LlamaIndex Workflows

"LlamaIndex Workflows: Data‑Connected Agents with Repeatable Retrieval Pipelines"
Modern LLM systems fail not only because models hallucinate, but because orchestration is hidden, retrieval is underspecified, and agent behavior is difficult to inspect or reproduce. This book is written for experienced AI engineers, platform architects, and advanced Python practitioners who need explicit control over data-connected applications. It shows how to use LlamaIndex Workflows as a disciplined runtime for building retrieval pipelines and agents whose behavior can be understood, tested, and operated with confidence.
Across the book, readers learn how workflows model execution through typed events, steps, context, and state; how repeatable retrieval systems are engineered from ingestion and indexing through reranking and synthesis; and how agent loops can be decomposed into inspectable planning, retrieval, tool-use, and termination stages. The coverage extends beyond implementation to durability, observability, testing, regression control, and production reliability, enabling readers to design systems that remain stable as data, prompts, packages, and operational conditions evolve.
Rather than presenting convenience abstractions as magic, the book treats them as design choices with measurable trade-offs. Readers should already be comfortable with Python, LLM application architecture, and retrieval-augmented generation concepts. In return, they gain a rigorous, architecture-first guide to building transparent, replayable, and maintainable LlamaIndex systems for serious production us