Preprocess prompts with one LLM to redact personal identifiable information (PII) before sending it to another LLM. Useful when processing prompts with LLMs hosted on a third party server, so the prompts are redacted by a local or self-hosted LLM beforehand.
Configuration settings
PII Detection Model - string - The LLM that is used for the redaction
Require Confirmation - boolean - Whether the redacted prompt needs to be confirmed by sending another message containing '--c'
Fallback Behavior - enum - The behavior for when redaction of PII has failed