IBM's Docling Hits 65K Stars Fixing the PDF Problem That Breaks RAG

The docling-project/docling repository has crossed 65,000 stars on GitHub, making it one of the most-watched open-source tools in the document-processing space. Released under an MIT license and hosted by the LF AI and Data Foundation, the project has become a default choice for teams that need to feed messy PDFs, spreadsheets, and slide decks into large language models.

Docling parses diverse document formats, handles advanced PDF understanding, and integrates with the generative AI ecosystem. The traction reflects a real pain point: most retrieval-augmented generation stacks fall apart the moment a PDF has a two-column layout, a nested table, or a scanned page, and Docling is one of the few open-source options that tries to solve all of those at once.

What the library actually does

Docling reads a document and emits a structured representation you can hand to an LLM, a vector store, or a chunker. Supported inputs include PDF, DOCX, PPTX, XLSX, HTML, EPUB, WAV, MP3, WebVTT, Box Notes, email formats (EML, MSG), images, LaTeX, DocLang, and plain text. PDF handling covers page layout, reading order, table structure, code, formulas, and image classification.