Very Demure, Very Mindful

Logo(s) generated by DALL-E and ChatGPT-4o

The name is a play on a current TikTok trend.

Motivation

After listening to this podcast where Brené Brown interviews Dr Amishi Jha to dive into the distinction between Focus, Attention and Memory, I have a renewed interest in my own mindfulness practices.

Dr. Gabor Mate writes in his book Scattered Minds about the importance of a mindfulness practice to help those with ADHD to manage and even reverse some of the attention deficit traits.

After trying a few different "mindfulness" apps, I wasn't convinced of their value with subscription pricing but could I have a more on-demand model? Turns out, yes I can.

Architecture

Click this arrow for detailed diagram

Architecture Diagram: Simple

Architecture Diagram: Detailed

Roadmap

Done

• get OpenAI to generate a guided mindfulness transcript
• get AWS Polly to use the text-to-speech to synthesize the audio.
• Add slower cadence and pauses. Leverage SSML
• Add AWS Bedrock intregration to leverage their models
• Deploy to CloudFront with security best practices (OAC)
• Deploy under custom domain name with SSL Certificate.

TODO

• Port simple HTML/CSS to React SPA
• Add Lambda@Edge and Cognito authenticated areas (Google SSO as MVP)
• Add API Gateway + Lambda API endpoint
  • Simple HelloWorld
  • List user protected content
  • Allow user to generate new content
• Consolidate the API Gateway Endpoint and the CloudFront static site under the same url but use Lambda@Edge to redirect /api/ requests to API Gateway instead
• Integrate Stripe or some other payment portal provider to make generating on-demand new audio tracks as a pay-per-generate option.

Quickstart

make dev

# Generate script for a "10 minute" session.
# TODO: Get the generated audio to actually pause with silence when cued.
python3 -m very_demure --duration 10 --voice Matthew

python3 -m http.server --directory docs

Samples

Checkout an example transcript and MP3 in the docs/samples/ folder