Docling
Python library and CLI that converts documents into structured data for AI
What it is
Docling parses PDFs, Office files, HTML, EPUB, email, images, audio and more into a unified document model, recovering layout, reading order, tables, formulas and code. It exports to Markdown, HTML, or lossless JSON, runs OCR on scans, and can use vision-language models or speech recognition, all locally. Developers building RAG pipelines and agents use it to feed clean document content into LangChain, LlamaIndex, Haystack, or any MCP-capable agent.