Output validation using Nemo guardrails

[Abhishekrkr]

Hello everyone! I want to validate the output generated by the system for things like profanity. I am currently using Nemo guardrails, specially Colang 2.x to validate user input. But after the result has been generated from the system(possibly using different (instance of) LLM), I want to validate it. As I understand, nemo will validate the output only during generation phase (even using ‘self check output’, as colang would not work for output) and you cannot pass message with role ‘assistant’ to somehow validate it. What is the best way to do so, if I want to stick with Nemo guardrails for this? Do I need to use llm.invoke() with prompts directly without using any guardrails? Any better way of doing this?