Nisarg0007/driftradar ? reverse-engineered prompt

Build me a web app called DriftRadar that watches structured developer data like GitHub releases or changelogs and tells me when the data starts going bad.

I want a clean dashboard where I can see each source, whether it is healthy, and when drift is detected. If an extraction suddenly comes back empty, missing fields, or with a different shape, the app should create an incident, show what changed, and give a simple diagnosis plus a suggested recovery action.

Then I want a human supervised recovery flow where I can review the proposal, approve a repair, trigger a scraper healing step, run a fresh extraction, and confirm that the data is actually back to normal. Please include a polished frontend, a backend API, sample data, and the full workflow from detect to verify.

Use the environment variables from the example file, and if you need current docs for Bright Data or the AI provider, look them up online.