paperless-weasyprint

Paperless-ngx mail parser that swaps the upstream Gotenberg/Chromium dependency for WeasyPrint. Pure Python, no browser, smaller attack surface for HTML email body rendering.

Built it because Gotenberg's bundled Chromium structurally lags Debian's security pocket by 1-2 weeks, producing a recurring ~90 CRITICAL CVE window. The Gotenberg maintainer formally treats it as "trusted content only" -- which doesn't fit an inbox receiving email from the open internet.

what it does

Subclasses upstream's MailDocumentParser (in paperless.parsers.mail) and overrides the four methods that called Gotenberg/Tika:

• generate_pdf_from_mail -- WeasyPrint instead of chromium.html_to_pdf
• generate_pdf_from_html -- WeasyPrint with cid: attachment rewriting
• generate_pdf (the merge step) -- pypdf instead of Gotenberg's merge route
• tika_parse -- stdlib HTML->text instead of a Tika network call

Registers via the paperless_ngx.parsers entrypoint with score=100, beating upstream's MailDocumentParser (score=10) at parser dispatch.

v0.2.0: embed attachments

Optional. After the body renders, walks mail.attachments, appends embeddable ones to the merged PDF (PDF + JPEG + PNG inline; everything else gets a "skipped" appendix listing).

Combined with consumption_scope=EMAIL_ONLY and pdf_layout=HTML_ONLY on the Paperless mail rule, you get one document per email containing body and attachments together. (Note: that means EMAIL_ONLY no longer strictly means "only the email body" once this plugin is active -- documented trade-off, not a bug.)

install

Build the included Dockerfile, layer onto whatever Paperless image you run. The plugin auto-registers on container start; no env vars to set.

FROM ghcr.io/<your-username>/paperless-weasyprint:<tag>

limitations

• Some HTML email shapes don't render cleanly under WeasyPrint -- Outlook MSO conditionals, VML namespaces, position:absolute marketing layouts. Receipt and transactional emails are usually fine.
• No PDF/A output. If you set PAPERLESS_OCR_OUTPUT_TYPE=pdfa, the generated PDFs aren't PDF/A compliant. Post-process with ocrmypdf --output-type pdfa --skip-text if you need it.
• Subclass risk: if upstream changes the signature of generate_pdf_from_mail or generate_pdf_from_html, this plugin breaks until updated. Pin paperless-ngx by digest, watch for upstream changes.
• Embed-attachments scope is PDF + JPEG + PNG only. Other types (.docx, .xlsx, .heic, .svg, .zip, etc.) skip with an appendix note.

maintenance

I built this quickly as a stopgap for the notable WTF with the Gotenberg project state and don't plan to actively maintain it -- though I'll do occasional updates. Happy to support anyone who wants to take it forward.

license

AGPL-3.0-or-later. Matches Paperless-ngx upstream.