No success building custom multilang Presidio Docker image

[bfisseler]

Describe the bug
I tried to modify presidio-analyzer Docker image to use Presidio with different languages, e.g. German.

To Reproduce
Steps to reproduce the behavior:

1. Clone the repo
2. Add additional language codes presidio_analyzer/conf/default.yaml, e.g.

  -
    lang_code: de
    model_name: de_core_news_md

3. Add supported languages to default_analyzer.yaml as well as default_recognizers.yaml, e.g.

     - en
     - de

4. build presidio-analyzer Docker image: docker build -t multilang_presidio .
5. Build process breaks after ~20 minutes with the following error:

...
27.60   - Installing shellingham (1.5.4)
1492.1 
1493.4 PEP517 build of a dependency failed
1493.4 
1493.4 Backend subprocess exited when trying to invoke build_wheel
...[followed by many more lines of Python gibberish]
1496.9 Note: This error originates from the build backend, and is likely not a problem with poetry but one of the following issues with numpy (2.0.2)
1496.9 
1496.9   - not supporting PEP 517 builds
1496.9   - not specifying PEP 517 build requirements correctly
1496.9   - the build requirements are incompatible with your operating system or Python version
1496.9   - the build requirements are missing system dependencies (eg: compilers, libraries, headers).

Do I need a specific version of numpy? And if so, how do I specify this in the dockerfile?

Expected behavior
I had expected a new customized Docker image.

Screenshots
If applicable, add screenshots to help explain your problem.

Additional context
I tried to build presidio-analzyer on macOS running Sequoia.

[bfisseler]

I changed the Dockerfile for presidio-analyzer from python:3.13-slim back to python:3.12-slim and had success building the image.

Btw the doc on "Installing Presidio" lists that Presidio supports Python 3.9 to 3.12. Why does the Dockerfile build on Python 3.13 then?

That said I would highly appreciate more details on how to modify the Docker images for multiple languages at one. Will open another issue for that.

[SharonHart]

@bfisseler I was able to build the image in my machine (Apple silicone Mac), on which OS are you trying? have you tried the most recent main?

[bfisseler]

@SharonHart I am building the image on Apple with Intel running latest Sequoia 15.5. And yes, I yesterday downloaded the most recent main. Successfully built an image this morning after modifying Python 3.13 slim back to 3.12 slim.

[bfisseler]

I was able to build the image based on Pyhon 3.13 (python:3.13-slim) successfully on the M1 Mac I use at work. So could it be a Mac-Intel issue? Is Docker's build process not independent of the hardware?

[SharonHart]

@bfisseler Thanks for the additional info.

Some python packages (mostly C extensioned) ship binary wheels that are specific to a particular architecture, so it makes sense.

You can try to build with buildx for multi-platform, from presidio-analyzer dir:

docker buildx build --platform linux/amd64,linux/arm64 .

[omri374]

Closing as we revamped the docker workflows, please reopen if needed.