RAG Foundations · 6-part series
From documents to reliable retrieval systems.
Six visual articles build one connected mental model: understand RAG, prepare evidence, retrieve candidates, combine and rerank them, evaluate what was found, then handle questions that need more than one search.
1What Is RAG?Retrieval, augmentation, generation, grounding, and the difference between a model and a system.Read →2How Documents Enter RAGPDF extraction, evidence objects, chunking, metadata, embeddings, and indexes.Read →3Keyword Search vs Vector SearchInverted indexes, BM25, embeddings, dense retrieval, and complementary failure modes.Read →4Hybrid Retrieval and RerankingCandidate generation, RRF, deduplication, cross-encoders, and the recall ceiling.Read →5Evaluating Retrieval SystemsPrecision, Recall, MRR, nDCG, gold evidence, and evidence completeness.Read →6Multi-hop, GraphRAG, and Agentic RAGQuery planning, graph traversal, iterative retrieval, verification, retry, and stopping.Read →