aws-samples/generative-ai-avatar-chat — reverse-engineered prompt

Reverse engineered prompt

GitHub

Build me a web app where people can chat with a friendly 3D avatar instead of a plain text bot. The avatar should be the main interface, with a chat area where users can ask questions and get streaming answers from Amazon Bedrock, using Claude Haiku by default.

I want it to work as a sample that can be deployed to AWS with CDK. It should include the frontend, the backend functions, and the AWS setup. Please make it easy to switch between normal chat and document based answers. I want to be able to use Bedrock Knowledge Base, Amazon Kendra, both, or neither, depending on a config setting.

Also include a way to add documents so the bot can answer from them, plus clear deploy, redeploy, local frontend run, and cleanup commands. For production, include optional WAF settings like allowed countries or IP ranges. Use sensible defaults for Tokyo region, and look up current AWS docs online if you need to.

Want more depth? Deep Reverse