Comment by raghavbali

Maybe I missed something but this is a round about way of doing things where an embedding + ML classifier would have done the job. We don't have to use an LLM just because it can be used IMO