Topics Boundaries¶
Note
This capability is provided by the “Topics Boundaries Guardrail” plugin, which you need to install. Please see Installing plugins.
This plugin is Not supported
The Topics Boundaries guardrail check that questions to a LLM or responses, either:
Discuss only about one of the allowed topics
Do not discuss about one of the forbidden topics
This is particularly useful to ensure that users cannot cause a Chatbot designed to discuss about a certain topic to veer off course.
Topics Boundaries guardrail leverages an auxiliary LLM for assessing whether a query or response respects the rule (“LLM as a judge” approach).
It can be configured to either:
block the request (with an error)
simply audit the off-track conversation
politely decline to respond