Agent building with Google Cloud
1. Agent building with Google Cloud
SPEAKER: Earlier, you learned about the evolution of GenAI applications, from chatbots to AI agents, and then to agentic AI. You also delved into an AI agent, how it works, with three components-- model, tools, and orchestration layer. Now it's time for a practical guide-- how to create an AI agent on Google Cloud. Google's GenAI architecture, as discussed at the beginning of this module, provides a comprehensive suite of AI agent tools and products. These offerings span foundation models, development tools, and application products. From the bottom up, Vertex AI Model Garden provides access to a wide range of Google's foundation models and third-party generative AI models, serving as the brain for your agent. Moving up to the development layer, a developer can build an AI agent using Vertex AI Agent Builder with either low-code or pro-code options. This platform offers tools such as Agent Development Kit, or ADK, Agent Engine, and Agent Garden, along with a managed environment to simplify agent development. It empowers developers, including software and ML engineers, to build AI agents end-to-end, from design to deployment. The top layer is the application or solution layer, designed for business users or analysts who want to build an AI agent with no code or minimum code. Gemini Enterprise is the hub for hosting and building AI agents to solve customized business problems. You can also rely on the Customer Engagement Suite to build conversational agents like customer service chatbots. How exciting. You're probably thrilled by the capabilities of AI agents, yet perhaps a bit overwhelmed by the sheer number of tools available. The big question is, how do you use them efficiently to build an AI agent? This is the exact challenge Bea, Ann, and Ian faced when they set out to build an AI agent to automate insurance quotes for their customers. Where to start? Let's dive deeper into how to use these different tools from an AI agent builder's perspective. Product choices and development journeys are influenced by two key factors, ease of use and flexibility. Ease of use ranges from no-code, ready-to-go solutions like Gemini Enterprise, progressing to more comprehensive code-based solutions for building your own. Flexibility varies from minimal configuration, like Gemini Enterprise, to full customization and high flexibility, like build your own. Let's focus on Gemini Enterprise and Vertex AI Agent Builder, as these are the tools most people interact with. Bea, Ann, and Ian want to prototype their ideas using an AI agent to automate insurance quotes. They turn to Gemini Enterprise, a user-friendly, no-code application designed for business users and analysts. Imagine a team of expert AI agents, each ready to tackle a specific business challenge, just like hiring a travel agent or general contractor. That's Gemini Enterprise. This powerful enterprise platform unites AI, Google-quality search, and your company's proprietary data in one secure space. Breaking down data silos, Gemini Enterprise makes your enterprise knowledge readily accessible and actionable in a secure manner. A core capability of Gemini Enterprise is its intelligent, multimodal search. It allows employees to quickly find contextual, cited answers from all organizational data formats-- documents, images, videos-- within a single interface, streamlining research and boosting productivity. Beyond search, Agentspace also provides a hub for AI agents that can automate workflows and complete tasks on behalf of users. It offers pre-built Google agents and a no-code agent designer for creating custom agents tailored to specific business needs. These agents can perform actions like updating project tickets, sending emails, or analyzing reports, reducing manual overhead and freeing up time for more strategic work. Among the powerful AI agents in Google Agentspace, NotebookLM stands out. It's your personal AI research assistant, a learning companion and study tutor designed to quickly help you understand and gain insights from vast amounts of information. Let's look at a demo on how to get started with NotebookLM, use it as your personal research assistant, and leverage it to discover insights and generate content. Welcome to NotebookLM, where you can tap into the power of generative AI. To begin, simply select the Create button to start a new notebook. The first step is to add your source documents. NotebookLM is flexible and supports a wide range of formats, including PDFs, text, markdown, and audio files. You can also pull documents directly from Google Drive, add links from websites or YouTube, or paste in text. For example, you could upload a report from McKinsey and a white paper from Google Cloud to create a new intelligent workspace. Once your documents are uploaded, you can start a conversation in the text box at the bottom. Think of it like talking to a super smart research assistant that uses only the documents you've provided for context. You can ask NotebookLM to perform a variety of tasks, such as creating a concise summary of your source materials with easy-to-reference citations. You can also ask specific questions, give it creative assignments, or have it summarize highly complex information. The right-hand Studio panel offers even more ways to work with your documents. In the Audio Overview section, NotebookLM can generate a podcast from two hosts based on your source documentation. In the Notes section, it can automatically create a study guide, a briefing document, an FAQ, or even a timeline. With these powerful features, it's your turn to explore and unlock the potential of your own data with NotebookLM. While Gemini Enterprise and NotebookLM are impressive for building AI agents, they fall short when it comes to tailoring agents for unique business needs. Bea, Ann, and Ian require an agent that can seamlessly integrate with their company's legacy applications and internal databases. They also need specific response behaviors, such as adhering to industry policies for claim reports while maintaining a warm and empathetic tone for customer communications. Gemini Enterprise couldn't accommodate these custom requirements, highlighting the need for tools like Vertex AI Agent Builder. Let's explore the Vertex AI components that facilitate end-to-end AI agent development from design to deployment. Vertex AI Agent Garden, a resource offering agent samples, customizable blueprints, and source code for various use cases, like data analysis and customer service, accelerating agent development so you don't have to start from scratch. Vertex AI Agent Development Kit, or ADK, an open source framework for developers seeking more control over agent logic, enabling the creation of production-ready agents using Python and seamless integration with Google Cloud Services. It's the agent version of an SDK, Software Development Kit. Feel free to check out ADK Quickstart in the Reading List. Vertex AI Agent Engine, a fully managed runtime for simplified deployment, scaling, and monitoring of agents in production, handling infrastructure so you can focus on capabilities. To summarize, a simplified decision tree can help navigate Google's AI agent ecosystem, including tools and products. Please note this is only a brief guide and may change as products evolve. Out-of-the-box solutions-- Gemini Enterprise and conversational agents are ideal for no-code deployment. This path is best for business users who need ready-to-use solutions with minimal setup. Customizable templates-- Agent Garden with Agent Builder offers pre-built sample agents that can be modified. This low-code approach is best for data scientists and analysts who need a starting point for creating more tailored solutions. Custom development-- the ADK or building your own solution from scratch is the pro-code path. This approach offers maximum flexibility and is best for developers like software and ML engineers who need to build highly tailored agents with complex logic and deep integrations. Armed with these amazing tools, you can now create your very own AI agent. Explore NotebookLM-- notebookLM.google.com. Now completely free for the general public, and find your AI research assistant, learning companion, or even tutor. Have fun.2. Let's practice!
Create Your Free Account
or
By continuing, you accept our Terms of Use, our Privacy Policy and that your data is stored in the USA.