LLM Design Patterns

Build production LLM and agentic AI systems using modern patterns for context engineering, memory, loops, harnesses, RAG, tools, MCP, observability, security, governance, and reliable deployment. Key Features Master modern LLM patterns for context engineering, loops, memory, harnesses, RAG, tools, and agents Apply evaluation, observability, security, governance, and cost controls across production AI Learn by building one evolving product capability per chapter, ending with an integrated capstone Book DescriptionBuilding an LLM application is easy; making it reliable, secure, observable, and cost-effective in production is far harder. This book presents a practical pattern language for LLM and agentic AI systems through a PolicyOps project. You’ll begin with LLM architecture and evaluation-driven development, then compare adaptation strategies and optimize test-time reasoning, latency, cost, and quality. Next, you’ll engineer context, build advanced and agentic RAG pipelines, integrate Graph RAG, and add governed long-term memory. Each chapter adds one production component or concern, including model boundaries, evaluation gates, context planning, retrieval, memory, secure extensions, agent loops, observability, deployment, and release review. You’ll extend the system with typed tools, MCP, Agent Skills, plugins, and hooks; design single-agent, multi-agent, browser, and computer-use workflows; and add durable harnesses, LLMOps, security, governance, and human oversight. The capstone unifies these patterns in a governed, observable, secure knowledge-and-action agent, with examples from healthcare, finance, customer support, and software engineering. By the end of the book, you’ll be able to design, secure, deploy, and operate production LLM systems against quality, cost,…