Extract text, tables, metadata, and images from PDF, Office, image, HTML, and archive documents.
npx skills add https://github.com/makramab/paddleocr-test --skill kreuzberg-makramab
Please help me install this Agent Skill. Skill: kreuzberg Source: https://github.com/makramab/paddleocr-test/tree/main Command: npx skills add https://github.com/makramab/paddleocr-test --skill kreuzberg-makramab
Kreuzberg provides high-performance, multi-format document extraction so developers can reliably convert PDFs, Office files, images, HTML, email and archive formats into clean text, structured tables, images and metadata for downstream processing. It eliminates brittle, manual parsing and inconsistent OCR results by offering configurable OCR backends, chunking, and language support across multiple language bindings.
Use the kreuzberg library to extract text, tables, metadata, and images from "document.pdf" and save the structured JSON output for downstream parsing.
High-intent search queries and answers about installing and using this skill.
You can extract text and tables from PDF and Office documents using multi-format document extraction tools that parse 75+ formats into structured JSON or markdown. These tools support configurable OCR backends like Tesseract to handle both native and scanned document text reliably.
The best way to batch process invoices for automated text extraction is using a library that supports sync and async workflows across Python, Node.js/TypeScript, and Rust. This allows server and client pipelines to process hundreds of mixed-format invoices and output structured JSON for LLM parsing or analytics.
Yes, you can use PaddleOCR or EasyOCR for scanned document OCR instead of Tesseract. The extraction framework offers built-in Tesseract by default and supports optional EasyOCR and Rust-native PaddleOCR backends, allowing you to configure the OCR engine based on your accuracy and language requirements.
Document extraction works with Python, Node.js/TypeScript, and Rust for server workflows, offering both synchronous and asynchronous modes. This multi-language binding support allows integration into batch jobs, server applications, and CLI pipelines for processing mixed-format archives prior to LLM parsing.
To feed extracted document text to an LLM for structured data parsing, you first extract the content from PDFs, emails, or images into clean text or markdown. You then pass this structured output directly to your LLM to reliably extract fields and perform analytics without brittle manual parsing.
Supported formats for extracting metadata and images include PDF, Office files like docx and xlsx, images, HTML, email, and archive formats. The extraction process handles over 75 formats, persisting table data, images, and metadata for downstream analytics and automated processing.
novvooExtract text, tables, metadata, and images from over 75 document formats.
droxeyExtract clean article text and structured news metadata from URLs into JSON.
damionrashfordExtract text and structure from images and PDFs using multiple OCR backends.
kevinnftExtract text from PDFs, scans, and images using OCR backends.
