Help with RAG pipeline

Im building a system like this

1. user give an input

1. LLM Use pandasqueryengine to query a CSV to see if it can find any matching rows,

1. If not found in the CSV file

1. then use a fine tuned model to predict whats the output based on the given input

can someone teach me how to do step 3 + 4? I dont know how to let LLM knows to make another decision IF its not found in csv