v2.1.0 - Complete PDF Specialist Removal & Clean Architecture

🎯 Major Architecture Cleanup

This release completes the removal of PDF specialist frameworks and cleans up all related filtering logic for a simpler, more maintainable codebase.

✅ What's Changed

• Complete removal: All PDF specialist frameworks (PyMuPDF, PDFPlumber, Playa) entirely removed from codebase
• Simplified architecture: Removed unnecessary filtering logic since PDF specialists no longer exist
• Cleaner CI/CD: Streamlined workflows to only run multi-format frameworks
• Updated configurations: Fresh AI assistant rules and documentation

🔧 Current Framework Support (6 frameworks)

1. Kreuzberg (sync/async + OCR variants) - 71MB installation
2. Extractous - Rust-based performance - ~100MB
3. Unstructured - Enterprise solution - 146MB
4. MarkItDown - Microsoft converter - 251MB
5. Docling - IBM Research ML - 1GB+

🚀 Technical Improvements

• Removed filter_pdf_specialists() function - no longer needed
• Simplified reporting.py initialization logic
• Cleaner generate_index.py without filtering calls
• Updated GitHub Actions workflows to exclude PDF specialists entirely

📊 Fresh Benchmark Results Coming

This release will trigger a complete benchmark run generating:

• Clean visualization charts without PDF specialists
• Updated performance rankings for multi-format frameworks only
• Fresh GitHub Pages deployment with corrected data

🎯 Impact

• Fairer comparisons: Only multi-format frameworks with similar scope
• Cleaner results: No more mixed single-format vs multi-format data
• Simpler maintenance: Less code, fewer edge cases
• Better insights: Focus on frameworks users actually compare

---

Breaking Change: PDF specialist frameworks are completely removed. Use previous releases if you need PDF-only framework benchmarks.