How to use docling-slim packages

[rrjohnson85]

Question

I see the recent version added support for slim packages, but it's not exactly clear how to use them. For example, if I need to convert docx to markdown, which packages to install?

I've tried docling-slim[format-office] (and docling-slim[format-office,convert-core,extract-core]) only and receive errors when importing DocumentConverter that indicate a dependency on DoclingParseDocumentBackend.

I then tried adding docling-core and docling-parse only to receive errors regarding the HTMLDocumentBackend and BeautifulSoup.

[dosubot]

Hi @rrjohnson85! I'm Dosu and I’m helping the docling team.

@rrjohnson85 The issue is that DocumentConverter eagerly imports backends for all supported formats (PDF, HTML, etc.), so even if you only need DOCX→Markdown, you'll hit import errors for missing backends unless those dependencies are installed too — or you're on a version with the recent import-guarding fixes.

For DOCX to Markdown, the minimum extras are:

pip install "docling-slim[convert-core,format-docx]"

Or with the office bundle:

pip install "docling-slim[convert-core,format-office]"

[1]

However, the DoclingParseDocumentBackend and HTMLDocumentBackend errors you're seeing are caused by DocumentConverter trying to import PDF and HTML backends at import time, even if you don't use them. This was addressed in PR #3375, which guards optional dependency imports so they don't fail on slim installations. Make sure you're on the latest version that includes this fix.

If you're still hitting these errors on the latest version, you can work around them by also installing the missing format extras:

pip install "docling-slim[convert-core,format-office,format-html,format-pdf]"

But this pulls in more dependencies than you strictly need. The proper fix is the guarded imports from PR #3375, so upgrading should resolve it without the extra packages.

Tip: extract-core already includes convert-core [2], so you can use extract-core instead of convert-core if you also need extraction features.

To reply, just mention @dosu.

---

Share context across your team and agents. Try Dosu.

[rrjohnson85]

As suggested, I've tried:

• docling-slim[convert-core,format-office,format-html,format-pdf]
• docling-slim[extract-core,format-office,format-html,format-pdf]

I now get issues about latex backend

[dosubot]

@rrjohnson85 Unfortunately, the core issue is that DocumentConverter eagerly imports all format backends at module load time — including LaTeX, HTML, PDF, etc. — with no conditional or lazy loading. So even if you only install a subset of format extras, the import will fail on whichever backend dependency is missing next.

For the LaTeX error specifically, you'd need to add format-latex:

pip install "docling-slim[extract-core,format-office,format-html,format-pdf,format-latex]"

But realistically, you'll likely keep hitting the next missing backend. Until the backend imports are refactored to be guarded/lazy (PR #3375 only addressed transformers in the datamodel layer, not backend imports [1]), the practical workaround for docling-slim is to install all format extras or just use the standard bundle:

pip install "docling-slim[standard]"

This is essentially the same as the full docling package in terms of dependencies, which defeats the purpose of slim for your use case. This looks like a gap that needs to be addressed — the backend imports need try-except guards or lazy loading so that only the backends for your installed extras are loaded.

To reply, just mention @dosu.

---

Share context across your team and agents. Try Dosu.

[rrjohnson85]

I've also tried with format-latex and format-markdown, but then get torch errors. I'll stick with full bundle/standard for now.