This repository provides a full-stack, serverless boilerplate for building AI applications on AWS, targeting developers who want a robust, cost-effective foundation for LLM-powered applications. It offers a complete architecture including backend APIs, event-driven services, a database, and a frontend, all designed to leverage AWS Bedrock for AI capabilities while ensuring data privacy.

How It Works

The stack utilizes AWS services like Lambda, API Gateway, DynamoDB, and EventBridge for a scalable, pay-as-you-go architecture. It features an Express.js backend for APIs and business logic, with a separate service for AI chat powered by AWS Bedrock, supporting multiple LLM models. Streaming responses are handled via Lambda Function URLs for the AI chat service, while other APIs use API Gateway. Serverless Compose manages configuration and deployment across services.

Quick Start & Requirements

• Install: npm install (in root), then serverless deploy
• Prerequisites: AWS Account, AWS Credentials configured, Serverless Framework (npm i -g serverless), Node.js. AWS Bedrock models (e.g., meta.llama3-70b-instruct-v1:0) must be enabled in the AWS console.
• Setup: Estimated setup time depends on AWS model enablement and deployment.
• Docs: awsaistack.com (Live Demo)

Highlighted Details

• Full serverless AI chat architecture with streaming responses.
• Data privacy: App data never leaves AWS and is not sent to model providers.
• Domain-oriented design allows easy removal of components like AI Chat or authentication.
• CI/CD integration with GitHub Actions for deployment.

Maintenance & Community

The project is maintained by serverless. The README does not explicitly mention community channels like Discord or Slack, nor does it list specific contributors or sponsorships.

Licensing & Compatibility

The README does not specify a license. This lack of explicit licensing means commercial use and closed-source linking are not clearly permitted.

Limitations & Caveats

The project is presented as a boilerplate and may require significant configuration for production, including custom domain names and secure secret management. Hosting the static website via Lambda is noted as not recommended for production. The README also mentions potential issues with enabling AWS Bedrock models for new or low-spend AWS accounts.