APIs for the "Guardrails Only" use case

[shaneutt]

Nemo Guardrails recognizes a use case for "guardrails only":

https://github.com/NVIDIA-NeMo/Guardrails/tree/develop/examples/configs/guardrails_only

Some LLM guardrail scenarios require invoking a guardrail configuration to check only the
input or the output (which was generated through other methods). In other words, the
interaction with the LLM will not happen through the guardrails layer but rather externally,
and the guardrails layer is only invoked to check the input/output.

However, based on using it, and searching through the issues I don't see that use case surfaced through the Guardrails Server, or any effort to do so yet. Is this correct? And further, what would that API surface look like, something perhaps in a similar vein to the moderation API?