Manage spaCy pipeline package using pipenv with Pipfile / Pipfile.lock

[darioprencipe]

Hello,

I am using pipenv to do package management in my python (Flask) dockerized application.

Need few components of your trained pipeline for Italian language. Now, in your nice docs here, you outline how to install trained pipelines in automated processes using pip.

However, there's no mention on how to handle - in this case - it_core_news_sm as a package using pipenv, whereas my standard flow with any other packages would be:

1. install with pipenv (pipenv install)
2. update Pipfile.lock file
3. use pipenv sync at the end of my Dockerfile

What's the way to go about installing spacy pipelines using pipenv?

Thanks a lot in advance!
Dario

[thomashacker]

Hey, thanks for your question.
A possible solution could be to directly download the models from spacy-models and install them locally / or handle them however you'd handle local/private wheel packages.

Hope that helps!

[darioprencipe]

Hello,
what worked exactly as intended for me was running pipenv install $(spacy info it_core_news_sm --url).

This added that package reference in my Pipfile such that I could lock it.

As less and less Python users use plain pip and requirements.txt to manage dependencies in their project, I think you should add a basic documentation support for pipenv usage.

Thanks!

[thomashacker]

Hey,
Thanks for the info! We'll add this to our list