Docify-RAG is a full-stack AI-powered chatbot that lets you upload any PDF file and ask natural language questions about its contents. It uses Retrieval-Augmented Generation (RAG) with Gemini AI to deliver accurate, contextual answers — like ChatGPT, but focused on your documents.
| Layer | Tech |
|---|---|
| Frontend | React (Vite) + Tailwind CSS |
| Backend | Flask (Python) |
| Vector DB | FAISS |
| Embeddings | Sentence Transformers |
| LLM | Gemini Pro (Google AI) |
| PDF Parsing | PyPDF |
| Auth | .env API key loading |
project-root/
├── Docify-RAG/ # React frontend (Vite)
├── Server/ # Flask backend
├── .gitignore
├── package.json # Runs both client and server
git clone https://github.com/your-username/docify-rag.git
cd docify-rag# Install React client cd Docify-RAG npm install # (Optional) Add Tailwind if not already: # npx tailwindcss init -p # Go back and create virtual environment for Flask cd ../Server python -m venv venv source venv/bin/activate # or venv\Scripts\activate on Windows pip install -r requirements.txt
Create a .env file in /Server:
GOOGLE_API_KEY=your_google_api_key_here
At the root of the project:
npm install concurrently --save-dev npm run dev
This will:
- Start Flask backend on
http://localhost:5000 - Start React frontend on
http://localhost:5173
-
Upload a PDF.
-
The backend extracts, chunks, and embeds the text.
-
When you ask a question:
- It retrieves top-k relevant chunks using FAISS
- Then it builds a prompt using Gemini + conversation history
-
Gemini generates a natural, contextual answer.
- Natural chat interface
- Context-aware RAG using FAISS
- Supports any text-based PDF
- Realtime answers powered by Gemini
- Works locally, no external database needed
- Streamed responses (typing effect)
- File management and history
- Support for image-based PDFs (OCR)
- Multi-user authentication
MIT License.