O'Reilly published "Hands-On RAG for Production" by Ofer Mendelevitch and Forrest Sheng Bao in June 2026. The 358-page intermediate-to-advanced guide spans 10 chapters covering the full RAG pipeline from document parsing, chunking, and embedding through to vector search, reranking, and LLM integration with code examples using LangChain, pgvector, and Anthropic Claude. The book addresses production challenges including hallucination detection and correction, prompt-injection defenses, cost optimisation, and index freshness. Later chapters extend to agentic RAG (tool calling, model context protocol, multi-agent systems, observability), multimodal RAG (tables, images, audio, video), and GraphRAG. A dedicated chapter compares DIY RAG against platform options using Vectara as a worked reference. The book targets software engineers, ML engineers, and data architects moving RAG from prototype to enterprise scale (O'Reilly).
What happened
O'Reilly published "Hands-On RAG for Production" by Ofer Mendelevitch and Forrest Sheng Bao in June 2026 (ISBN 9798341621701). The 358-page, intermediate-to-advanced guide covers end-to-end RAG pipeline design with working code examples and quizzes, targeting engineers and architects already building retrieval-augmented applications. Forewords come from Sharon Zhou and Jim Dowling.
Technical scope
The 10-chapter book opens with the base RAG stack: document parsing (including vision-language model parsing), chunking strategies, embedding model selection, approximate nearest neighbor vector search with pgvector, and LLM integration with Anthropic Claude for generation (O'Reilly). It then covers production scaling topics including index freshness, cost management, hallucination detection and correction, prompt-injection defenses, hybrid retrieval, reranking, and RAG user experience design.
Advanced RAG coverage
The final three chapters extend the stack. Chapter 7 covers agentic RAG with tool calling, model context protocol (MCP), agent-to-agent communication, LangChain and LlamaIndex frameworks, agentic memory, and observability/tracing. Chapter 8 addresses multimodal RAG with embedded tables, images, audio, and video. Chapter 9 covers GraphRAG with knowledge graph construction, graph querying, and accuracy vs cost trade-offs. A closing chapter discusses the future of RAG including edge small language models, federated retrieval, and context engineering.
Industry context
Production RAG deployments frequently fail at retrieval misses, generation hallucinations, high latency under load, ingestion pipeline ownership, and data security gaps. A structured guide pairing explicit coverage of those failure modes with runnable code addresses gaps vendor documentation and short tutorials typically leave open.
Context and significance
The evaluation chapter -- covering LLM-as-a-judge, offline and online RAG evaluation, latency, uptime, and cost metrics -- is notable for practitioners who often lack formal evaluation pipelines. The DIY-versus-platform chapter, using Vectara as the worked reference, gives architects a decision framework beyond generic component comparisons. Available on O'Reilly Learning and Amazon.
Scoring Rationale #
A major publisher's production-focused RAG guide with code examples across base RAG, agentic, multimodal, and GraphRAG is a solid practitioner resource. It is not an event -- no new model, benchmark, or regulatory development -- but it addresses a gap between tutorial-level RAG content and enterprise deployment realities. Score sits at the lower end of the Solid range.
Practice with real FinTech & Trading data
90 SQL & Python problems · 15 industry datasets
[Active Verified Users by Income TierEasy](/problems/sql/active-verified-users-by-income)
[Technology Stocks with High BetaMedium](/problems/sql/technology-stocks-with-high-beta)
[Portfolio Performance ScorecardHard](/problems/sql/portfolio-performance-scorecard)
250 free problems · No credit card
See all FinTech & Trading problems