Skip to content

vrushankdhande/vectorless-rag

Repository files navigation

🚀 Vecless RAG (Vectorless Retrieval-Augmented Generation)

📌 Overview

Vecless RAG is an end-to-end Retrieval-Augmented Generation (RAG) system that eliminates the need for traditional vector databases. Instead of embeddings, it uses PageIndex for structured document retrieval and Groq LLMs for fast, context-aware response generation.

This approach simplifies architecture, reduces computational overhead, and enables efficient document querying.

Get Your API's By Creating Account.

⚙️ Features

  • 🔍 Vectorless document retrieval (no embeddings required)
  • ⚡ Fast inference using Groq LLM APIs
  • 🧠 Context-aware response generation
  • 📄 Structured document indexing via PageIndex
  • 🛠️ Lightweight and scalable architecture

🏗️ Architecture

  • Ingestion → Documents are indexed using PageIndex
  • Query Input → User provides a natural language query
  • Relevant Node Retrieval → PageIndex fetches relevant sections
  • LLM Processing → Groq LLM generates final response

🛠️ Tech Stack

  • Python
  • Groq API (LLM inference)
  • PageIndex API (document retrieval)
  • JSON processing

📂 Project Structure

vecless_rag/
│── vectorless_rag.ipynb   # Main Code File
│── constants.py         # Stores API key
│── requirements.txt     # Dependencies
└── README.md            # Documentation

🔑 Setup & Installation

  • 1️⃣ Clone the repository
git clone https://github.com/your-username/vecless_rag.git
cd vecless_rag
  • 2️⃣ Install dependencies
uv add -r requirements.txt
  • 3️⃣ Add API Keys -- Create a constants.py file:
Groq_API_KEY = "your_groq_api_key"
PAGEINDEX_API_KEY = "your_pageindex_api_key
  • ▶️ Usage -- Run the notebook:
jupyter notebook vecless_rag.ipynb

Steps:

  1. Load your document/index
  2. Enter a query
  3. Retrieve relevant nodes
  4. Generate response using LLM

📊 Example

  • Input Query:
What is Handwriting Recognition?
  • Output:
Handwriting recognition is a machine learning application that converts handwritten text into digital format...

✅ Advantages

  • No need for vector databases (FAISS, Pinecone, etc.)
  • Lower cost and complexity
  • Faster setup and deployment
  • Easier debugging and transparency

🚧 Future Improvements

  • Add evaluation metrics for retrieval quality
  • Integrate UI (Streamlit / React)
  • Support multiple document formats (PDF, DOCX)
  • Add caching for faster responses

🤝 Contributing

Contributions are welcome!

  • Fork the repository
  • Create a new branch
  • Submit a pull request

👤 Author

Vrushank Dhande Aspiring Data Scientist | Machine Learning Engineer

⭐ Support

If you found this project useful, please give it a ⭐ on GitHub!

About

Implements a vectorless RAG architecture using PageIndex APIs and Groq LLMs, enabling efficient document retrieval and response generation without traditional vector databases.

Topics

Resources

Stars

2 stars

Watchers

0 watching

Forks

Releases

No releases published

Packages

 
 
 

Contributors