RAGify

A lightweight Retrieval-Augmented Generation (RAG) system built with Python (FastAPI) + React.js.
It allows users to upload documents (PDF, DOCX, CSV), store embeddings in a vector database, and ask natural-language questions over their own data.

---

Backend Features

• Upload documents (PDF, DOCX, CSV)
• Generate embeddings and store them in ChromaDB
• Query documents using local LLM (Ollama + TinyLlama)
• Support for multiple documents via unique doc_id
• REST APIs powered by FastAPI
• Uses LangChain for loaders, splitters, retrievers, and chains
• Fully offline-capable once models are installed

Frontend Features

• Ask natural-language questions over uploaded documents
• Real-time answers powered by the RAG backend
• Minimal and clean UI
• Fully decoupled (API-driven)
• Responsive design (desktop-first)

---

Tech Stack

• Language: Python, TypeScript
• Framework: FastAPI, React
• RAG Framework: LangChain
• Vector Database: ChromaDB
• LLM Runtime: Ollama
• LLM Model: TinyLlama
• Embedding Model: nomic-embed-text
• Styling: TailwindCSS
• UI Components: shadcn/ui

---

Project Structure (Backend and Frontend)

backend/
│
├── app.py
│
├── routes/
│   └── routes.py
│
├── controllers/
│   ├── csv_controller.py
│   ├── doc_controller.py
│   ├── pdf_controller.py
│   └── process_query.py
│
├── lib/
│   ├── state.py
│   └── validateFile.py

frontend/
│
├── src/
│   ├── components/
|   |   ├── HeaderLogo.tsx
│   │   ├── Nav.tsx
│   │   └── ui/
|   |       ├── button.tsx
│   │       └── input.tsx
│   │
│   ├── pages/
|   |   ├── Docs.tsx
|   |   ├── Explore.tsx
│   │   ├── Home.tsx
│   │   ├── Index.tsx
|   |   ├── NotFound.tsx
│   │   └── Rag.tsx
│   │
│   ├── utils/
│   │   └── Steps.ts
│   │
│   ├── App.tsx
│   ├── App.css
│   ├── index.css
│   └── main.tsx
│
├── index.html
├── tailwind.config.ts
├── vite.config.ts
└── package.json

---

Setup Instructions (Backend Only)

Clone the Repository

git clone https://github.com/08abhinav/RAGify.git 
cd backend

Create and Activate Virtual Environment

python -m venv venv
source venv/Scripts/activate

Install Python dependencies

pip install -r requirements.txt

Run backend

uvicorn app:app --reload

Access the API

• Server: http://localhost:8000
• Swagger UI: http://localhost:8000/docs

Important Note (LLM Setup)

This project uses TinyLLaMA via Ollama for local LLM inference. Before running the backend, make sure you:

• Install Ollama locally (Download from: https://ollama.com)
• Pull the TinyLLaMA model

• ollama pull tinyllama

• Run the model (Ollama runs on port 11434 by default)

  ollama run tinyllama

Setup Instructions (Frontend Only)

cd frontend

Install React dependencies

npm i

Run frontend

npm run dev

Access the Frontend

• Local URL: http://localhost:5173