forthfate/openorbit ? reverse-engineered prompt

Build me a local web app for managing recurring AI evaluations and improvements.

I want a control room where I can create a build for an AI system, run it once or on a schedule, and then review everything that happened in one place. The app should show a dashboard with the current health of the system, recent changes, and active runs. Each run should keep logs, evidence, browser screenshots, and traces so I can understand what the AI actually did. I also want a clear review flow for supervisor feedback, issues, and suggested improvements, with human approval required before any change gets applied.

Please include a quick start experience so I can set up common use cases like a smoke test, site exploration review, agent self improvement, or behavior drift monitoring without wiring everything together by hand. Keep the data local, let me stop active work from the app, and make the whole thing feel polished and easy to use. If you need to, look up current docs online.