Skip to content

README.md

Latest commit

 

History

History
209 lines (157 loc) · 5 KB

README.md

File metadata and controls

209 lines (157 loc) · 5 KB

🎓 QED-Scribe - Hybrid RAG

This project is a production-grade Retrieval-Augmented Generation (RAG) Tutor built to understand and explain handwritten notes of QEDS, even with OCR noise.

Python Stack OCR

It uses:

  • 🔍 Hybrid Retrieval (BM25 + BGE-M3)
  • 🔄 RAG Fusion using FLAN-T5 paraphraser
  • 🔥 Cross-Encoder Reranking
  • ✂️ Contextual Compression
  • 🧹 OCR Noise Sanitization
  • 🧠 LLaMA3 (Ollama) for answering

Perfect for academic notes, handwritten documents, mathematical derivations, and noisy OCR text.

📸 Demo

Asking about Gini Coefficient:

App Demo

Asking questions about Homogenous Differential Equations:

App Demo (The AI retrieves the correct handwritten module, fixes the math symbols, and explains the concept)

🔍 What Problem Does This Solve?

Students often struggle with:

  • Handwritten lecture notes
  • OCR errors in scanned material
  • Fragmented definitions and formulas
  • Losing context while asking follow-up questions

QEDS-GPT solves this by:

  • Indexing cleaned OCR notes into a semantic vector database
  • Retrieving the most relevant modules and topics
  • Correcting equations and notations
  • Allowing multi-turn, memory-aware academic conversations

🚀 Features

🧠 Conversational RAG with Memory

  • Multi-turn chat interface
  • Context retained across refreshes and sessions
  • No login required

📚 Semantic Retrieval over Notes

  • ChromaDB vector store
  • Hugging Face bge-m3 embeddings
  • Module-aware retrieval with metadata:
    • Semester
    • Subject
    • Module

✍️ OCR Noise Correction

  • Fixes broken equations and symbols
  • Rewrites math in clean LaTeX
  • Repairs fragmented sentences

🚫 Hallucination Controls

  • Vague-query detection
  • Relevance filtering
  • Uses academic knowledge only to reconstruct missing context

📦 Production Deployment

  • Dockerized application
  • Deployed on Hugging Face Spaces
  • Git LFS support for vector index files
  • Secure API key management

🧠 Architecture

RAG Pipeline

Basic RAG Pipeline

Detailed Retrieval & Embedding Flow

Detailed Architecture

User
↓
Streamlit Chat UI
↓
Semantic Retrieval (ChromaDB)
↓
OCR Cleanup & Context Sanitization
↓
LLM Reasoning (Groq)
↓
Answer + Updated Memory

🧪 Example Capabilities

Query: "Explain the Slutsky substitution effect."

System Action: Retrieves Economics notes from Semester 3, fixes OCR typos in the definition, and presents the derivation.

Query: "Solve the Bernoulli differential equation from Module 1."

System Action: Filters for "Semester 4 - Diff Eq", finds the specific raw formula, converts it to clean LaTeX, and explains the solution steps.

🔒 Notes

Handwritten notes are private and not included

Vector database is stored using Git LFS

Memory is stored per user locally via SQLite

📁 Project Structure

QEDS-RAG-Project/
│
├── chroma_db/              # Vector database
│
├── src/
│ ├── streamlit_app.py      # Main conversational RAG app   
│
├── Dockerfile
├── requirements.txt
└── README.md

⚙️ Tech Stack

- Language: Python
- UI: Streamlit
- Vector DB: ChromaDB
- Embeddings: Hugging Face `BAAI/bge-m3`
- LLM: Groq (LLaMA-3.1-8B-Instant)
- Deployment: Docker + Hugging Face Spaces

🛠️ Installation & Usage

1. Prerequisites

Make sure Ollama is installed and running with Llama 3:

ollama run llama3

2. Setup Environment

git clone https://github.com/apooorv19/QEDS-RAG-Project.git
cd QEDS-RAG-Project
pip install -r requirements.txt
  1. Set environment variable
export GROQ_API_KEY=your_api_key_here
  1. Run the app
streamlit run src/streamlit_app.py

🐳 Docker Deployment

docker build -t qeds-gpt .
docker run -p 8501:8501 -e GROQ_API_KEY=your_api_key qeds-gpt

🌐 Live Demo

https://huggingface.co/spaces/Apooorv69/QEDS-RAG-Project

👤 Author

Apurva Mishra
IMSc Quantitative Economics & Data Science
Birla Institute of Technology, Mesra

GitHub: https://github.com/apooorv19
LinkedIn: https://www.linkedin.com/in/apooorv/

📜 Citations & Credits

@misc{paruchuri2025surya,
  author       = {Vikas Paruchuri and Datalab Team},
  title        = {Surya: A lightweight document OCR and analysis toolkit},
  year         = {2025},
  howpublished = {\url{https://github.com/VikParuchuri/surya}},
  note         = {GitHub repository},
}