gets (old) mc-providers via pip?

[philbudne]

requirements.txt has "mc-providers" which will bring in the last version published to pypi? Maybe update to use a github URL??

Looks like it's only used for stop words? I know moving that functionality to metadata-lib would be wrong, so perhaps move the functions and data to a new repo brought in by github URL by both s-c and mc-providers?

[philbudne]

Also:
sous_chef/tasks/language_tasks.py (lazy) imports mc_providers.language for top_dectected which isn't used ANYWHERE else (ie; in web-search). top_detected brings in fasttext, which ISTR was mentioned as a heavy requirement in slack.

The story-indexer pipeline gets language from mcmetadata, which looks for HTML hints and/or uses py3langid