Faiss Vector Store

​

Node Details

• Name: Faiss

• Type: Faiss

• Version: 1.0

• Category: Vector Stores

​

Base Classes

• Faiss

• VectorStoreRetriever

• BaseRetriever

​

Input Parameters

1. Document (optional, list)

   • Type: Document

   • Description: List of documents to be stored in the vector store

2. Embeddings

   • Type: Embeddings

   • Description: Embedding model used to convert documents into vector representations

3. Base Path to load

   • Type: string

   • Description: Path to load or save the faiss.index file

   • Placeholder: C:\Users\User\Desktop

4. Top K (optional, additional parameter)

   • Type: number

   • Description: Number of top results to fetch (default: 4)

   • Placeholder: 4

​

Outputs

1. Faiss Retriever

   • Name: retriever

   • Base Classes: [Faiss, VectorStoreRetriever, BaseRetriever]

2. Faiss Vector Store

   • Name: vectorStore

   • Base Classes: [Faiss, …FaissStore base classes]

​

Functionality

The Faiss Vector Store node provides two main functions:

1. Upsert:

   • Processes input documents

   • Creates vector embeddings using the provided embedding model

   • Stores the vectors in a Faiss index

   • Saves the index to the specified base path

2. Init:

   • Loads an existing Faiss index from the specified base path

   • Creates either a retriever or a vector store based on the output selection

   • Configures the similarity search function to avoid illegal invocation errors

​

Use Cases

• Efficient similarity search in large document collections

• Building retrieval-augmented generation (RAG) systems

• Creating semantic search engines

• Implementing recommendation systems based on content similarity

​

Notes

• The node includes a custom implementation of similaritySearchVectorWithScore to handle potential issues with the number of requested results exceeding the total number of stored vectors.

• It’s designed to work seamlessly within a larger system, likely a node-based workflow for natural language processing tasks.