requirements.txt

# Core PDF processing libraries
PyMuPDF>=1.23.0  # fitz - Fast, comprehensive PDF processing
pdfplumber>=0.10.0  # Excellent text extraction with layout awareness

# Table extraction libraries
camelot-py[cv]>=0.11.0  # Advanced table extraction with lattice and stream modes
tabula-py>=2.8.0  # Alternative table extraction using Java-based Tabula
JPype1>=1.4.1  # Java integration required by tabula-py

# Image processing (required by camelot)
opencv-python>=4.8.0  # Computer vision library for image processing
Pillow>=10.0.0  # Image processing

# Data handling
pandas>=2.0.0  # Data manipulation and analysis

# Token-efficient serialization for LLM input
toon-format>=0.1.0  # TOON format encoder/decoder - 30-60% token reduction vs JSON

# Additional dependencies for camelot
ghostscript>=0.7  # PostScript and PDF interpreter

# Math/Formula detection (classic ML - no GPU required)
scikit-learn>=1.3.0  # Machine learning for formula detection
numpy>=1.24.0  # Numerical operations
scipy>=1.11.0  # Scientific computing