microsoft/markitdown

MarkItDown is a lightweight Python utility for converting various files into Markdown. In plain terms, Markdown is a simple text format with light markup for headings, lists, tables, and links, and this tool reads in a PDF, Word document, spreadsheet, or other file and spits out the same content as Markdown text. The README explains that the goal is to feed that text into large language models and other text analysis pipelines, where Markdown is preferred because mainstream LLMs natively understand it and it is token-efficient. According to the README, MarkItDown can convert PDF, PowerPoint, Word, Excel, images (extracting EXIF metadata and running OCR), audio (EXIF metadata and speech transcription), HTML, text-based formats like CSV, JSON, and XML, ZIP archives, YouTube URLs, EPubs, and more, while trying to preserve structure such as headings, lists, tables, and links. It is most comparable to textract but with a focus on preserving document structure. You install it from PyPI with pip install 'markitdown[all]', or pick specific format extras like [pdf, docx, pptx]. Python 3.10 or higher is required. From there you can call it as a command-line tool, as a Python library via a MarkItDown class with a convert method, or through a Docker image. The library also supports third-party plugins, an OCR plugin that uses an LLM vision client, and an option to delegate conversion to Azure Document Intelligence. You would use MarkItDown when you want to feed mixed office documents and media into an LLM-powered pipeline and need a uniform text representation rather than handling each format yourself. The full README is longer than what was provided.