🚀 Documize Conversion: community package for Documize's Word import / PDF export service

[vitetj]

Documize Community delegates two things to a separate service: Word document import and PDF export. Out of the box it points at Documize's shared hosted endpoint, and upstream recommends self-hosting the documize/conversion container instead and changing the conversion endpoint URL to your own.
So here it is as a Cloudron app. No UI, no login — it just answers the API calls your Documize instance makes.

Repo: https://github.com/vitetj/documizeconversion-cloudron
Upstream version: 3.4.0

Install

App Store → Install from URL:
https://raw.githubusercontent.com/vitetj/documizeconversion-cloudron/main/CloudronVersions.json

or:
cloudron install --versions-url https://raw.githubusercontent.com/vitetj/documizeconversion-cloudron/main/CloudronVersions.json --location conversion --domain example.com

Then, in Documize as an administrator, set the conversion endpoint to the app's origin. That's the whole setup.
Check it is alive with curl https://<app-domain>/api/version → 3.4.0.

Packaging notes

Upstream publishes no source release, only the Docker image, so the Dockerfile lifts the artifacts out of documize/conversion:3.4.0 and re-assembles them on cloudron/base:5.1.0: the Go API binary, the Aspose/PlantUML jars (run with OpenJDK 11 from the Ubuntu archive), puppet.js + node_modules with the Chromium puppeteer bundles, and the upstream TrueType fonts — without those, Word→PDF layout drifts.
Two upstream quirks shaped the start script, in case they help anyone packaging something similar:

• the API binary resolves java/ and puppet.js relative to its working directory and writes a log there, so it can't run from the read-only /app/code — it starts from a writable /run/documize that symlinks the code;
• its scratch directory follows TMPDIR, pointed at /app/data/tmp, because a large .docx would blow through the small /tmp tmpfs.
  Other details: httpPort 5010, health check /api/version, localstorage addon, runs as cloudron (never root), memoryLimit 2 GB because Word conversion spawns a JVM with -Xmx1000m on top of Chromium. The image is ~3.6 GB, mostly fonts and the bundled Chromium.

Heads up: the API is unauthenticated

That's how the upstream service works — Documize calls it server-to-server with no credentials, so the package can't put SSO or proxyAuth in front of it without breaking it. Anyone who knows the domain can submit a document for conversion. If the endpoint doesn't need to be public, restrict it at the firewall to your Documize server's IP.

Tested

Health check, a real .docx converted through POST /api/1/word (the Aspose/Java path), PDF rendering with the bundled Chromium as the cloudron user, and a restart with existing app data. Feedback and issues welcome on the repo.

[joseph]

@vitetj you can publish your package at https://ca.cloudron.io/ for others to find it more easily

[vitetj]

@joseph here we go ! https://ca.cloudron.io/app/documize-conversion