Class VectorStore

The VectorStore class provides a high-level abstraction for storing and retrieving documents. It mainly consists of two modules - embedding model and vector store.

It supports embedding text using AI and interfacing with pluggable vector store backends such as FAISS or ChromaDB. This class handles initialization, insertion, similarity-based retrieval, and cleanup.

Typical usage involves:

Initializing the store via initialize()
Inserting documents via insert()
Querying similar documents via retrieve()

The embedding model and vector store are defined dynamically within the provided runtime.

Index

Constructors

constructor

Methods

clear define delete embedding insert retrieve

Constructors

constructor

new VectorStore(runtime: Runtime): VectorStore
Parameters
- runtime: Runtime
Returns VectorStore
- Defined in vector_store.ts:59

Methods

clear

clear(): Promise<void>
Returns Promise<void>
- Defined in vector_store.ts:158

define

define(
    embeddingModelName: "BAAI/bge-m3",
    embeddingModelAttrs: Record<string, any>,
    vectorStoreName: VectorStoreName,
    vectorStoreAttrs: Record<string, any>,
): Promise<void>
Defines the embedding model and vector store components to the runtime. This must be called before using any other method in the class. If already defined, this is a no-op.
Parameters
- embeddingModelName: "BAAI/bge-m3"
- embeddingModelAttrs: Record<string, any>
- vectorStoreName: VectorStoreName
- vectorStoreAttrs: Record<string, any>
Returns Promise<void>
- Defined in vector_store.ts:73

delete

delete(): Promise<void>
Delete resources from the runtime. This should be called when the VectorStore is no longer needed. If already deleted, this is a no-op.

Returns Promise<void>
- Defined in vector_store.ts:121

embedding

embedding(text: string): Promise<NDArray>
Generates an embedding vector for the given input text using the embedding model
Parameters
- text: string
  Input text to embed
Returns Promise<NDArray>
- Defined in vector_store.ts:163

insert

insert(item: VectorStoreInsertItem): Promise<void>
Inserts a new document into the vector store
Parameters
- item: VectorStoreInsertItem
Returns Promise<void>
- Defined in vector_store.ts:129

retrieve

retrieve(query: string, topK?: number): Promise<VectorStoreRetrieveItem[]>
Retrieves the top-K most similar documents to the given query
Parameters
- query: string
  The input query string to search for similar content
- topK: number = 5
  Number of top similar documents to retrieve
Returns Promise<VectorStoreRetrieveItem[]>
- Defined in vector_store.ts:139

Class VectorStore

Index

Constructors

Methods

Constructors

constructor

Parameters

Returns VectorStore

Methods

clear

Returns Promise<void>

define

Parameters

Returns Promise<void>

delete

Returns Promise<void>

embedding

Parameters

Returns Promise<NDArray>

insert

Parameters

Returns Promise<void>

retrieve

Parameters

Returns Promise<VectorStoreRetrieveItem[]>

Settings

On This Page