MAYA-Platform/ReadTheRoom ? reverse-engineered prompt

Build me a local web app that helps me tune how my AI assistant behaves, so it stops sounding generic and starts sounding like me. I want a short guided session where I can adjust things like directness, warmth, humor, profanity, how it handles corrections, brainstorming, short replies, and when it should use tools or stay quiet.

The app should show me a side by side comparison between the default behavior and the calibrated version, and it should make it easy to review exactly what changed and why. Keep everything local and private, with no account, no telemetry, and nothing saved silently. I want it to work nicely on desktop and mobile, and to feel simple enough that I can run it locally and test it right away.

If you need current docs or best practices for any part of it, look them up online as you build.