refactor: migrate to dedicated unoserver container

The old `unoconv` is not supported anymore upstream and needs to
be replaced. Instead of running all the subprocess management code
ourselves, we now just let this happen in it's own container. This
should improve service stability significantly, as it's now cleanly
separated.

While we're at it, also update the base image to newest debian Trixie

Author: winged

.github/workflows/tests.yml

@@ -57,7 +57,6 @@ jobs:
       - name: Set UID
         run: |
           echo "UID=$(id --user)" > .env
-          echo "ISOLATE_UNOCONV=true" >> .env
       - name: Build docker containers
         run: docker compose up -d --build
       - name: Run pytest
@@ -71,8 +70,6 @@ jobs:
       fail-fast: false
       matrix:
         version:
-          - "3.10"
-          - "3.11"
           - "3.12"
           - "3.13"
         database:
@@ -91,6 +88,10 @@ jobs:
           --health-retries 5
         ports:
           - 5432:5432
+      unoserver:
+        image: ghcr.io/unoconv/unoserver-docker:0.4.3
+        ports:
+          - 2003:2003
     steps:
       - uses: actions/checkout@v5
       - name: Install poetry
@@ -105,15 +106,15 @@ jobs:
       - name: Install dependendies
         run: |
           sudo apt-get update
-          sudo apt-get install -y --no-install-recommends util-linux unoconv libreoffice-writer libmagic1
+          sudo apt-get install -y --no-install-recommends libmagic1 media-types util-linux wait-for-it
           poetry install --extras full --with dev
       - name: Set environment
         run: |
           echo "ENV=dev" >> .env
           echo "UID=$(id --user)" >> .env
-          echo "ISOLATE_UNOCONV=true" >> .env
           echo "DATABASE_DIR=${{ runner.temp }}/document-merge-service/data" >> .env
           echo "MEDIA_ROOT=${{ runner.temp }}/document-merge-service/media" >> .env
+          echo "UNOSERVER_HOST=localhost" >> .env
       - name: Configure postgres
         if: ${{ matrix.database == 'postgres' }}
         run: |

.mypy.ini

@@ -4,6 +4,9 @@ exclude = document_merge_service/api/migrations
 [mypy-rest_framework.*]
 ignore_missing_imports = True
 
+[mypy-unoserver.*]
+ignore_missing_imports = True
+
 [mypy-babel.dates.*]
 ignore_missing_imports = True
 

CONFIGURATION.md

@@ -21,11 +21,13 @@ To scale the service a different database storage is needed. Any database suppor
 - `DATABASE_USER`: Username to use when connecting to the database
 - `DATABASE_PASSWORD`: Password to use when connecting to database
 
-## Unoconv
+## Unoconv/Unoserver
 
-- `UNOCONV_ALLOWED_TYPES`: List of types allowed to convert to. See `unoconv --show` (default: ['pdf'])
-- `UNOCONV_PYTHON`: String, defaults to "/usr/bin/python3.5"
-- `UNOCONV_PATH`: String, defaults to "/usr/bin/unoconv"
+Unoserver runs separately in a container. You can configure the connection by the following env variables:
+
+- `UNOSERVER_HOST`: String, defaults to "unoserver"
+- `UNOSERVER_PORT`: Integer, defaults to 2003
+- `UNOCONV_ALLOWED_TYPES`: List of types allowed to convert to. Defaults to  ['pdf']
 
 ## python-docx-template
 

Dockerfile

@@ -1,4 +1,4 @@
-FROM python:3.13-bookworm AS build
+FROM python:3.13-trixie AS build
 
 ARG ENV=docker
 ARG APP_HOME=/app
@@ -24,7 +24,7 @@ RUN \
   --mount=type=cache,target=.cache/pypoetry \
   poetry install --only-root
 
-FROM python:3.13-slim-bookworm
+FROM python:3.13-slim-trixie
 
 ARG UID=901
 ARG APP_HOME=/app
@@ -54,8 +54,7 @@ RUN \
   --mount=type=cache,target=/var/lib/apt,sharing=locked \
   apt-get update && apt-get install -y --no-install-recommends \
     libmagic1 \
-    libreoffice-writer \
-    unoconv \
+    media-types \
     util-linux \
     wait-for-it \
   && rm -rf /var/lib/apt/lists/*

Makefile

@@ -8,7 +8,7 @@ start: ## Start the development server
 	@docker compose up -d --build
 
 test: ## Test the project
-	@docker compose exec document-merge-service sh -c "ruff format --diff --fix . && ruff check --diff . && mypy document_merge_service && pytest --no-cov-on-fail --cov --create-db"
+	@docker compose exec document-merge-service sh -c "ruff format --diff . && ruff check --diff . && mypy document_merge_service && pytest --no-cov-on-fail --cov --create-db"
 
 shell: ## Shell into document merge service
 	@docker compose exec document-merge-service bash

docker-compose.yml

@@ -15,6 +15,8 @@ services:
       # - SECRET_KEY=
       # https://docs.djangoproject.com/en/2.1/ref/settings/#allowed-hosts
       # - ALLOWED_HOSTS=
+  unoserver:
+    image: ghcr.io/unoconv/unoserver-docker:0.4.3
 
 
 volumes:

document_merge_service/api/file_converter.py

@@ -4,7 +4,7 @@
 from django.conf import settings
 from django.http import HttpResponse
 
-from .unoconv import Unoconv
+from .unoconv import UnoConverter
 
 
 class FileConverter:
@@ -14,10 +14,7 @@ def convert(file_contents, target_format):
 
         with NamedTemporaryFile("wb", dir=dir) as tmp:
             tmp.write(file_contents)
-            unoconv = Unoconv(
-                pythonpath=settings.UNOCONV_PYTHON,
-                unoconvpath=settings.UNOCONV_PATH,
-            )
+            unoconv = UnoConverter()
             result = unoconv.process(tmp.name, target_format)
 
         status = 200 if result.returncode == 0 else 500