RAG-Enabled Visual Image Product Search

Demo

visual-search-demo.webm

Description

This project implements a Retrieval Augmented Generation (RAG) pipeline for visual image product search, built with the MERN stack (MongoDB, Express, React, Node.js). It leverages the Milvus vector database, Transformers-based encoders for image and text embeddings, and a local Ollama-based Gemma LLM to handle intelligent query responses. Licensed under the MIT License, this repository demonstrates how to combine state-of-the-art vector indexing, machine learning models, and custom logic for a visual similarity search system.

About

The goal of this repository is to showcase how to integrate:

• A MERN web application for user-facing functionality and backend logic.
• Milvus as a high-performance vector database to store and manage product embeddings.
• Transformers library for embedding generation and semantic representation.
• Ollama-based Gemma LLM to enhance search queries via natural language understanding.

Key Points

• Full-stack implementation with React and Node.js for a seamless developer experience.
• Simple REST API endpoints for searching and retrieving relevant products based on visual similarity.
• Customizable search strategies and easy extension with new models or data sources.
• Clear instructions for local deployment, with minimal overhead and minimal dependencies.