BUILDING AI AGENTS WITH LLMS, TOOLS, AND MEMORY: A Hands-On Python Guide to Agentic AI, Function Calling, RAG, and Multi-Agent Workflows - Couverture souple

Thornqvist, Leif

Couverture souple

ISBN 13 : 9798166096166

Editeur : Independently published, 2026

Synopsis

Building an AI agent is not just about making a model respond. It is about creating a system that can act within clear, testable limits.

A tool-using application must decide what information to retrieve, what to remember, which action is permitted, and how to respond when an API fails. A fluent answer cannot confirm that a task was completed; a valid tool call cannot replace permission checks; and longer workflows introduce new risks around state, cost, and reliability.
This hands-on Python guide shows how to move from basic language-model integration to bounded, evidence-aware agent systems. It explains the architecture behind each capability and provides examples, exercises, and complete projects that connect the design choices to working code.
Learn how to:

• Create controlled reasoning-and-action loops with explicit stopping rules.
• Build structured function calls and validate tools, arguments, and outcomes.
• Ground responses in retrieved documents using retrieval-augmented generation.
• Manage context and persistent memory while tracking relevance and provenance.
• Plan multi-step tasks, checkpoints, approval gates, and specialist-agent handoffs.
• Evaluate, secure, monitor, and prepare agent applications for deployment.

Starting with agent fundamentals and Python model interfaces, the book covers tool execution, errors and retries, document search, embeddings, memory, planning, multi-agent coordination, generate–verify–revise patterns, testing, observability, security, scaling, and cost control. The final projects explore a knowledge assistant, a memory-enabled workflow assistant, and a coordinated research agent.
Ideal for developers and technically curious learners with basic Python experience, this book focuses on using existing models in practical software. Machine-learning expertise is not required. Simulated examples support reproducible learning; production integrations still need current documentation, testing, and appropriate permissions.

Get your copy and start developing AI agents with clearer architecture, better evidence, and deliberate control over what they do.

Détails de l'édition

• Éditeur: Independently published
• Date d'édition: 2026
• Langue: anglais
• ISBN 13: 9798166096166
• Reliure: Broché
• Nombre de pages: 215
• Coordonnées du fabricant: non disponible
• Personne responsable: non disponible