Lecture Parser

A production-ready tool that converts PDF lecture slides into structured study notes using AI vision and text models via the Groq API.

Features

• 📄 Converts PDF slides to detailed summaries using vision AI
• 📝 Generates problem-solving-focused study notes
• 🔄 Batched processing for efficient API usage
• 📊 Progress tracking with tqdm
• 🔧 Configurable via environment variables
• ⚡ Automatic retry with exponential backoff

Prerequisites

• Python 3.8+
• Groq API key (Get one here)
• poppler-utils (for PDF to image conversion)

Installing poppler

macOS:

brew install poppler

Ubuntu/Debian:

sudo apt-get install poppler-utils

Windows: Download from poppler releases and add to PATH.

Installation

1. Clone the repository:

git clone <repository-url>
cd lecture_parsing

2. Create a virtual environment:

python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install dependencies:

pip install -r requirements.txt

4. Configure environment:

cp .env.template .env
# Edit .env and add your GROQ_API_KEY

Usage

1. Place your PDF lecture files in the PDFs/ directory

2. Run the parser:

python main.py

3. Find generated notes in the Summaries/ directory

Configuration

All configuration can be done via the .env file:

Required

• GROQ_API_KEY: Your Groq API key

Optional

• VISION_MODEL: Vision model to use (default: meta-llama/llama-4-scout-17b-16e-instruct)
• TEXT_MODEL: Text model to use (default: openai/gpt-oss-20b)
• PDF_DIR: Input directory for PDFs (default: PDFs)
• SUMMARY_DIR: Output directory for summaries (default: Summaries)
• TMP_IMG_DIR: Temporary directory for slide images (default: tmp_slide_imgs)
• SLIDES_PER_BATCH: Number of slides to process per batch (default: 20)
• MAX_IMG_PIXELS: Maximum pixels per image (default: 33000000)
• MAX_IMG_SIZE_MB: Maximum image file size in MB (default: 4)

Output Format

Each processed PDF generates a Markdown file containing:

• Study Notes: Problem-solving focused summary with:
  • Core formulas and definitions
  • Key concepts and intuition
  • Problem-solving strategies
  • Common pitfalls
• Raw Slide Summaries: Detailed summaries of each slide

Error Handling

• Automatic retry with exponential backoff for API calls
• Individual PDF processing errors don't stop the entire batch

Project Structure

lecture_parsing/
├── main.py                 # Main application
├── requirements.txt        # Python dependencies
├── .env.template          # Environment variable template
├── .env                   # Your configuration (git-ignored)
├── README.md              # This file
├── PDFs/                  # Input PDFs
└── Summaries/             # Generated notes (auto-created)