Docling — разбор документов для RAG и ИИ

★ 65K

Docling is a document-parsing library that turns PDF, DOCX, PPTX, HTML, images, and other formats into a structured form (Markdown or JSON) convenient for AI and RAG. Reach for it when you need to reliably extract document content while preserving structure: text, headings, tables, reading order, and page elements — to then feed a RAG pipeline or process programmatically. It works as a Python library, can OCR scans, recognizes tables and layout, and integrates with popular frameworks (LangChain, LlamaIndex). Its focus is programmatically turning heterogeneous documents into clean structure for downstream processing, not a web app with a UI or table extraction alone (excalibur for tables from text PDFs): the value is a single high-quality document parser for AI pipelines. In a RAG stack it sits before indexing and the vector store. It is developed with IBM Research.

• #Documents