Struct MultiVectorBuilder 

pub struct MultiVectorBuilder { /* private fields */ }

Builds a multi-vector index from per-document token embeddings.

Implementations

impl MultiVectorBuilder

pub fn new(dim: usize) -> Self

Create a builder for the given embedding dimension.

pub fn set_m(&mut self, m: usize) -> &mut Self

Set the HNSW M parameter.

pub fn set_ef_construction(&mut self, ef: usize) -> &mut Self

Set the HNSW efConstruction parameter.

pub fn insert( &mut self, doc_id: u32, embeddings: Array2<f32>, metadata: HashMap<String, Value>, ) -> &mut Self

Insert a document’s token embeddings.

embeddings has shape [num_tokens, dim].

pub fn build(&self, index_path: &Path) -> Result<()>

Build the index and write it to index_path.

Produces:

• <index_path>.index — HNSW binary index
• <index_path>.labels.json — per-node token labels
• <index_path>.emb.npy — raw embedding matrix for exact reranking