Workflow 'Mitigate Hallucinations with LLM using RAG'

Is there an example of a workflow that does the following:

1. If the answer is not found on the vector i.e. spreadsheet of answers can the LLM provide an answer to the prompt instead? See the scenario below.

2. The scenario I tested with said workflow (see name on title) - I entered in the ‘Questions to the store’ a prompt that was not on the ‘KNIME DL Keras Integration Node Description’ and it returned with a ‘I'm sorry, I don't have information about a concert by Widespread Panic on Oct 22, 2001.’