Kreuzberg

Kreuzberg is an advanced Python library designed for seamless text extraction from a variety of documents including PDFs, images, office documents, and more. This library features a unified asynchronous interface, making it simple and efficient to retrieve text content.

Key Features

• User-Friendly API: Kreuzberg offers a straightforward and clean API that requires no complex configuration, making text extraction hassle-free.
• Local Processing: The library operates completely offline, eliminating the need for external API calls or cloud services, enhancing privacy and security.
• Lightweight and Efficient: Kreuzberg is built to be resource-efficient, with minimal dependencies that prevent heavy resource usage, even without GPU support.
• Extensive Format Compatibility: It supports a wide range of document and image formats ensuring comprehensive text extraction capabilities.
• Modern Asynchronous Design: Implements modern Python features such as async/await and type hints, fostering an optimized experience for developers and applications.
• Open Source: Kreuzberg and its dependencies are released under a permissive open-source license, encouraging collaboration and contribution.

Architectural Highlights

Kreuzberg combines several powerful tools for its functionality:

• PDF text extraction utilizes pdfium2 for searchable content and Tesseract OCR for scanned PDFs.
• Document conversion capabilities are powered by Pandoc, alongside support for various popular file types like PowerPoint (python-pptx), HTML, and Excel (calamine).
• Smart text processing includes features like automatic encoding detection and support for Markdown, plain text, and many more markup formats.

Supported Document Formats

Kreuzberg supports a multitude of formats, including:

• Documents: PDF, Word (.docx), PowerPoint (.pptx), OpenDocument Text (.odt), EPUB, LaTeX, and more.
• Markup and Text: HTML, Markdown, Plain Text, reStructuredText, and more.
• Data Files: Excel (various formats), CSV, Jupyter Notebooks, and more.
• Images: JPEG, PNG, TIFF, BMP, and several others.

Easy-to-Use APIs

The library provides both asynchronous and synchronous APIs for text extraction. Essential functions include:

async def extract_document():
    pdf_result = await extract_file("document.pdf")
    img_result = await extract_file("scan.png", language="deu")
    docx_result = await extract_file(Path("document.docx"))

Kreuzberg also supports batch processing for efficient handling of multiple files or contents, facilitating both async and sync operations:

async def process_documents(file_paths: list[Path]) -> None:
    results = await batch_extract_file(file_paths)

Robust Error Handling

Kreuzberg features comprehensive error handling to assist users with any potential issues that may arise during text extraction, including handling for validation errors, OCR inaccuracies, and dependency issues.

Contribution and Development

Kreuzberg welcomes contributions from the community. Interested developers can contribute by following the local development guidelines provided in the repository.

Explore Kreuzberg to enhance your text extraction tasks and streamline workflows with its powerful capabilities.