Hello,
I’m new to RASA and now I’m trying to build a chatbot using CALM, I read the documents but couldn’t find a straightforward tutorial about giving the CALM chatbot the ability to reply to Chitchat and FAQs from a predefined answers without “RAG” and without “Enterprise search feature”, please advise.
There’s a built in chitchat feature by default. It’s called pattern_completed.
If you don’t want to use the built-in RAG feature then you could create flows for each FAQ or you could use the NLU based response selector.
But the CALM way to do FAQ’s is with the built in enterprise search or connect an existing RAG with Custom Information Retrieval
Thanks for your reply. if I used the NLU based response selector. what is the best way to combine it inside CALM
Do you want to use an LLM or avoid use of an LLM? Why not use a RAG for FAQ?
I don’t want to use RAG for FAQ, to avoid hallucinations and make sure that all answers are accurate. If I used RAG in production I would have to use guardrails also, increasing the cost and latency.
Another alternative is RAG with extractive search