Minimal RAG NextJS

Overview

Minimal RAG NextJS is a proof-of-concept demonstrating that you can implement Retrieval-Augmented Generation (RAG) without relying on expensive vector database subscriptions. This project leverages the power of pgvector, a simple PostgreSQL extension, combined with OpenAI embeddings to create a cost-effective, local vector search solution.

Features

• Uses pgvector extension for PostgreSQL
• Integrates OpenAI embeddings for vector creation
• Implements a NextJS web application for file upload and search
• Chunks and stores vectors in a local database
• Provides a simple, intuitive user interface

Prerequisites

• Node.js (v14 or later)
• Docker and Docker Compose
• OpenAI API key

Installation

1. Clone the repository:

   git clone https://github.com/MartinKondor/minimal-rag-nextjs.git
   cd minimal-rag-nextjs
   

2. Install dependencies:

   npm install
   

3. Set up the database:

   docker-compose up -d
   npx prisma generate
   npx prisma db push --force-reset
   

4. Modify the given .env file in the root directory (if you wish):

   SKIP_ENV_VALIDATION=0
   NODE_ENV="development"
   POSTGRES_URL="postgres://postgresuser:postgrespassword@localhost:54322"
   POSTGRES_URL_NON_POOLING="postgres://postgresuser:postgrespassword@localhost:54322?pool=false"
   

Usage

1. Start the development server:

   npm run dev
   

2. Open your browser and navigate to http://localhost:3000.

3. Follow the on-screen instructions:

   • Enter your OpenAI API key (this is used client-side and not stored)
   • Upload a text file (max 20,000 characters)
   • Enter a search query
   • View the most relevant chunks from your uploaded file

How It Works

1. File Upload: The app chunks your uploaded text file and creates embeddings using OpenAI's API.
2. Vector Storage: These embeddings are stored in the local PostgreSQL database using pgvector.
3. Search: When you enter a query, it's converted to an embedding and compared against the stored vectors.
4. Results: The most similar text chunks are retrieved and displayed, ranked by relevance.

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

Future Improvements

• Implement user authentication
• Add support for multiple file types
• Optimize vector search for larger datasets
• Implement more advanced RAG techniques

Acknowledgements

• OpenAI for their embedding API
• pgvector for enabling vector operations in PostgreSQL
• Next.js for the React framework

Contact

Martin Kondor - https://martinkondor.github.io/

Project Link: https://github.com/MartinKondor/minimal-rag-nextjs

License

Copyright © Martin Kondor

MIT License. See the LICENSE file for more details.