RAG: bring an open book to generation

Picture it

Retrieval finds evidence outside the model’s weights and puts it into the prompt. The generator can still misread it or invent an answer. Retrieval quality, permissions, freshness, and citation checks are separate engineering problems.

What to understand

• Chunking, embeddings, lexical/vector/hybrid search, reranking
• Retrieval recall, answer faithfulness, ACLs, freshness, and evaluation

Follow the references

1. 1 · Paper

   Lewis et al. · Retrieval-Augmented Generation ↗

   Read the retrieval-plus-generation diagram and identify where retrieved documents enter.

2. 2 · Documentation

   Meta · Faiss similarity search ↗

   Use the index-selection guidance to connect search recall with latency and memory.