Use natural language to create application design in Google Cloud with Gemini Cloud Assist

Key Concepts

Gemini Cloud Assist, Application Design Center (ADC), application design, multi-product application templates, natural language processing, multi-turn conversations, Vertex AI, ticket categorization, priority assignment, infrastructure application design, Terraform code, AI-driven knowledge suggestions, embeddings, Google Cloud Console.

Application Design with AI in Google Cloud

Introduction

Afrina, a product manager at Google Cloud, demonstrates how to use AI to create an application design in Google Cloud using Gemini Cloud Assist and Application Design Center (ADC).

Prerequisites

Two services need to be activated in the Google Cloud Console:

1. Gemini Cloud Assist: An AI collaborator in Google Cloud that provides AI assistance for various cloud operation and management tasks.
2. Application Design Center (ADC): A service for designing and deploying application templates with built-in governance, security, and observability.

Links to learn more about these products and how to activate them are provided in the video description.

Designing an Application with Gemini Cloud Assist

• Gemini Cloud Assist Chat can be used to design multi-product application templates in Google Cloud via ADC.
• Users can describe their intent using natural language and refine the design through multi-turn conversations.
• Example: The presenter prompts Gemini to design an internal application on Google Cloud for an IT-support department to enable faster resolutions and improve productivity using AI.
• Gemini generates a comprehensive application design, including a visual diagram showing global load balancers, front-end and back-end services, a database, and Secret Manager for storing database credentials.
• Gemini provides rationale for its recommendations, making the architecture easy to understand.

Iterating on the Design

• The design can be further improved by adding new features through multi-turn conversations.
• Example: The presenter asks Gemini to improve the design so that the system uses Vertex AI to categorize the type and priority of the ticket based on the keywords it finds.
• Gemini adds Vertex AI for ticket categorization and priority assignment to the design.

Leveraging Application Design Center (ADC)

• The application design can be viewed and edited within the Application Design Center.
• Clicking "Edit App Design" opens the infrastructure application design directly within ADC.
• Creating a Template: Before making changes, a template needs to be created to save the design in its present state. This allows users to return to it later directly from ADC and easily share it with other team members.
• The design generated by Gemini is displayed on the ADC Canvas, surrounded by options to customize, deploy, or view the Terraform code.

Customizing the Design in ADC

• The design can be further customized by adding AI-driven knowledge suggestions and responses to created tickets.
• Example: The presenter adds a bucket to store knowledge articles and a pipeline to generate and store the embeddings in a separate database.
• Gemini proposes relevant changes to the architecture.
• Users can iterate further, examine the proposed Terraform code changes, or accept the suggestion.
• Components can be added to the canvas using the tab or the Plus icon.
• Gemini provides context-aware suggestions based on the canvas and changes made.
• All design changes automatically generate Terraform code, which can be viewed in line.

Deploying the Application

• Once the application design is finalized, it can be brought to life by creating the application from the template.
• Clicking "Configure App" prepares the application for deployment.

Conclusion

The demo showcases how AI assistance from Gemini Cloud Assist can be used to quickly design and refine an application template via Application Design Center. Designing applications has become easier. Users are encouraged to try it themselves by logging into Google Cloud Console.