AI Agents With Python: Build, Evaluate, Secure, and Deploy Real AI Agents in Python - From Your First LLM Call to Production (AI Engineering) (en Inglés)
Reseña del libro "AI Agents With Python: Build, Evaluate, Secure, and Deploy Real AI Agents in Python - From Your First LLM Call to Production (AI Engineering) (en Inglés)"
Build AI Agents in Python — from your first LLM call to a real, deployable system. Includes complete source code on GitHub: https://github.com/masud-dot/ai-agents-with-python AI agents are more than chatbots with a clever prompt. A useful agent needs to reason about tasks, use tools, retrieve information, remember context, handle failures, protect sensitive operations, and produce results you can evaluate and trust. AI Agents with Python: From Beginner to Production teaches these capabilities step by step, using Python and practical projects instead of hiding the important machinery behind framework abstractions. You will start with the fundamentals of LLM-powered applications, then build the core pieces of an agent yourself before exploring frameworks and production patterns. By understanding what happens underneath the abstractions, you can make better architectural decisions and troubleshoot systems when they fail. Inside this book, you will learn how to: • Understand what makes an AI agent different from a script, workflow, or chatbot • Connect Python applications to large language models and work with structured output • Build reliable tool-calling agents and control their execution • Design memory for conversations, tasks, and long-term preferences • Create embeddings, semantic search, vector stores, and RAG pipelines • Build knowledge agents that answer from retrieved sources with citations • Connect agents to APIs, web services, and read-only databases safely • Understand and implement the Model Context Protocol (MCP) • Design multi-step planning and human-in-the-loop approval workflows • Compare single-agent, workflow, and multi-agent architectures • Build supervisor/worker and other multi-agent coordination patterns • Evaluate agent behavior with measurable tests, datasets, and regression gates • Protect agents against prompt injection, excessive agency, and unsafe actions • Add observability, tracing, cost controls, caching, and model tiering • Package, serve, containerize, and deploy an agent with Python, FastAPI, and Docker Learn by building. The book includes eight progressive projects, each introducing a real agent capability, followed by Atlas — a complete research-and-reporting agent that brings the concepts together. Atlas can plan research, retrieve information from documents and the web, optionally query structured data, remember conversation context, produce grounded reports with citations, and incorporate evaluation, security, observability, cost controls, and deployment practices. The GitHub repository includes: • Complete source code for Projects 1–8 • The complete Atlas capstone implementation • Reusable shared AI-agent components • Evaluation and testing examples • Project-specific README files and setup guidance • Configuration examples for running the projects safely Who is this book for? This book is designed for Python developers, software engineers, AI/ML practitioners, automation engineers, data engineers, and technical learners who understand basic Python and want to move from experimenting with LLMs to building real AI-agent systems. You do not need prior machine learning experience or advanced mathematics. Build the foundations. Understand the architecture. Create real AI agents.
