AI Creators
Google Cloud Tech
How to build and deploy powerful data apps with agents
Oct 16, 20261 min watch
Quick Overview
This video is a practical tutorial presented by a Google Cloud developer demonstrating the Looker data apps development workflow. It guides viewers through the steps of configuring a local workspace, prompting AI coding agents, and deploying an interactive dashboard application into Looker. The tutorial was created to showcase how AI agents accelerate building custom, secure business intelligence interfaces.
Key Points
- 1.Data applications can be hosted directly inside Looker while preserving the governance and security of the LookML semantic layer.
- 2.Initial setup requires executing a terminal installation script and providing a Looker instance URL, API credentials, and a target folder ID.
- 3.Developers can use AI coding agents such as Antigravity or Claude Code to generate multi-tab user interfaces from natural language prompts.
- 4.Applications can be previewed locally, customized using uploaded images or logos for color themes, and published to Looker with a single prompt.
- 5.Deploying data apps enables JavaScript interactivity and custom visualizations while querying underlying Looker explores.
Summary
The presentation opens with a demonstration of a flight operations center data application hosted directly within Looker. The presenter highlights its modern user interface, responsive widgets, and compliance with data governance, noting that the entire application was deployed in minutes using Looker data apps tools.
To recreate the project, the workflow starts in a development terminal where the presenter runs a shell installer script fetched directly from Google's distribution URL. After specifying a project working directory name, the setup script requests Looker connection details. The developer provides the Looker instance URL, the API client ID, and the API client secret. The installer then prompts for the target Looker folder ID, which the presenter explains can be found by navigating to personal folders in Looker and copying the trailing numeric identifier from the browser address bar.
Next, the presenter opens an AI coding agent inside the workspace, highlighting tools such as Antigravity or Claude Code. A prompt is entered instructing the agent to build an interactive FAA flight data app featuring a three-tab layout, key performance indicators, a origin visualization map, and a data architecture tab. The agent inspects the LookML explore schemas, processes the flight and airport definitions, generates the code, and spins up a local preview server on localhost.
After reviewing the local preview, the presenter demonstrates visual customization. By providing an image of a cloud logo to the agent with an instruction to update the color palette, the tool extracts primary and accent colors and updates the UI theme across both light and dark display modes. Once the design meets expectations, the developer instructs the agent with a single command to publish the app. The agent registers the app directly within the target Looker instance as an extension linked to a dashboard ID, making it accessible to authorized Looker users.
The finished application in Looker features interactive flight trajectory maps, origin hub filtering, carrier volume breakdowns, and a data lineage tab detailing the underlying Cloud SQL tables and LookML explores. The presenter concludes by emphasizing that this approach combines the visual freedom and interactivity of custom JavaScript development with the underlying governance, security, and access controls of Looker's semantic layer.
Environment Setup and Looker Configuration
The build process begins in a terminal workspace by running an installer script via a curl command. The user specifies a working directory name, inputs the Looker instance URL, and provides the API client ID and client secret. To determine where the application will be stored, the user locates the numeric folder ID at the end of their Looker personal folder URL and enters it into the configuration prompt.
Prompting AI Agents to Generate the Application
With the workspace configured, the developer launches an AI agent such as Antigravity or Claude Code. By submitting a natural language prompt, the developer requests a three-tab custom interface connected to governed LookML semantic models, using FAA flight data as the example dataset. The agent reads LookML explore schemas, builds the application structure, and serves a local preview link.
Refining, Theming, and Publishing the App
The developer inspects the local preview to evaluate layout, widgets, and themes, with options to toggle between light and dark modes. Theme colors can be updated by uploading an image or logo, from which the agent extracts a palette to restyle the interface. Finally, issuing a single publish prompt compiles the application and deploys it directly into Looker as a shared dashboard extension.
The Bottom Line
The video establishes a straightforward workflow for using AI development agents to generate and deploy custom interactive web applications directly into Looker. It demonstrates that developers can achieve full interface flexibility with JavaScript and responsive layouts without sacrificing the governance and security of LookML models. While it proves how quickly an application can be generated and published, it leaves aside advanced enterprise concerns such as multi-environment deployment pipelines and ongoing version management.
FAQ
What is a Looker data app and how does it integrate with Looker?
A Looker data app is a custom interactive application featuring responsive widgets and tailored interfaces that is hosted directly inside Looker. It connects directly to Looker's governed LookML semantic layer, ensuring that queries adhere to existing data security and governance models.
How is the development environment configured for building Looker data apps?
The developer pipes an installation script into the terminal workspace, names a working directory, and supplies the Looker instance URL, API client ID, API client secret, and the numeric target folder ID taken from the Looker folder URL.
Which AI agents are highlighted for generating custom code in Looker data apps?
The video specifically highlights Antigravity and Claude Code as AI coding agents that can run in the workspace to generate the application structure and interface from text prompts.
How can developers customize the visual appearance of a Looker data app?
Developers can prompt the agent to adjust layouts, switch between light and dark modes, or upload an image or logo so the agent can extract color values and apply matching themes to the application.
How is the finished data app published to the Looker instance?
The developer enters a single prompt instructing the AI agent to publish the app, which automatically packages the code, creates an inline extension, and makes it available as a dashboard in Looker.
Worth watching for
Data analysts, Looker developers, and business intelligence engineers interested in using AI coding agents to create custom, interactive data applications within Looker.
- looker
- google-cloud
- ai-agents
- data-apps
- lookml