[Feature]: Add monitor mode support to Lakera guardrails #18015
Description
The Feature
Current State
Lakera guardrail always blocks on violations. There is no way to log violations without blocking such that Lakera can learn from the risks.
Other guardrails have a "monitor" or "block" mode. Lakera needs the same
Target State
Support on_flagged: "monitor" (log but don't block)
Support on_flagged: "block" (current behavior, default)
Works with existing pre_call and during_call hook
Motivation, pitch
I'm working with a user of LiteLLM who really wants this feature so they can implement security as observability to start, before moving into a blocking state. They like using LiteLLM as a gateway.
LiteLLM is hiring a founding backend engineer, are you interested in joining us and shipping to all our users?
No