redacting#

class scikitplot.cleanprompt.redacting(secrets=None, logger=None)[source]#

Scrub known secrets from log records for the duration of a block.

Parameters:
secretsiterable of str, optional

Surfaces to scrub.

loggerlogging.Logger, optional

Logger to attach to. Defaults to this submodule’s root logger.

Parameters:

Notes

User notes. Wrap anything that logs while holding personal data:

with redacting(vault.export().values()):
    do_work()

Developer notes. The filter is attached to every logger in this namespace, not to a handler, so it applies however the application has configured its output. Attaching it to the namespace root alone looked equivalent and was not: a logger’s filters see only that logger’s own records, never its children’s (CP-059). It is removed on exit, including when the block raises, so a long-lived process does not accumulate filters holding the secrets of every request it has ever served — which would be the very leak this is meant to prevent.

Examples

>>> with redacting(["topsecret@example.com"]):
...     pass