docling-project/docling is a Python tool that parses documents in multiple formats, PDF among them, organizes them into a common document structure, and exports them as Markdown, HTML, JSON and more. The repository description sets out the goal of preparing documents for generative AI use. The latest release in the provided record is v2.145.0. However, the record does not state what changed in this version.
The tool normalizes input documents into a common representation called DoclingDocument and then converts them into multiple formats. The key point is that it unifies the different structures that documents have, so they can be handed off to further processing. The README summary lists the supported features: page layout and reading order, table and formula understanding, OCR, and audio ASR. ASR is the process of turning speech in audio into text. It also describes LangChain integration, a command-line tool, and the docling-serve API server. The repository license is MIT.
The verifiable features are the ones described in the repository and the README, and the provided record contains no changelog for v2.145.0. So the features added or modified in this version cannot be pinned down here. The purpose of handling document parsing and conversion as a single flow is confirmed, but version-by-version changes are hard to judge from the given information alone.