Skip to content

RealtimeOutputGuardrail

A guardrail that checks the output of the agent.

execute: OutputGuardrailFunction<"text">;

The function that performs the guardrail check.

OutputGuardrail.execute


name: string;

The name of the guardrail.

OutputGuardrail.name


optional policyHint: string;

This will be passed to the model to inform it about why the guardrail was triggered and to correct the behavior. If it’s not specified the name of your guardrail will be passed instead.