kreuzberg-dev/kreuzberg v2.0.0

on GitHub

Enhancements:

• add sync methods (feature)
• add corrupt PDF searchable text detection with automatic OCR fallback (feature)
• add metadata extraction using Pandoc (feature)
• add multi-sheet worksheet (excel etc.) extraction (feature)
• add language, psm and pax_processes keyword arguments (enhancement; api)
• gated typing-extensions to Python 3.10 and below (enhancement; dependencies)
• added multi-loop compatibility by switching from asyncio to using anyio (enhancement; compatibility)
• added managed worker processes for Pandoc and Tesseract using anyio.to_process (enhancement; performance)
• replaced xslx2csv with python-calamine and improved implementation to extract all sheets in a workbook (enhancement; performance)

Breaking Changes:

• updated ExtractionResult to include metadata (breaking change; api)
• changed force_ocr to a kwarg (breaking change; api)

Internal:

• split the _extractors namespace into smaller packages and reorganized source code
• add matrix tests against all supported Python versions (internal; testing)
• refined ruff rules to enhance linting strictness
• increase coverage to >=99% coverage