Struct MultiVectorRetriever 

pub struct MultiVectorRetriever<S: VectorStore> { /* private fields */ }

A retriever that maps multiple child vectors back to parent documents.

Each parent document can have multiple sub-documents (e.g., summaries, smaller chunks) stored in the vector store. Retrieval finds relevant sub-documents, then returns the original parent documents.

Implementations

impl<S: VectorStore + 'static> MultiVectorRetriever<S>

pub fn new( vectorstore: Arc<S>, embeddings: Arc<dyn Embeddings>, k: usize, ) -> Self

Create a new MultiVectorRetriever.

• vectorstore: the vector store to search for child documents.
• embeddings: the embeddings provider for embedding child documents.
• k: the number of child documents to retrieve for parent lookup.

pub fn with_id_key(self, key: impl Into<String>) -> Self

Set a custom metadata key linking child documents to their parent ID. Defaults to "parent_id".

pub async fn add_documents( &self, parent_docs: Vec<Document>, child_docs: Vec<Document>, ) -> Result<(), SynapticError>

Add parent documents and their associated child documents.

Parents are stored in the internal docstore. Children are embedded and added to the vector store. Each child document must have the id_key metadata field set to the parent document’s ID.

Trait Implementations

impl<S: VectorStore + 'static> Retriever for MultiVectorRetriever<S>

fn retrieve<'life0, 'life1, 'async_trait>( &'life0 self, query: &'life1 str, top_k: usize, ) -> Pin<Box<dyn Future<Output = Result<Vec<Document>, SynapticError>> + Send + 'async_trait>>

where Self: 'async_trait, 'life0: 'async_trait, 'life1: 'async_trait,