book

AI Agents and Applications

Name: AI Agents and Applications
Author: Roberto Infante
ISBN: 9781633436541

by Roberto Infante

March 2026

Intermediate to advanced

448 pages

12h 44m

English

Manning Publications

Read now

Unlock full access

AI Agents and Applications
copyright
contents
dedication
preface
acknowledgments
about this book
about the author
about the cover illustration
Part 1 Getting started with LLMs

1 Introduction to AI agents and applications
1.1 Building LLM-based applications and agents1.1.1 LLM-based applications: Summarization and Q&A engines1.1.2 LLM-based chatbots1.1.3 AI agents1.2 Introducing LangChain1.2.1 LangChain architecture1.2.2 LangChain’s core object model1.3 Typical LLM use cases1.4 How to adapt an LLM to your needs1.4.1 Prompt engineering1.4.2 RAG1.4.3 Fine-tuning1.5 Which LLMs to choose1.6 What you’ll learn from this book
2 Executing prompts programmatically
2.1 Running prompts programmatically2.1.1 Setting up the environment for this chapter2.1.2 Minimal prompt execution2.2 Running prompts with LangChain2.3 Prompt templates2.3.1 Implementing a prompt template with a Python function2.3.2 Using LangChain’s PromptTemplate2.4 Prompt types2.4.1 Text classification2.4.2 Sentiment analysis2.4.3 Text summarization2.4.4 Composing text2.4.5 Question answering2.4.6 Reasoning2.5 Reasoning in detail2.5.1 One-shot learning2.5.2 Two-shot learning2.5.3 Providing steps2.5.4 Few-shot learning2.5.5 Implementing few-shot learning with LangChain2.5.6 Chain of Thought2.6 Prompt structure
Part 2 Summarization
3 Summarizing text using LangChain
3.1 Summarizing a document bigger than the context window3.1.1 Chunking the text into Document objects3.1.2 Split3.1.3 Map3.1.4 Reduce3.1.5 MapReduce combined chain3.1.6 MapReduce execution3.2 Summarizing across documents3.2.1 Creating a list of Document objects3.2.2 Wikipedia content3.2.3 File-based content3.2.4 Creating the Document list3.2.5 Progressively refining the final summary3.3 Summarization flowchart
4 Building a research summarization engine
4.1 Overview of a research summarization engine4.2 Setting up the project4.3 Implementing the core functionality4.3.1 Implementing web searching4.3.2 Implementing web scraping4.3.3 Instantiating the LLM client4.3.4 JSON to Python object converter4.4 Enhancing the architecture with query rewriting4.5 Prompt engineering4.5.1 Crafting web search prompts4.5.2 Crafting summarization prompts4.5.3 Research report prompt4.6 Initial implementation4.6.1 Importing functions and prompt templates4.6.2 Setting constants and input variables4.6.3 Instantiating the LLM client4.6.4 Generating the web searches and collecting the results4.6.5 Scraping the web results4.6.6 Summarizing the web results4.6.7 Generating the research report4.7 Reimplementing the research summary engine in LCEL4.7.1 Assistant Instructions chain4.7.2 Web Searches chain4.7.3 Search and Summarization chain4.7.4 Web Research chain
5 Agentic workflows with LangGraph
5.1 Understanding agentic workflows and agents5.1.1 Workflows5.1.2 Agents5.1.3 When to use agent-based architectures5.1.4 Agent development frameworks5.2 LangGraph basics5.3 Moving from LangChain chains to LangGraph5.4 LangGraph core components5.4.1 StateGraph structure5.4.2 State management and typing5.4.3 Node functions and edge definitions5.4.4 Entry points and end conditions5.5 Turning the web research assistant into an AI agent5.5.1 Original LangChain implementation overview5.5.2 Identifying components for conversion5.5.3 Step-by-step transformation process5.5.4 Code comparison and benefits realized
Part 3 Q&A chatbots
6 RAG fundamentals with ChromaDB
6.1 Semantic search6.1.1 A basic Q&A chatbot over a single document6.1.2 A more complex Q&A chatbot over a knowledge base6.1.3 The RAG design pattern6.2 Vector stores6.2.1 What’s a vector store?6.2.2 How do vector stores work?6.2.3 Vector libraries vs. vector databases6.2.4 Most popular vector stores6.2.5 Storing text and performing a semantic search using Chroma6.3 Implementing RAG from scratch6.3.1 Retrieving content from the vector database6.3.2 Invoking the LLM6.3.3 Building the chatbot6.3.4 Recap of RAG terminology
7 Q&A chatbots with LangChain and LangSmith
7.1 LangChain object model for Q&A chatbots7.1.1 Content ingestion (indexing) stage7.1.2 Q&A (retrieval and generation) stage7.2 Vector store content ingestion7.2.1 Splitting and storing the documents7.2.2 Removing duplication7.2.3 Ingesting multiple documents from a folder7.3 Q&A across stored documents7.3.1 Querying the vector store directly7.3.2 Asking a question through a LangChain chain7.3.3 Completing the RAG chain setup7.3.4 Follow-up question7.4 Chatbot memory of message history7.4.1 Amending the prompt7.4.2 Updating the chat message history7.4.3 Feeding the chat history to the RAG chain7.4.4 Putting everything together7.5 Tracing execution with LangSmith
Part 4 Advanced RAG
8 Advanced indexing
8.1 Improving RAG accuracy8.1.1 Content ingestion stage8.1.2 Question-answering stage8.2 Advanced document indexing8.3 Splitting strategy8.3.1 Splitting strategies8.3.2 Factors to consider8.3.3 Choosing the right strategy8.3.4 Splitting by HTML header8.4 Embedding strategy8.4.1 Embedding child chunks with ParentDocumentRetriever8.4.2 Embedding child chunks with MultiVectorRetriever8.4.3 Embedding document summaries8.4.4 Embedding hypothetical questions8.5 Granular chunk expansion8.6 Semi-structured content8.7 Multimodal RAG
9 Question transformations
9.1 Rewrite-Retrieve-Read9.1.1 Retrieving content using the original user question9.1.2 Setting up the query rewriter chain9.1.3 Retrieving content with the rewritten query9.1.4 Combining everything into a single RAG chain9.2 Generating multiple queries9.2.1 Setting up the chain for generating multiple queries9.2.2 Setting up a custom multi-query retriever9.2.3 Using a standard MultiQueryRetriever instance9.3 Step-back question9.3.1 Setting up the chain to generate a step-back question9.3.2 Incorporating step-back question generation into the RAG chain9.4 Hypothetical Document Embeddings (HyDE)9.4.1 Generating a hypothetical document for the user question9.4.2 Integrating the HyDE chain into the RAG chain9.5 Single-step and multi-step decomposition
10 Query generation, routing, and retrieval postprocessing
10.1 Content database query generation10.2 Self-querying (metadata query enrichment)10.2.1 Ingestion: Metadata enrichment10.2.2 Q&A on a metadata-enriched collection10.3 Generating a structured SQL query10.3.1 Installing SQLite10.3.2 Setting up and connecting to the database10.3.3 Generating SQL queries from natural language10.3.4 Executing the SQL query10.4 Generating a semantic SQL query10.4.1 Standard SQL query10.4.2 Semantic SQL query10.4.3 Creating the embeddings10.4.4 Performing a semantic SQL search10.4.5 Automating semantic SQL search10.4.6 Benefits of a semantic SQL search10.5 Generating queries for a graph database10.6 Chain routing10.6.1 Setting up data retrievers10.6.2 Setting up the query router10.6.3 Integrating the chain router into a full RAG chain10.7 Retrieval postprocessing10.7.1 Similarity postprocessors10.7.2 Keyword postprocessors10.7.3 Time weighting10.7.4 RAG fusion (Reciprocal Rank Fusion)
Part 5 AI agents
11 Building tool-based agents with LangGraph
11.1 Starting simple: Building a single-tool travel info agent11.1.1 Project setup11.1.2 Loading environment variables11.1.3 Preparing the travel information vector store11.2 Enabling agents to call tools11.2.1 From function calling to tool calling11.2.2 How tool calling works with LLMs11.2.3 Registering tools with the LLM11.2.4 Agent state: Tracking the conversation11.2.5 Executing tool calls11.2.6 The LLM node: Coordinating reasoning and action11.3 Assembling the agent graph11.4 Understanding the agent graph structure11.5 Running the agent chatbot: The Read-Eval-Print Loop11.6 Executing a request11.6.1 Step-by-step debugging11.7 Expanding your agent: Adding a weather forecast tool11.7.1 Implementing a mock weather service11.7.2 Creating the weather forecast tool11.7.3 Updating the agent for multi-tool support11.8 Executing the multi-tool agent11.8.1 Running the multi-tool agent (initial behavior)11.8.2 Improving LLM tool usage with system guidance11.9 Using prebuilt components for rapid development11.9.1 Refactoring to use the LangGraph ReAct agent11.9.2 Running the prebuilt agent11.9.3 Observing and debugging with LangSmith11.9.4 Enabling LangSmith tracing
12 Multi-agent systems
12.1 Building an accommodation booking agent12.1.1 Hotel booking tool12.1.2 B&B booking tool12.1.3 ReAct accommodation booking agent12.2 Building a router-based travel assistant12.2.1 Designing the router agent12.2.2 Routing logic12.2.3 Building the multi-agent graph12.2.4 Trying out the router agent12.3 Handling multi-agent requests with a Supervisor component12.3.1 The Supervisor pattern: An agent of agents12.3.2 From “one-way” to “return ticket” interactions12.3.3 Trying out the Supervisor agent
13 Building and consuming MCP servers
13.1 Introduction to MCP servers13.1.1 The problem: Context integration at scale13.1.2 The solution: The Model Context Protocol13.1.3 The MCP ecosystem13.2 How to build MCP servers13.2.1 Essential resources for MCP server development13.2.2 Official language-specific MCP SDKs13.2.3 Consuming MCP servers in LLM applications and agents13.3 Building a weather MCP server13.3.1 Implementing the MCP server13.3.2 Trying out the MCP server with MCP Inspector13.3.3 Consuming the MCP server from a test MCP host13.4 Integrating the Weather MCP tool into an agent13.4.1 Preparing the travel agent for live weather data13.4.2 Integrating the AccuWeather MCP tool13.4.3 Updating the agent chat loop13.4.4 Combining local and remote tools13.4.5 Testing and verification13.4.6 Using the agent for complex queries
14 Productionizing AI agents: Memory, guardrails, and beyond
14.1 Memory14.1.1 Types of memory14.1.2 Why short-term memory is needed14.1.3 Checkpoints in LangGraph14.1.4 Adding short-term memory to our travel assistant14.1.5 Executing the checkpointer-enabled assistant14.1.6 Rewinding the state to a past checkpoint14.2 Guardrails14.2.1 Implementing guardrails to reject nontravel-related questions14.2.2 Implementing more restrictive guardrails at the agent level14.3 Beyond this chapter14.3.1 Long-term user and application memory14.3.2 Human-in-the-loop14.3.3 Post-model guardrails14.3.4 Evaluation of AI agents and applications14.3.5 Deployment on the LangGraph platform and Open Agent Platform
appendix A Trying out LangChain
A.1 Trying out LangChain in a Jupyter Notebook environmentA.1.1 Sentence completion exampleA.1.2 Prompt engineering examplesA.1.3 Creating chains and executing them with LCEL
appendix B Setting up a Jupyter Notebook environment
B.1 Installing the Python interpreter or a Python distributionB.2 Creating a virtual environment with venv and upgrading pipB.3 Setting up a Jupyter Notebook
appendix C Choosing an LLM
C.1 Popular large language modelsC.1.1 OpenAI GPT seriesC.1.2 GeminiC.1.3 GemmaC.1.4 ClaudeC.1.5 CohereC.1.6 LlamaC.1.7 FalconC.1.8 MistralC.1.9 QwenC.1.10 GrokC.1.11 PhiC.1.12 DeepSeekC.2 How to choose a modelC.2.1 Model purposeC.2.2 Proprietary vs. open sourceC.2.3 Model size (number of parameters)C.2.4 Context window sizeC.2.5 Multilingual supportC.2.6 Accuracy vs. speedC.2.7 Cost and hardware requirementsC.2.8 Task suitability (standard benchmarks)C.2.9 Safety and BiasC.2.10 A practical exampleC.3 A word of caution
appendix D Installing SQLite on Windows
D.1 Installing SQLite
appendix E Open source LLMs
E.1 Benefits of open source LLMsE.1.1 TransparencyE.1.2 PrivacyE.1.3 Community drivenE.1.4 Cost savingsE.2 Popular open source LLMsE.3 Considerations on running open source LLMs locallyE.3.1 Limitations of consumer hardwareE.3.2 QuantizationE.3.3 OpenAI REST API compatibilityE.4 Local inference enginesE.4.1 llama.cppE.4.2 OllamaE.4.3 vLLME.4.4 llamafileE.4.5 LM StudioE.4.6 LocalAIE.4.7 GPT4AllE.4.8 Comparing local inference enginesE.4.9 Choosing a local inference engineE.5 Inference via the Hugging Face Transformers libraryE.5.1 Hugging Face Transformers libraryE.5.2 LangChain’s Hugging Face pipelineE.6 Building a local summarization engineE.6.1 Choosing the inference engineE.6.2 Starting up the OpenAI-compatible serverE.6.3 Modifying the original solutionE.6.4 Running the summarization engine through the local LLME.6.5 Comparison between OpenAI and local LLM

Content preview from AI Agents and Applications

6 RAG fundamentals with ChromaDB

This chapter covers

Implementing semantic search using the RAG architecture
Understanding vector stores and their functionality
Implementing RAG with ChromaDB and OpenAI

In this chapter, we’ll explore two key concepts: semantic search and Retrieval Augmented Generation (RAG). You’ll see how large language models (LLMs) are used for semantic search through a chatbot, enabling you to query a system for information across multiple documents and retrieve the fragments that best match the meaning of your question, rather than just matching keywords. This approach is also known as Q&A over documents or querying a knowledge base.

In earlier chapters, you learned about summarization, a typical use case for ...

Become an O’Reilly member and get unlimited access to this title plus top books and audiobooks from O’Reilly and nearly 200 top publishers, thousands of courses curated by job role, 150+ live events each month,
and much more.

Read now

Unlock full access

More than 5,000 organizations count on O’Reilly

O’Reilly covers everything we've got, with content to help us build a world-class technology community, upgrade the capabilities and competencies of our teams, and improve overall team performance as well as their engagement.

Julian F.

Head of Cybersecurity

I wanted to learn C and C++, but it didn't click for me until I picked up an O'Reilly book. When I went on the O’Reilly platform, I was astonished to find all the books there, plus live events and sandboxes so you could play around with the technology.

Addison B.

Field Engineer

I’ve been on the O’Reilly platform for more than eight years. I use a couple of learning platforms, but I'm on O'Reilly more than anybody else. When you're there, you start learning. I'm never disappointed.

Amir M.

Data Platform Tech Lead

I'm always learning. So when I got on to O'Reilly, I was like a kid in a candy store. There are playlists. There are answers. There's on-demand training. It's worth its weight in gold, in terms of what it allows me to do.

Mark W.

Embedded Software Engineer

Publisher Resources

ISBN: 9781633436541Publisher Support Other Publisher Website Purchase Link

Cloud Computing

Data Engineering

Data Science

AI & ML

Programming Languages

Software Architecture

IT/Ops

Security

Design

Business

Soft Skills

AI Agents and Applications

by Roberto Infante