rajeevranjan4348-arch/Mamta-Kumari — reverse-engineered prompt

Build me a clean, working version of this Gemini powered AI Studio web app. The README is pretty minimal, so first inspect the existing code and keep the app close to what is already there.

I want a simple browser experience where a user can open the app, enter or speak a prompt if the existing live API code supports it, send it to Gemini using the configured API key, and see the response clearly on the page. Make the interface feel polished and easy for a normal person to use, with helpful loading states, error messages if the key is missing or the request fails, and a basic empty state explaining what to do.

Please wire up the local development flow so it runs with npm install and npm run dev, uses the environment example correctly, and does not expose secrets in the browser if the server file is meant to proxy requests. Look up current Gemini docs online if needed.