Generative AI Design Patterns: Solutions to Common Challenges When Building GenAI Agents and Applications

by Valliappa Lakshmanan, Hannes Hapke

Generative AI enables powerful new capabilities, but they come with some serious limitations that you'll have to tackle to ship a reliable application or agent. Luckily, experts in the field have compiled a library of 32 tried-and-true design patterns to address the challenges you're likely to encounter when building applications using LLMs, such as hallucinations, nondeterministic responses, and knowledge cutoffs.


This book codifies research and real-world experience into advice you can incorporate into your projects. Each pattern describes a problem, shows a proven way to solve it with a fully coded example, and discusses trade-offs.



Design around the limitations of LLMs
Ensure that generated content follows a specific style, tone, or format
Maximize creativity while balancing different types of risk
Build agents that plan, self-correct, take action, and collaborate with other agents
Compose patterns into agentic applications for a variety of use cases

Genres: Artificial Intelligence, Technical, Technology

506 pages, Paperback

Published November 11, 2025

Reviews

[Özgür]

Certainly a gap in the current literature and good coverage of patterns overall, however the book although touching the right points looks a bit outdated at release and limited in depth e.g. seems to be missing key patterns like reasoning models, telemetry in observability, RL in LLM context, no graph RAG, multi-modality (book is named Generative AI not LLM's)...The curse of writing a book in such a fast developing field.

ps - I like Chip Huyen's books and AI Engineering seems to be hugely successful. However I think she didn't do good for the community when she started writing books in stateless long tweet thread format which this book mentions at the beginning suggesting the book is following Chip's style...which causes a lack of overall informational unity and integrity of the content and how things are related, data backed context on why x is better than y etc...

[Lucille Nguyen]

Note: This review was written on the pre-release version on O'Reilly for Public Libraries

An outstanding book that goes over multiple design patterns for deploying generative AI applications. Is very well-researched and generalizes many cases that I already knew about, but hadn't thought about applying it in certain cases. A lot more general than their last book on Machine Learning Design Patterns, a very good read and very ambitious given how quickly this domain is changing.

[Xianshun Chen]

Highly recommend for learning genai, I came across various patterns throughout other readings, bit this books categorized and organized them into a nice structured framework with some pretty nice and easy to follow illustrations

[Afonso]

I love the book's structure and its approach to presenting design patterns with problems and solutions, and found many chapters to be genuinely useful, but much of the book is simply outdated by 2026 standards. That is the curse of writing on such a fast-moving topic as GenAI, I guess.