kreuzberg-dev/kreuzberg v3.0.0

on GitHub

Enhancements:

• added support for multiple OCR backends: added PaddleOCR and Easy OCR (feature)
• added support for having no OCR backend (feature)
• changed Tesseract OCR to optional (enhancement)
• added support for registering creating custom extractors (feature)
• added support for overriding builtin extractors (feature)
• added support for post-processing hooks (feature)
• added support for validation hooks (feature)
• added PDF metadata extraction using Playa-PDF (feature)
• added optional chunking support (feature)
• added documentation site (documentation)

Breaking Changes:

• Changed ExtractionResults from NamedTuple to TypedDict (breaking change; api)

Internal:

• Rework internals to allow extensibility by changing to a class-based architecture (internal; architecture)