The text discusses the concept of Retrieval Augmented Generation (RAG), a central paradigm in Large Language Model (LLM) application development to address the limitations of LLMs being only periodically trained on public data. It introduces Self-Reflective RAG, which captures the idea of using an LLM to self-correct poor quality retrieval and/or generations. The text also highlights the importance of feedback, re-generating the question, and re-retrieving documents in this process. Additionally, it mentions LangGraph, an easy way to implement LLM state machines, which supports the more general process of "flow engineering" for RAG with specific decision points and loops. Two approaches are discussed: Corrective RAG (CRAG) and Self-RAG, both of which use a graph-based representation to outline their workflows in LangGraph. CRAG introduces a lightweight retrieval evaluator to assess the quality of retrieved documents and performs web-based document retrieval to supplement context. In contrast, Self-RAG trains an LLM to generate self-reflection tokens that govern various stages in the RAG process. The text concludes that self-reflection can greatly enhance RAG, enabling correction of poor quality retrieval or generations, and provides cookbooks for implementing ideas from two interesting papers, CRAG and Self-RAG.