Google Cloud strengthened its artificial intelligence tools for databases and developers today with the addition of its most powerful generative AI model Gemini that will provide integrations for the analysis of data and the creation of AI agents.

Today at Google Cloud Next ’24, the company announced updates to how databases and data are managed with the addition of Gemini in Databases, part of Gemini for Google Cloud. This will deliver a powerful AI-based assistant that will work with every aspect of database creation and management to simplify everything from migration to application development.

The new AI assistant focuses on giving database operators powerful management across three different interfaces including code development within Database Studio, fleet management with Database Center and assisted migrations with Database Migration Service.

Database Studio is a rich SQL editor console that allows developers to quickly generate and summarize SQL code with an assistant directly in the editor. Database Center provides security, management and intelligent dashboards for monitoring the health of multiple databases. The Database Migration Service acts as a one-stop-shop where Gemini can help examine and convert any database code from one to another and make code recommendations in plain language.

Google announced the public preview of Gemini in BigQuery, a fully managed data warehouse from Google Cloud that allows users to analyze and manage data at large scale. Adding generative AI capabilities to BigQuery will open up a new range of analytic opportunities for business through access to metadata by providing valuable insights not just through conversational chat, but also enabling visualization of data by providing context, Google said.

Gemini in BigQuery will extend to query recommendations, semantic search capabilities, low-code visual data pipeline development tools and AI-powered recommendations for performance improvement. It will allow users to quickly understand and see what the data means under the surface without needing to look at charts and graphs.

Google also announced the public preview of Gemini in Looker, a business intelligence platform that organizes data from multiple clouds within a real-time view for insights analytics. With the power of Gemini’s conversational and visual engine, users will be able to receive analysis, automated Google Slide generation, report and formula generation, and more that can be integrated directly with Workspace.

Building AI experiences with Vertex AI Agent Builder

Using Vertex AI, Google’s managed cloud services that companies can use to build and customize generative AI models, developers will be able to build and deploy conversational AI agents using a no-code tool using foundation models, including Gemini, combined open-source tools such as LangChain and Google Search.

Vertex AI Agent Builder provides a powerful out-of-the-box system for developers to simply define a goal they want, provide step-by-step instructions that the generative AI agent will follow to achieve that goal, provide examples for the agent and the builder will produce an agent that can be deployed for enterprise purposes. For more complex use cases, multiple agents can be linked together, with a lead agent acting as an overseer and others acting as workers under it.

“The Vertex AI agents console also features advanced tooling to make agent building, orchestration, and maintenance easier,” said Burak Gokturk, vice president and general manager of cloud AI and industry solutions at Google Cloud. “Including the ability to create production-grade, high-quality agents out of a prototype, monitor the performance of agents in real-time, and improve responses for specific queries by training them using natural language.”

It is easy to augment Vertex AI agents with enterprise data using out-of-the-box grounded truth using Google Search and RAG, or retrieval-augmented generation, a process of optimizing the output of large language models with external data to make it more accurate. This allows for the input of fresh, accurate data to reduce the likelihood of hallucinations and increase answer completeness.

Additionally, Vertex AI extensions, data connections and function calls can be built in to connect AI models to tools to allow them to bring in external data and execute tasks.

Vertex AI is already home to more than 130 generative AI models. Today Google announced the addition of expanded access to newer models. These include Gemini 1.5 Pro, Google’s most complex and largest model to date, it comes in two sizes of context windows 128K tokens and 1 million tokens, available in public preview with the ability to process audio files, including video with audio. Others include Anthropic PBC’s Claude 3 family of AI models, Google’s lightweight CodeGemma coding assistance models and Google’s Imagen 2.0 text-to-image generation models.

Enabling developers with Code Assist

Duet AI for Developers has evolved into Gemini Code Assist, which is now built with Google’s most powerful AI model, and offers enterprise-grade coding assistance within editors such as VS Code and JetBrains, it is also capable of fully supporting private and public repositories such as Gitlab, GitHub and BitBucket.

Currently in private preview, Code Assist is powered by Gemini 1.5 and allows developers to perform large-scale changes across an entire codebase, including the addition of new features and cross-file dependency changes. It uses the model’s extremely large 1 million-token context window to retain a view of the codebase to allow for a very wide scale and comprehensive code review assistance as well.

Enterprise customers can also connect the AI assistant to their private codebase and tailor it to their specific internal style to help their developers remain in line with company standards at all times. Additionally, the company said, it has the capability to connect to multiple repositories at once for additional context.

In addition to coding, Google announced Gemini Cloud Assist, an AI agent that helps cloud teams design, deploy workloads, manage applications and troubleshoot issues in cloud environments. It includes capabilities for goal-driven design that allows users to describe their desired outcome and Cloud Assist will generate configurations based on user needs and even explain reasoning and provide additional recommendations. For troubleshooting and optimization, the assistant can help diagnose issues, pinpoint root causes and help prioritize savings, performance or high availability depending on user needs.

Image: Google