Kuligin, Leonid / Zaldívar, Jorge / Tschochohei, Maximilian Generative AI on Google Cloud with LangChain

Design scalable generative AI solutions with Python, LangChain, and Vertex AI on Google Cloud
1. Auflage 2025
ISBN: 978-1-83588-933-6
Verlag: De Gruyter
Format: EPUB
Kopierschutz: 0 - No protection

E-Book, Englisch, 308 Seiten

The rapid transformation and enterprise adoption of GenAI has created an urgent demand for developers to quickly build and deploy AI applications that deliver real value. Written by three distinguished Google AI engineers and LangChain contributors who have shaped Google Cloud's integration with LangChain and implemented AI solutions for Fortune 500 companies, this book bridges the gap between concept and implementation, exploring LangChain and Google Cloud's enterprise-ready tools for scalable AI solutions.
You'll start by exploring the fundamentals of LLMs and how LangChain simplifies the development of AI workflows by connecting LLMs with external data and services. This book guides you through using essential tools like the Gemini and PaLM 2 APIs, Vertex AI, and Vertex AI Search to create sophisticated, production-ready GenAI applications. You'll also overcome the context limitations of LLMs by mastering advanced techniques like RAG and external memory layers.
Through practical patterns and real-world examples, you'll gain everything you need to harness Google Cloud's AI ecosystem, reducing the time to market while ensuring enterprise scalability. You'll have the expertise to build robust GenAI applications that can be tailored to solve real-world business challenges.

*Email sign-up and proof of purchase required

Autoren/Hrsg.

Kuligin, Leonid, Leonid Kuligin

Zaldívar, Jorge, Jorge Zaldívar

Tschochohei, Maximilian, Maximilian Tschochohei

Chase, Harrison

Weitere Mitwirkende

Chase, Harrison

Fachgebiete

• Mathematik | Informatik EDV | Informatik Informatik Künstliche Intelligenz

Leseproben

Preface

Generative AI is on the rise, and its enterprise adoption is growing quickly. Developers need to understand the new technology, and they’re requested to deliver business value quickly and reduce time to market. That’s where LangChain as a framework for the quick development of generative AI applications comes in, together with enterprise-ready and highly scalable solutions from Google Cloud that expose foundational models, vector search, and other capabilities required by such applications.

In this book, we’ll explore the basics of the LangChain framework and its core interfaces and then we’ll start building our applications on Google Cloud using the Gemini model family and Vertex AI platform. We’ll learn how to compose generative AI workflows, access external knowledge, and chain large language models (LLMs) to solve specific domain problems. You’ll learn about some commonly used patterns and techniques such as retrieval-augmented generation (RAG), ways to process long documents that don’t fit into the context of the LLM, implementing external memory layers, using third-party APIs to enhance LLMs’ capabilities, and developing agentic workflows.

Who this book is for

If you’re an application developer who is enthusiastic about generative AI, or you’re an experienced ML engineer new to LangChain or Google Cloud, this book is for you. The book is also excellent for professionals who haven’t used generative AI on Google Cloud and would like to learn how to navigate across the landscape of enterprise products for generative AI developers offered by Google.

What this book covers

, , introduces LangChain and explains how LangChain orchestrates LLMs and coordinates the execution of complex tasks. It also examines LangChain primitives and explains how they can be composed together using the LangChain Expression Language (LCEL).

, , guides you on how to select and implement the right Google Cloud foundational model in LangChain for many common use cases.

, , focuses on the problem of hallucinations in LLMs and how to address them with RAG. It describes the components of a RAG system and introduces Vertex AI Agent Builder, a managed Google Cloud service for building RAG-based applications.

, , explores the architecture of a vector search pipeline and discusses different searching techniques. It showcases how the combined power of Google Cloud and LangChain can be harnessed to develop search components for RAG applications.

, , explains the range of document parsing capabilities available within Google Cloud and LangChain. It demonstrates the use of LangChain document loaders, Document AI, and Vertex AI’s Agent Builder for ingesting documents of a variety of formats.

, , discusses multimodal LLMs, how to compose a multimodal input with LangChain, and advanced methods of building multimodal RAGs to enhance their capabilities.

, , teaches you how to use LangChain to summarize documents, including how to handle long documents and different modalities such as audio and video, as well as how to efficiently implement question-and-answering on long documents.

, , guides readers on how to build chatbots with LangChain, covering conversation engineering principles, memory implementation, intent routing, and integration with RAG.

, , explores how to improve the reasoning capabilities of LLMs through the use of tools. It introduces the concept of tool calling and examines different ways to construct tools programmatically with LangChain. It also discusses ReAct – one of the foundational patterns in the multi-step reasoning process.

, , explains the concept and components of an agent and gives examples of how to build one with two different methods: LangChain on Google Cloud with the Gemini SDK and Vertex AI Agent Builder.

, , delves into the fundamentals of agentic architectures and shows two examples of real-world architectures: agentic RAG and natural language to SQL using LangChain, LangGraph, and Google Cloud.

, , emphasizes the importance of evaluating generative AI applications to ensure they meet product requirements and quality expectations. It discusses the difference between traditional evaluation and where generative AI evaluation is similar or different. It also introduces LangSmith, a tool from LangChain that helps developers trace and evaluate generative AI applications, and explains how to use Vertex AI evaluation, a managed service for evaluating generative AI applications on Google Cloud.

, , walks you through designing and building generative AI systems, addressing challenges such as non-deterministic outcomes and rapid technology evolution while emphasizing responsible AI principles.

, , discusses the principles behind generative AI and LLMs. It explains very briefly how these models are trained, their unique characteristics, and the importance of AI alignment and discusses key criteria for a successful proof of concept (POC) in generative AI.

, , provides a foundational guide to Google Cloud for those new to the platform, covering organization setup, user management, billing, networking, and AI/ML development environments, enabling you to effectively leverage Google Cloud for LangChain applications.

To get the most out of this book

In order to learn from this book, you should have a basic knowledge of Python, minimal experience at the beginner level in machine learning, and some hands-on experience in generative AI. You need to have a Google Cloud project with enabled Vertex AI APIs. For most of the examples provided in this book, your costs will be minimal (less than a few dozen dollars). The only expensive workload is developing your own RAG on a massive amount of PDFs using Vertex Agent Builder and Vertex Vector Search, but you can skip those examples if you don’t need the scale Google Cloud provides for these use cases.

We also encourage you to register and get an API key on LangSmith (but the free tier would be enough for running through this book).

Software/hardware covered in the book

Operating system requirements

Python with the required libraries installed

Windows, macOS, or Linux and ChromeOS

Vertex AI API on Google Cloud

A custom Google Search enabled and a Google Cloud API key

Vertex Vector Search, Vertex Agent Builder

LangSmith (free tier)

We encourage you to create a Python virtual environment (using Python venv or conda) and install all the dependencies in your virtual environment with pip or conda install.

If you are using the digital version of this book, we advise you to type the code yourself or access the code from the book’s GitHub repository (a link is available in the next section). Doing so will help you avoid any potential errors related to the copying and pasting of code.

Download the example code files

You can download the example code files for this book from GitHub at https://github.com/PacktPublishing/Generative-AI-on-Google-Cloud-with-LangChain. If there’s an update to the code, it will be updated in the GitHub repository.

Discussions

We encourage readers to take advantage of the Discussions feature in the repository to have conversations, ask...

Über Autor(innen)

Kuligin Leonid :

Leonid Kuligin is a staff AI engineer at Google Cloud, working on generative AI and classical machine learning solutions (such as demand forecasting or optimization problems). Leonid is one of the key maintainers of Google Cloud integrations on LangChain, and a visiting lecturer at CDTM (TUM and LMU). Prior to Google, Leonid gained more than 20 years of experience in building B2C and B2B applications based on complex machine learning and data processing solutions such as search, maps, and investment management in German, Russian, and US technological, financial, and retail companies.Zaldívar Jorge :

Jorge Zaldivar is an AI Engineer at Google and also a contributor to LangChain's integrations with Google. He has a decade of experience building complex Machine Learning applications and products applied to the energy and financial industries.Tschochohei Maximilian :

Maximilian Tschochohei leads AI engineering at Google Cloud Consulting EMEA. He implements LangChain applications together with Google Cloud customers. Before Google, he worked in strategy and technology consulting with Boston Consulting Group.