What Is MarkItDown?

The problem: LLMs read text, your files aren't text

An LLM prompt is a string. Almost nothing you actually need to feed it starts out as a string: it's a .docx memo, a .pptx deck, an .xlsx spreadsheet, a .pdf spec sheet, a .png screenshot, or a web page. Before any of that can go into a prompt, into a RAG index, or into a tool call, something has to turn it into text.

MarkItDown is Microsoft's open-source answer to that problem: a Python library that converts a wide range of file formats into Markdown. Not plain text, Markdown specifically, because Markdown still carries structure an LLM can use: headings stay headings, tables stay tables, list items stay list items. That structure is often the difference between an LLM correctly answering "what's the third bullet under Key Requirements" and it guessing.

Where MarkItDown fits next to this repo's other converters

This repo also has a LiteParse course. Both libraries turn documents into text an LLM can consume, but they solve different problems:

Neither replaces the other. If everything you handle is PDFs and you need to catch every table cell and scanned page, LiteParse's narrower focus pays off, that comparison is Lesson 11. If your input folder is a mixed bag of office documents, images, and text files, and a "probably right" Markdown conversion is enough, MarkItDown's breadth is the better fit. This course assumes no prior exposure to either, but Lessons 10 and 11 lean on basic familiarity with the LlamaIndex course (Lessons 1-4) and the existence of the LiteParse course.

Installing MarkItDown: the [all] extra matters

MarkItDown ships with format support split into optional extras, so you only install what you need:

pip install markitdown            # base package: plain text, html, csv, zip, a handful of others
pip install markitdown[pdf]       # adds PDF support
pip install markitdown[docx]      # adds .docx support
pip install markitdown[pptx]      # adds .pptx support
pip install markitdown[xlsx]      # adds .xlsx support
pip install markitdown[all]       # every format extra, including pdf/docx/pptx/xlsx

If you install the bare package and try to convert a PDF, you don't get a vague error, you get a MissingDependencyException that names the exact extra you're missing. This project's pyproject.toml already specifies markitdown[all], and uv sync installed it, so every format used in this course works out of the box. Nothing else to install, no API key required for anything through Lesson 5.

The code, piece by piece

version = importlib.metadata.version("markitdown")

Confirms the package is actually installed and reports its version, the same idea as Lesson 1 of the Ollama course checking ollama list() before assuming anything works.

from markitdown import MarkItDown
md = MarkItDown()

Constructing MarkItDown() with no arguments registers every built-in converter whose dependencies are satisfied. This succeeds even without markitdown[all], individual converters just wouldn't be able to run yet, the MissingDependencyException only appears when you actually try to .convert() an unsupported format.