Hands-On RAG for Production: Design, Develop, and Deploy Production-Ready RAG Applications
by O'Reilly Media (Ofer Mendelevitch, Forrest Sheng Bao)
Take a RAG system from a base pgvector stack to evaluated, production-scale retrieval, with agentic, multimodal and GraphRAG chapters.
Overview
Hands-On RAG for Production: Design, Develop, and Deploy Production-Ready RAG Applications is an O'Reilly book by Ofer Mendelevitch and Forrest Sheng Bao, with forewords by Sharon Zhou and Jim Dowling. Ebookpoint lists 358 pages and an ebook release date of May 27, 2026. It has ten chapters: Introduction to Retrieval-Augmented Generation, The Base RAG Stack, Scaling Your RAG Stack, Deploying RAG to Production, The RAG Platform, Evaluating Your RAG Application, From RAG to AI Agents, Multimodal RAG, Knowledge-Enhanced RAG, and The Future of RAG. According to a Let's Data Science summary, the base stack runs from document parsing, including vision-language model parsing, through chunking, embedding model selection and approximate nearest neighbor search with pgvector to generation with Anthropic Claude, and the code uses LangChain. Production material covers index freshness, cost management, hallucination detection and correction, prompt-injection defenses, hybrid retrieval, reranking and RAG user experience design. The evaluation material covers LLM-as-a-judge, offline and online evaluation, and latency, uptime and cost metrics. The agent chapter covers tool calling, the Model Context Protocol, agent-to-agent communication, agentic memory and tracing. The multimodal chapter handles tables, images, audio and video, and the GraphRAG material covers knowledge graph construction, graph querying and accuracy versus cost trade-offs. The build-versus-buy framework uses Vectara as its worked reference, so read that material knowing it centers on one vendor. The summary rates the book intermediate to advanced, aimed at engineers already building retrieval applications, and says it includes working code examples and quizzes.
At a Glance
- Topic
- RAG
- Level
- Intermediate
- Format
- Book
- Cost
- Paid
- Duration
- 358 pages, self-paced
- Provider
- O'Reilly Media (Ofer Mendelevitch, Forrest Sheng Bao)
- Hands-on
- Yes — code/exercises
- Certificate
- None
What You’ll Learn
- ✓Build a base RAG stack with document parsing, chunking, embeddings and pgvector search
- ✓Scale retrieval with hybrid search and reranking while managing index freshness and cost
- ✓Detect and correct hallucinations, and add defenses against prompt injection in RAG applications
- ✓Evaluate RAG quality offline and online using LLM-as-a-judge plus latency, uptime and cost metrics
- ✓Decide whether to build RAG in-house or use a RAG-as-a-service platform
- ✓Extend retrieval into agents using tool calling, MCP, agent-to-agent communication, memory and tracing
- ✓Handle tables, images, audio and video in multimodal RAG, and construct and query knowledge graphs
Highlights
- •A full chapter on evaluation covers offline and online testing alongside latency, uptime and cost
- •Production concerns such as index freshness, prompt injection and RAG user experience get direct coverage alongside the base stack
- •The later chapters extend past text retrieval into agents, multimodal data and knowledge graphs
- •The platform comparison uses Vectara as its worked reference, so treat its build-versus-buy conclusions as one vendor's example
- •The code stack is concrete: LangChain, pgvector and Anthropic Claude, with quizzes to check understanding
Who It’s For
Best For
- ✓Software and ML engineers scaling a RAG prototype toward production
- ✓Data architects planning enterprise retrieval systems with privacy and explainability requirements
- ✓Teams deciding between an in-house RAG stack and a RAG-as-a-service platform
Prerequisites
- •Experience building or maintaining a retrieval-augmented application, since that is the stated audience
- •Enough Python and SQL familiarity to follow LangChain and pgvector code examples
FAQ
What is Hands-On RAG for Production: Design, Develop, and Deploy Production-Ready RAG Applications?
Hands-On RAG for Production is a 2026 O'Reilly book for software engineers, ML engineers and data architects who already build retrieval applications. After reading it, you can build a base RAG stack, measure hallucinations, latency and cost, defend against prompt injection, and decide between building in-house and buying a platform.
Is Hands-On RAG for Production: Design, Develop, and Deploy Production-Ready RAG Applications free?
Hands-On RAG for Production: Design, Develop, and Deploy Production-Ready RAG Applications is a paid resource.
What level is Hands-On RAG for Production: Design, Develop, and Deploy Production-Ready RAG Applications for?
Hands-On RAG for Production: Design, Develop, and Deploy Production-Ready RAG Applications is aimed at a intermediate audience. Recommended background: Experience building or maintaining a retrieval-augmented application, since that is the stated audience, Enough Python and SQL familiarity to follow LangChain and pgvector code examples.
How long does Hands-On RAG for Production: Design, Develop, and Deploy Production-Ready RAG Applications take?
Expect roughly 358 pages, self-paced. Most learners work through it at their own pace.
What will I learn from Hands-On RAG for Production: Design, Develop, and Deploy Production-Ready RAG Applications?
You'll learn: Build a base RAG stack with document parsing, chunking, embeddings and pgvector search; Scale retrieval with hybrid search and reranking while managing index freshness and cost; Detect and correct hallucinations, and add defenses against prompt injection in RAG applications; Evaluate RAG quality offline and online using LLM-as-a-judge plus latency, uptime and cost metrics; Decide whether to build RAG in-house or use a RAG-as-a-service platform; Extend retrieval into agents using tool calling, MCP, agent-to-agent communication, memory and tracing; Handle tables, images, audio and video in multimodal RAG, and construct and query knowledge graphs.
Topics
Sources
This page was written from 3 sources, 2 on domains other than amazon.com.