What is Agentic RAG

This means using as many possible LLMs during RAG pipeline, ie considering 1 LLM for every distinct task

Simple System
Prompt -----> |LLM| -------> Response

Agentic RAG
            ----Yes--> |vectorDB|---> |Retriever| --> Documents --> Augumented Prompt -->|LLM|--> Reponse
            |               /\                          |
            |                |-------|Evaluator LLM|----|
            |
prompt -> |Router LLM|
          Does it need
          call to vectorDB
            |
            |--No-->|LLM| -----> Reponse