Docling
Converts PDFs and other files into structured data for search, RAG, and AI agents.
Description
Docling is an open-source document-processing library for generative AI workflows. It extracts content from PDFs, office files, HTML, images, and additional formats.
Its PDF pipeline detects reading order, tables, formulas, and page structure, with OCR available for scans. Outputs include Markdown and JSON and can be accessed through a Python API, CLI, or RAG integrations.
Core processing runs locally after downloading required models. Complex documents and OCR may demand substantial resources; library code is MIT-licensed, individual models may differ, and IBM's managed service is a separate offering.
