What is Retrieval Augmented Generation (RAG) for documents?

Retrieval Augmented Generation (RAG) is a technique that combines a language model with an external knowledge base to improve factual accuracy. The process starts by indexing a corpus of documents—often using vector embeddings from models like Sentence‑BERT or BM25 for keyword retrieval. When a user query arrives, a retrieval component fetches the top‑k relevant passages. These passages are concatenated with the prompt and fed to a generative model (e.g., GPT‑3 or a fine‑tuned LLM). The model then generates an answer conditioned on both the query and the retrieved context. RAG reduces hallucinations because the LLM is anchored to real documents. It is widely used in question‑answering systems, legal document review, and customer support bots. Implementations can be built with open‑source stacks (FAISS + HuggingFace) or managed services like AWS Bedrock’s Retrieval Augmented Generation feature.