In-Memory Vector Store

​

Node Details

• Name: memoryVectorStore

• Label: In-Memory Vector Store

• Version: 1.0

• Type: Memory

• Category: Vector Stores

​

Base Classes

• Memory

• VectorStoreRetriever

• BaseRetriever

​

Input Parameters

1. Document (optional, list)

   • Type: Document

   • Description: A list of documents to be stored in the vector store.

2. Embeddings

   • Type: Embeddings

   • Description: The embeddings model used to convert documents into vector representations.

3. Top K (optional)

   • Type: number

   • Description: Number of top results to fetch. Defaults to 4 if not specified.

   • Placeholder: 4

​

Outputs

1. Memory Retriever

   • Name: retriever

   • Base Classes: [Memory, VectorStoreRetriever, BaseRetriever]

2. Memory Vector Store

   • Name: vectorStore

   • Base Classes: [Memory, …BaseClasses of MemoryVectorStore]

​

Functionality

1. Initialization:

   • Creates a MemoryVectorStore from the provided documents and embeddings.

   • Can return either a retriever or the vector store itself based on the specified output.

2. Upsert Method:

   • Allows adding new documents to the vector store.

   • Returns the number of added documents and the added documents themselves.

​

Usage

This node is particularly useful in scenarios where:

• You need a quick, in-memory solution for storing and retrieving vector embeddings.

• The dataset is small enough to fit in memory.

• You want to perform similarity searches on document embeddings.

• You need a simple integration within a larger language model pipeline.

​

Note

The in-memory nature of this vector store means that it’s not persistent between runs and may not be suitable for very large datasets or scenarios requiring data persistence.