docling-colab

One-click Google Colab notebook: upload PDFs in your browser, get clean Markdown files back. No local Python install, no GPU, no model downloads beyond what Docling needs for fast text extraction.

Open in Colab

Click the badge, then Runtime → Run all (or Ctrl+F9). The file picker appears when the upload cell runs; pick one or more PDFs. Markdown files auto-download when the last cell runs.

What it does

1. Installs Docling.
2. Prompts you to upload PDFs via google.colab.files.upload().
3. Runs Docling with OCR and table-structure detection disabled — text-only, fast, no heavy ML model downloads.
4. Writes a <name>.md next to each PDF in the Colab filesystem.
5. Auto-downloads each .md to your browser.

Need OCR or structured tables?

Open the third cell and flip the flags:

pipeline_options.do_ocr = True            # for scanned PDFs
pipeline_options.do_table_structure = True  # preserve table row/col structure

Enabling either triggers a one-time model download (~1 GB on first run inside that Colab session).

Notes

• The Colab runtime is ephemeral — files and installs vanish when the session ends. Re-run from the top if you come back later.
• Large uploads can be slow through the browser picker. For many PDFs, mount Google Drive and point the loop at a Drive folder instead.
• Free-tier Colab is CPU-only but that's fine for this notebook since the ML models are off.

Want to run it locally / batch hundreds of PDFs?

See docling-batch — parallel, fault-tolerant CLI with resume.

License

MIT — see LICENSE.