Review of Retrieval-Augmented Generation (RAG) System Architecture, Limitations, and Practical Challenges

This paper presents a comprehensive review of Retrieval-Augmented Generation (RAG) systems, focusing on their architecture, retrieval pipeline, indexing mechanisms, embedding models, vector databases, and large language model integration. It also discusses key limitations including retrieval failures, hallucinations, context fragmentation, latency, scalability, security, privacy, and evaluation challenges. The paper synthesizes findings from publicly available academic literature and industry standards to provide students, researchers, and practitioners with an overview of current RAG systems and future research directions.