paperless-ngx-erechnung

German E-Rechnung parser plugin for Paperless-ngx v3.

Handles the two formats produced by the German B2B E-Rechnung mandate and makes their data visible and searchable in Paperless-ngx:

• XRechnung — pure XML invoice (UBL Invoice, UBL CreditNote, or UN/CEFACT CII).
• ZUGFeRD / Factur-X — hybrid PDF/A-3 with an embedded XML invoice.

How it plugs in

Paperless-ngx exposes a parser plugin framework that scans the paperless_ngx.parsers Python entry-point group on startup (see paperless/parsers/registry.py in paperless-ngx). This package declares two parsers under that group:

Both parsers gate their score() on content inspection — XRechnung confirms the root element namespace, ZUGFeRD looks for an embedded factur-x.xml / ZUGFeRD-invoice.xml. When the file does not match, score() returns None and Paperless's built-in parsers take over normally.

Built-in parsers score 10; this plugin scores 100, so it cleanly outranks defaults for matching files without affecting anything else.

What this plugin does

• Archive PDF rendering for XRechnung via the official KoSIT XRechnung-Visualization XSLT (xr-pdf.xsl) plus Apache FOP (XSL-FO → PDF).
• Pass-through archive for ZUGFeRD PDFs (preserves PDF/A-3 conformance and the Factur-X signature).
• Extracted invoice fields (number, date, due date, seller, totals, …) surfaced in Paperless's metadata sidebar and prepended to the searchable text body.

What this plugin does NOT do

• Creating Custom Fields (no plugin hook for this exists yet) from XML invoice fields
• Generation / creation of E-Rechnungen (Paperless-ngx is not a data entry tool)
• OCR fallback for scanned/non-Factur-X PDFs — those continue to use Paperless's built-in Tesseract path.

Vendored KoSIT assets

src/paperless_ngx_erechnung/xslt/ contains a pinned snapshot of the KoSIT XRechnung-Visualization project (Apache-2.0): the XSLT stylesheets, the Apache FOP configuration (conf/fop.xconf), and the Source Serif Pro TTF fonts (conf/fonts/, SIL OFL) that the FOP config registers. See xslt/README.md for the pinned version and re-vendoring instructions.

Installation

Unfortunately, installing this plugin is more than a single click and depends very much on your setup. Two pieces need to land in Paperless-ngx's runtime environment:

1. The Python package itself, installed into the same interpreter Paperless runs under, so its paperless_ngx.parsers entry-points are discovered at startup.
2. Apache FOP + a headless Java JRE, available on PATH, so the XRechnung → archive PDF stage works.

Once both are in place, Paperless picks up the parsers at startup — look for Loaded third-party parser 'XRechnung' … in the logs to confirm succesful installation.

Docker / Docker Compose

1. Set up Paperless-ngx with Docker Compose (e.g. by using one of the official compose files). Since this upstream image ships neither FOP nor this plugin, we need to extend it with a small Dockerfile.
2. Create the following Dockerfile next to your preferred docker-compose.(XYZ).yml:

FROM ghcr.io/paperless-ngx/paperless-ngx:3.0

USER root
RUN apt-get update && apt-get install -y --no-install-recommends \
  default-jre-headless fop \
&& rm -rf /var/lib/apt/lists/*
RUN pip install --no-cache-dir paperless-ngx-erechnung
USER paperless

3. Build this modified image. You can choose any image and tag name that you like for this custom image. Here, we'll be using paperless-ngx-erechnung:latest:

docker build -t paperless-ngx-erechnung:latest .

4. Then point the webserver service in your existing docker-compose.yml at the newly created image instead of the upstream one:

services:
  webserver:
    image: paperless-ngx-erechnung:latest
    # ...keep the rest of your existing config (env, volumes, depends_on, ...)

5. Restart your Compose stack to pick up the new image:

docker compose up -d

(Alternatively, if you want to do this in a single step, you can let Compose handle the build by setting build: . on the service instead of image: and run docker compose build webserver && docker compose up -d.)

services:
  webserver:
    build: .
    # ...keep the rest of your existing config (env, volumes, depends_on, ...)

Bare-metal

If Paperless-ngx runs directly on the host, install Apache FOP and a JRE via the system package manager, then install the plugin into the same Python environment Paperless uses (commonly a venv under /opt/paperless/). Paths and service names depend on how you provisioned Paperless — substitute yours:

# System-side: FOP + JRE
sudo apt install default-jre-headless fop          # Debian/Ubuntu

# Plugin into Paperless's Python environment
sudo -u paperless /opt/paperless/.venv/bin/pip install paperless-ngx-erechnung

# Restart so the entry-point group is rescanned
sudo systemctl restart paperless-webserver paperless-consumer paperless-scheduler

Development

Archive PDF rendering shells out to Apache FOP, which needs a JRE. The Dockerfile installs both (default-jre-headless + fop). For local development, install via your package manager:

brew install fop          # macOS — pulls in openjdk
apt install fop           # Debian/Ubuntu — pulls in default-jre-headless

The XSLT-dependent rendering tests skip if the fop binary is not on PATH. All other runtime dependencies (qpdf for pikepdf, libxml2/libxslt for lxml, libpdfium, Saxon-HE native libs, Pillow) are bundled in their respective wheels and need no system install.

Then:

uv sync
uv run pytest

AI Disclaimer

This plugin was developed with the help of an AI Agent (Claude Code). However, it was not "vibecoded" (as in "a human did not look at or understand this code") or "one-shotted". Every line of this code was reviewed by a capable developer and the whole project is tested, refined and documented by real humans.

License

GPL-3.0-only — matching Paperless-ngx itself, since this plugin runs in-process with Paperless and the combined work is most cleanly distributed under a single license. See LICENSE for the full text.

Bundled KoSIT XSLT under src/paperless_ngx_erechnung/xslt/ is Apache-2.0 (GPL-3.0-compatible); see src/paperless_ngx_erechnung/xslt/LICENSE.