book

AI Agents in Practice

by Valentina Alto

August 2025

Intermediate to advanced

282 pages

6h 42m

English

Packt Publishing

Read now

Unlock full access

Subscribe for a Free eBook
Who this book is forWhat this book coversTo get the most out of this bookDisclaimer on AI usageGet in touchJoin our Discord and Reddit spaceYour Book Comes with Exclusive Perks – Here’s How to Unlock ThemUnlock Your Book’s Exclusive BenefitsHow to unlock these benefits in three easy steps
Technical requirementsUnderstanding foundation models and the rise of LLMsFrom narrow AI to foundation modelsUnder the hood of an LLMHow do we consume LLMs?Latest significant breakthroughsSmall language models and fine-tuningModel distillationReasoning modelsDeepSeekRoad to AI agentsText generationChat with your dataMultimodalityThe need for an additional layer of intelligence: introducing AI agentsSummaryReferences
Technical requirementsEvolution of agents from RPA to AI agentsComponents of an AI agentDifferent types of AI agentsRetrieval agentsTask agentsAutonomous agentsSummaryReferencesSubscribe for a Free eBook
Introduction to AI orchestratorsAutonomyAbstraction and modularityCore components of an AI orchestratorWorkflow managementMemory and context handlingTool and API integrationError handling and monitoringSecurity and complianceOverview of the most popular AI orchestrators in the marketHow to choose the right orchestrator for your AI agentSummaryReferences
Different types of memoryShort-term memoryLong-term memorySemantic memoryEpisodic memoryProcedural memoryIn between short-term memory and long-term memory – the role of semantic cachesManaging context windowsStoring, retrieving, and refreshing memoryTemporal and spatial reasoning in AI agentsPopular tools to manage memoryLangMemMem0LeTTA (formerly MemGPT)SummaryReferencesSubscribe for a Free eBook
Technical requirementsThe anatomy of an AI agent’s toolsHardcoded and semantic functionsHardcoded functionsSemantic functionsAPIs and web servicesWeb APIsInternal or enterprise APIsBackend function APIs (service mesh or microservices)Serverless functions/Lightweight APIsDatabases and knowledge basesStructured dataUnstructured dataSynchronous versus asynchronous callsSummaryReferences
Technical requirementsIntroduction to the LangChain ecosystemBuild – the architectural foundationRun – the operational layerManage – observability and iterationOverview of out-of-the-box componentsUse case – e-commerce AI agentScenario descriptionAskMamma’s building blocksDeveloping the agentObservability, traceability, and evaluationInfusing the AI agent in the mobile appSummaryReferencesSubscribe for a Free eBook

Technical requirementsIntroduction to multi-agent systemsUnderstanding and designing different workflows for your multi-agent systemOverview of multi-agent orchestratorsAutoGenTaskWeaverOpenAI Agents SDKLangGraphBuilding your first multi-agent application with LangGraphSummaryReferences
Technical requirementsWhat is a protocol?Understanding the Model Context ProtocolAgent2AgentAgent Commerce ProtocolToward an agentic webFrom the traditional web to the agentic webKey components of NLWebCurrent progress and applicationsSummaryReferencesSubscribe for a Free eBook
Ethical challenges in AI – fairness, transparency, privacy, and accountabilityFairness and biasTransparency and explainabilityPrivacy and data protectionAccountability and liabilitySafety and reliabilityAgentic AI autonomy and its unique ethical challengesAutonomy versus human controlDeception and manipulationUnintended consequences and liability for agent actionsResponsible AI principles and practicesFrom principles to practiceGuardrails for safe and ethical AIWhat are AI guardrails?Content filtering and moderation in AI systemsWhy content filtering is neededHow content filtering worksEthical considerations in content moderationAddressing the challenges: governance, regulations, and collaborationOrganizational governance and cultureIndustry collaboration and self-regulationGovernment regulations and policySummaryReferences
Join our Discord and Reddit space

Content preview from AI Agents in Practice

5 The Need for Tools and External Integrations

As we mentioned in the preceding chapters, one of the key features and differentiations of AI agents is that they can interact with the world. While LLMs can understand, reason, and generate text, they are ultimately limited by what exists within their training data and current context window. To go beyond passive conversation and perform real, useful actions—such as booking appointments, querying databases, retrieving live information, or executing multi-step workflows—AI agents must be equipped with tools.

Tools are the functional extensions of an agent’s intelligence. They allow agents to call APIs, access external systems, retrieve fresh data, and even manipulate structured knowledge bases. ...