GenAI-Based Application Architecture Designer

Problem Statement

Designing scalable, secure, and efficient cloud application architectures using AWS resources can be a daunting task, particularly for users who lack in-depth technical knowledge, such as product managers and non-technical stakeholders.

This project addresses that challenge by creating an application using Generative AI (GenAI) that assists naive users and product managers in designing the high-level architecture of an application utilizing AWS resources. The tool provides insights into the application's architecture complexity, service requirements, and associated costs.

Objective

The main objective of this project is to:

• Simplify the cloud architecture design process for users with minimal technical expertise.
• Leverage AWS resources to generate a scalable, efficient, and secure architecture based on the input requirements.
• Provide Terraform scripts for engineers to use Infrastructure as Code (IaC) to quickly deploy the suggested architecture.
• Utilize AWS Bedrock with LLaMA as the GenAI to provide intelligent and tailored architecture designs based on user-provided inputs.

Solution Overview

This application uses AWS Bedrock with LLaMA, a Generative AI service, to analyze user inputs and design an application architecture that fits their needs using AWS services. The AI selects appropriate components such as compute, storage, networking, and security, and generates a high-level architecture that includes AWS resources, a logical flow of components, and cost estimations.

Key Features:

• Automated AWS Architecture Design: Generates detailed architecture diagrams using appropriate AWS services.
• AI-Powered via AWS Bedrock with LLaMA: GenAI intelligently selects AWS components based on user input.
• User-Friendly: Aimed at non-technical users like product managers to help them visualize the high-level architecture.
• Terraform Scripts for IaC: Provides Terraform scripts to help engineers implement the designed architecture efficiently.
• Cost Estimation: Offers an estimated cost breakdown for running the application using AWS resources.

Challenges Faced

1. Lucidchart Developer Account Access: We faced challenges in obtaining a developer account for Lucidchart, which hindered our ability to automatically generate system architecture diagrams.
2. API Limitations: Some AWS Bedrock services have API limits that slowed down the architecture generation process during high-traffic times.
3. Cost Optimization: Balancing resource selection with cost efficiency was a challenge when dealing with GenAI model suggestions.
4. Terraform Script Generation: Handling dynamic and complex architectures required constant iteration on Terraform script generation logic.

AWS Services Used

• Compute: EC2, Lambda
• Storage: S3
• Networking: VPC, Route 53, CloudFront
• Security: IAM
• Database: DynamoDB
• GenAI Engine: AWS Bedrock with LLaMA to generate application architecture

Tech Stack

• AWS Bedrock with LLaMA: Used for generating application architecture through GenAI.
• Terraform: Used to create Infrastructure as Code (IaC) to deploy AWS resources.
• Selenium: Integrated for interacting with Lucidchart to construct system architecture diagrams.
• Python: Core programming language used for system logic, integrating with AWS APIs and Selenium.

Terraform Integration

For technical users such as cloud engineers and developers, this project includes Terraform scripts for deploying the designed architecture on AWS. The Terraform scripts serve as Infrastructure as Code (IaC), making the deployment process faster, more efficient, and repeatable.

Terraform Features:

• Automates the creation and management of AWS resources.
• Ensures consistency across environments by utilizing Infrastructure as Code.
• Makes it easy for engineers to adapt and scale the generated architecture.

Usage

1. Users provide basic information about their application's goals and required features.
2. The GenAI-powered engine (AWS Bedrock with LLaMA) processes the input and designs a high-level architecture using AWS services.
3. Users receive an architecture diagram, a description of the AWS resources used, and an estimated monthly cost for running the application.
4. The system also provides Terraform scripts that engineers can use to deploy the designed architecture as Infrastructure as Code (IaC).

Conclusion

This AI-powered tool, driven by AWS Bedrock with LLaMA, simplifies cloud architecture design and empowers product managers and non-technical users to visualize the complexity of their applications on AWS. With added support for Terraform-based Infrastructure as Code, engineers can quickly deploy these AI-generated architectures, making the entire process efficient and scalable.

Built With

• amazon-web-services
• bedrock
• dynamodb
• gemini
• lambda
• llama
• lucidchart
• python
• react