No format-specific setup

Lessons 1-3 only ever touched a PDF. This lesson runs the exact same DocumentConverter instance over project_plan.docx and team_update.pptx too, with no extra configuration, docling detects the format from the file (extension, or content sniffing for a stream) and routes it to the right backend automatically.

converter = DocumentConverter()
for filename in ["quarterly_report.pdf", "project_plan.docx", "team_update.pptx"]:
    result = converter.convert(SAMPLE_DATA / filename)

This is the same pattern markitdown's course used for its own multi-format lesson, one converter object, a loop over files, docling just does more work per file to get there (a trained layout model for PDF, versus reading DOCX/PPTX's own native structure directly, since those formats already carry heading and paragraph information in their XML).

Why DOCX and PPTX convert with zero tables or pictures here

project_plan.docx and team_update.pptx are plain text and bullet lists, this course's fixtures for those formats don't happen to include an embedded table or image. That's not a limitation, docling reads native tables and images from DOCX/PPTX just as well as it does from PDF, this fixture set simply doesn't exercise that path. Lesson 8 (tables) and Lesson 10 (pictures) both work from quarterly_report.pdf specifically because that's the fixture built with a table and a picture.

Checkpoint

  • One DocumentConverter, many formats: no per-format setup, format is detected from the file itself.
  • Backend differs by format under the hood: PDF goes through a layout model, DOCX/PPTX read their own native XML structure, both end up as the same kind of DoclingDocument.
  • A top-level heading (# Title) versus a section header (## Title): DOCX's Heading 0/title style maps to a top-level heading in Markdown, distinct from the section_header labels seen on the PDF in Lesson 3.

If anything here still feels unclear, ask before moving to Lesson 5.