PDF Structure Extractor using Tiny LLM

Project Overview

This project extracts and restructures the content of a PDF file by identifying headings, subheadings, and plain text using a lightweight language model (Tiny LLM ~60MB). Due to LLM input token limitations (1024 tokens), large PDFs are broken into manageable chunks before processing. The tool outputs structured JSON and a final PDF.

Project Structure

Got it! Here's your Project Structure section rewritten in a clean, numbered format (ideal for GitHub READMEs):

---

📁 Project Structure

1. uploads/ Stores uploaded PDF files from the frontend.

2. index.js Main Node.js backend file to handle file upload and processing.

3. processpdf.js Contains logic to split PDFs into chunks and manage token limits for LLM.

4. frontend/ (React + Vite app)

   • public/ – Static assets
   • src/ – React components and frontend logic
   • package.json – Frontend dependencies and scripts

5. llm-server/ (Python backend with FastAPI)

   • main.py – API endpoints for processing text using Tiny LLM
   • download_model.py – Downloads and prepares the LLM model
   • requirements.txt – Python dependencies for the server

6. .gitignore Lists files and folders ignored by Git (like node_modules and venv).

7. README.md Documentation of the project.

---

Let me know if you want this embedded back into the full README and exported as an updated PDF.

Features

• Upload a PDF via UI
• Automatically chunks into ~500-word parts
• Sends to Tiny LLM and extracts structure
• Displays JSON result
• Outputs a structured final PDF Setup Instructions

1. Clone the Repository: git clone https://github.com/your-username/pdf-structure-extractor.git cd pdf-structure-extractor
2. Setup Python Environment: PDF Structure Extractor using Tiny LLM cd llm-server python -m venv venv source venv/bin/activate (Windows: venv\Scripts\activate) pip install -r requirements.txt
3. Install Node Dependencies: Run 'npm install' in root and 'frontend/' folders. Running the App
4. Start Node Backend: node index.js
5. Start LLM Server: cd llm-server uvicorn main:app --host 127.0.0.1 --port 8000
6. Start Frontend: cd frontend npm run dev How to Use
7. Open the app at http://localhost:5173
8. Upload a PDF
9. Wait for JSON and PDF outputs
10. JSON is displayed; PDF is downloadable Important Notes

• Tiny LLM has a 1024-token input limit
• We split large PDFs into ~500-word chunks
• Python, pip, and virtualenv are required
• Run 'npm install' in each folder with a package.json PDF Structure Extractor using Tiny LLM