

Dispatches based on mode:

• SearchMode::Semantic — pure dense vector search via SearchIndex::rank.
• SearchMode::Keyword — pure BM25 keyword search, truncated to top_k.
• SearchMode::Hybrid — retrieves both ranked lists, fuses them with rrf_fuse, then truncates to top_k.

Scores are min-max normalized to [0, 1] regardless of mode, so a threshold of 0.5 always means “above midpoint of the score range” whether the underlying scores are cosine similarity, BM25, or RRF.

pub fn chunks(&self) -> &[CodeChunk]

All chunks in the index.