دانلود کتاب سیستم‌های عامل هوش مصنوعی چندوجهی و بلادرنگ

You’ve likely seen the term ‘AI Agent’ popping up everywhere, often described as the next big thing in artificial intelligence. But what does that term actually mean? Beyond the hype, how does an agent really work “under the hood”? This chapter is designed to answer those questions by taking you on a practical journey from the ground up. We’ll begin by exploring the five distinct stages of agent evolution, starting with the foundational ideas from the pre-generative AI era. We’ll then move through the rise of Foundation Models, the addition of knowledge with RAG, and the birth of agents that can act. With that historical context, we’ll establish a new, robust definition of what an agent truly is, moving beyond a simple buzzword to a concrete engineering concept. Once we have our definition, we’ll get our hands dirty. We’ll walk you through the core components of an agent and show you how to build the logic for one from scratch. Finally, we’ll scale up, first by giving agents memory for complex, multi-turn conversations, and then by seeing how the field has come full circle to orchestrate powerful multi-agent systems, fulfilling a vision that started decades ago. The way we interact with artificial intelligence is undergoing a fundamental shift. We’re moving beyond AI as a passive chatbot that simply answers our questions and toward AI as an active partner that gets things done on our behalf. This evolution from a simple conversationalist to a complex actor is what defines the exciting new world of AI Agents. As we’ll explore, and as you can see in Figure 1-1, this journey has five distinct stages. We began with legacy agents before the birth of Generative AI, moved on to powerful Foundation Models (FMs) that respond to prompts, enhanced them with specialized knowledge using Retrieval Augmented Generation (RAG), and arrived at true Agents that can take action, which are now being orchestrated into collaborative multi-agent systems. Let’s walk through each step.