e Cloud By Richard Seroter New and shiny Your agents got an upgrade with the new Vertex AI Agent Engine Memory Bank. Our interactions with agents may span sessions and extend for some time. You’ve had to solve this with your own state solutions. But with the new Memory Bank, you get a managed service that “just works” for agents running in Vertex AI Agent Engine. See the blog post and docs for more. BigQuery gets some powerful new aggregation functions. This won’t appeal to all of you, but to those dealing with rich data sets, this might be a BIG deal. These three new categories of aggregation functions —group by extension, user-defined aggregation functions, and approximate aggregate functions—may introduce massive performance and efficiency improvements. Cloud Run introduces worker pools and new Docker integrations. How do you deploy a background task that continuously chomps on work? Now you can run these in Cloud Run worker pools. It’s great for processing data from Apache Kafka, and we also introduced a new open source Cloud Run Kafka Autoscaler. We also shared new integrations that make it easier to deploy Docker Compose definitions to Cloud Run. Dig into all this new functionality. Watch this Take a fun tour through agent building tech. Here’s an excellent video that uses great imagery and code to clearly explain concepts like agents, tools, and MCP. A captured graphic of a Google Cloud video where Megan is explaining the types of tools that ADK provides to agent builders. Community cuts Send emails from Google Cloud using the Application Integration service. It’s true we don’t offer a built-in managed email service. But you can still send emails directly from your cloud apps using the Send Mail task included in the Application Integration service. Learn more in this blog post. Which service mesh is right for you? It’s the age-old question: managed service or run it yourself? We give you both options in Google Cloud. Rahul and Shrey compare the managed and unmanaged flavors of Google Cloud Service Mesh. Building and evaluating ADK agents. Our friends at W&B wrote up a good look at the Agent Development Kit and how you can trace their behavior. Learn and grow Try out Cloud Run worker pools by running this code lab. We recently added this functionality to deploy continuously-processing background workers to Cloud Run. This code lab walks you through creating and using one as a Cloud Pub/Sub reader. Build intelligent agents with the Agent Development Kit (ADK). Today, we published a new course in Google Cloud Skills Boost for developers ready to start building their own agents with ADK. It's packed with hands-on labs, core concepts, real-world applications, and resources from Google Developer Experts (GDEs) and the broader developer community. Protect your (BigQuery) data from eager AI agents. Here’s a great post with good tips for creating lines of defense around your BigQuery environment to ensure agents don’t do too much, at too high a cost. Get smart(er) about Google Cloud networking. It’s hard to be good at Cloud without some working knowledge of networking configurations. This quick reference guide points you to assets that help you learn more. Set up a GraphRAG setup with Vertex AI and Spanner Graph. If you’ve got graph data that you’d like to ground your AI applications with, you might explore the GraphRAG option. I like this new comprehensive architecture guide that helps you configure this on Google Cloud. Building production-ready AI Agents? The Agent Factory video podcast dives deep into design, build, & deployment on Google Cloud. The podcast covers code, architecture , & practical tips. Subscribe now! ➡️ https://goo.gle/theagentfactory One more thing Batch mode comes to the Gemini API. Large jobs where you don’t need an immediate response? The new Gemini API Batch mode returns results within 24 hours, and is half the cost. An X post from Logan announcing the new Batch mode in the Gemini API Follow us on Social media. Add your voice to the conversation and get the latest news, views, and insights. X Google Cloud blog LinkedIn Facebook YouTube GitHub No-cost cloud training 35 credits on Google Cloud Skills Boost. Redeem now Already on Google Cloud? Go to your console Google LLC 1600 Amphitheatre Parkway, Mountain View, CA 94043 This email was sent to pfeifferj@archlinux.ch because you indicated that you want news and program announcements from the Google Cloud Innovators program. If you don't want to receive such emails in the future, please unsubscribe. You can also change your preferences by visiting https://developers.google.com/profile/u/me/my-community and editing your selection under “community email preferences” for Google Cloud Innovators.The overwhelmed person’s guide to Google Cloud: July 22Google Cloud Innovators undefinedpfeifferj@archlinux.ch pfeifferj@archlinux.ch