How to make Advanced RAG work in practice with Evaluations, Agentic Patterns and Generative AI with LLM