I feel a playground, or at least much more comprehensive examples, would be needed to get me interested.
Perhaps for small B2C applications it can work well, but I struggle to see how it could handle 80% of the stuff that hits the support mail for our complex B2B application.
"Handling 80% of queries via LLM" sounds great, until you note that it's equivalent to "20% of support inquiries are answered by a hallucinating robot that can't actually solve their problem"
I don't think it hallucinates very much. If a customer asks a question that is answerable based on documents you uploaded, it will find the answer! I am working on a playground to showcase the RAG.
Perhaps for small B2C applications it can work well, but I struggle to see how it could handle 80% of the stuff that hits the support mail for our complex B2B application.