Class InMemoryVectorStore

In-memory vector store implementation.

Uses cosine similarity for vector search and stores all documents in memory. Suitable for development, testing, and small-scale applications.

For production use with large datasets, consider using persistent vector stores like Pinecone, Weaviate, or Chroma.

Example

const store = new InMemoryVectorStore(llmProvider, 'text-embedding-3-small');

await store.add([
  { id: '1', content: 'AI agents are autonomous entities' },
  { id: '2', content: 'Machine learning powers modern AI' }
]);

const results = await store.search({
  query: 'What are AI agents?',
  limit: 3
});

Implements

  • VectorStore

Constructors

constructor

Methods

add search get update delete deleteMany count clear listAll size

Constructors

constructor

Methods

add

  • add(documents: Document[]): Promise<number>

    Add documents to the vector store

    Parameters

    • documents: Document[]

      Documents to add (embeddings will be generated if not provided)

    Returns Promise<number>

    Number of documents added

search

  • search(query: VectorSearchQuery): Promise<VectorSearchResult[]>

    Search for similar documents

    Parameters

    • query: VectorSearchQuery

      Search query

    Returns Promise<VectorSearchResult[]>

    Matching documents with similarity scores

get

  • get(id: string): Promise<null | Document>

    Get a document by ID

    Parameters

    • id: string

      Document ID

    Returns Promise<null | Document>

    The document or null if not found

update

  • update(id: string, document: Partial<Document>): Promise<boolean>

    Update a document

    Parameters

    • id: string

      Document ID

    • document: Partial<Document>

      Updated document

    Returns Promise<boolean>

    true if updated, false if not found

delete

  • delete(id: string): Promise<boolean>

    Delete a document

    Parameters

    • id: string

      Document ID

    Returns Promise<boolean>

    true if deleted, false if not found

deleteMany

  • deleteMany(filter: MetadataFilter): Promise<number>

    Delete multiple documents matching a filter

    Parameters

    • filter: MetadataFilter

      Metadata filter

    Returns Promise<number>

    Number of documents deleted

count

clear

listAll

  • listAll(filter?: MetadataFilter): Promise<Document[]>

    List all documents (optionally filtered)

    Parameters

    • Optionalfilter: MetadataFilter

      Optional metadata filter

    Returns Promise<Document[]>

    All matching documents

size

Settings

Member Visibility

On This Page

Constructors
Methods