Faiss

Faiss is a library for efficient similarity search and clustering of dense vectors. Unlike cloud vector stores, Faiss runs locally on your machine and saves data to your file system.

Faiss Node

Prerequisites

Method 1: NPM/Source Installation

For local installations, ensure your machine can compile C++ modules.

• Linux: Install build-essential.
• MacOS: Run xcode-select --install to get Clang/C++ tools.
• Windows: Install Visual Studio Build Tools and Python.

Method 2: Docker

If you are running CiniterFlow using Docker, Faiss is ready to use.

Setup

Inputs

Input	Description
Document	Connect any node from the Document Loader category.
Embeddings	Connect any node from the Embeddings category.

Parameters

• Base Path: The folder path where the index files (faiss.index and docstore.json) will be saved.
  • If left blank, data is stored in RAM and will be lost when CiniterFlow restarts.
  • Docker Path Example: /root/.ciniterflow/faiss

  • Note: Ensure the path is mapped to a volume in your docker-compose.yml file to prevent data loss upon container restart.

  • Local Path Example: C:\Users\YourName\ciniterflow_data

Configuration and Ingestion

1. Add a new Faiss node on the canvas.
2. Enter a folder path in the Faiss node’s Base Path field.
3. Connect your Document Loader and Embeddings nodes to the Faiss node.
4. Click Upsert Vector Database to process your documents.

Verify

To verify your data has been upserted, navigate to the folder you specified in the Base Path field. You should see files faiss.index and docstore.json.

Resources

• LangChain JS Faiss Documentation