engine_readiness#

scikitplot.cleanprompt.engine_readiness(name, language='en', model=None, size='sm', check_assets=False)[source]#

Decide whether one engine can run for a language, and how to fix it if not.

Parameters:
namestr

"spacy" or "nltk".

languagestr, default=’en’

Language code.

modelstr, optional

Explicit spaCy model, overriding language and size.

sizestr, default=’sm’

Preferred spaCy model size.

check_assetsbool, default=False

Also check NLTK’s data packages, which imports NLTK. spaCy’s model is always checked, because that check imports nothing.

Returns:
EngineReadiness

The decision, its reason and its remedy.

Raises:
PolicyError

If name is not a known engine, or size is not a known size.

Parameters:
Return type:

EngineReadiness

Notes

Developer notes — the order of the checks is the order of the remedies. Language first, because no installation fixes an engine that cannot read the language. Then the package, because a model cannot be downloaded into a library that is not there. Then the data. The first failure decides the status and the remedy, so a user is told the next step rather than all of them at once.

This is CP-093. doctor --ner reported entity detection as active on a machine with spaCy installed and no model, and the next inspect --ner failed on the missing model; the same for NLTK without its data. Both decisions read the package and never the data. A diagnosis that disagrees with the run it describes is worse than none, because the run is where the user finds out — after deciding the text was safe to send.

Examples

>>> engine_readiness("nltk", "de").reason
"nltk does not support language 'de'; it handles en"