Module vector_store 

Vector-store insertion and similarity-search interfaces. VectorStoreIndex implementations also implement Tool when their filter is deserializable and uses JSON values.

use rig_core::vector_store::in_memory_store::InMemoryVectorStore;

let store = InMemoryVectorStore::<String>::from_documents([]);
assert!(store.is_empty());

Re-exports

pub use request::VectorSearchRequest;

Modules

in_memory_store

In-memory vector storage with cosine scoring. A search scores every document by default; InMemoryVectorStore::builder opts into scoring only approximate candidates (IndexStrategy::LSH) on large stores.

lsh

Random-projection hashing for approximate vector-search candidates.

request

Vector-search requests and composable metadata filters.

Structs

VectorSearchIdResult

One document ID returned by VectorStoreIndex::top_n_ids.

VectorSearchResult

One document returned by VectorStoreIndex::top_n.

Enums

IndexStrategy

How the in-memory vector store picks the documents a search scores.

VectorStoreError

Errors from vector store operations.

Traits

InsertDocuments

Trait for inserting documents and embeddings into a vector store.

VectorStoreIndex

Trait for querying a vector store by similarity.

Functions

flatten_embedded

Serializes each document to JSON once, then applies f to every (document, embedding) pair, flattening the results into a single vector.