google-gemini/angular-webxr-art-sample ? reverse-engineered prompt

Build me an Angular web app that works like a WebXR art gallery where people can browse generative AI images in a nice immersive experience.

I want the app to create new images using Google Vertex AI image generation, then automatically generate captions for each image using Gemini so the gallery feels more alive and descriptive. The main experience should be smooth in the browser, work as a normal local app during development, and make it easy to connect either to a local Python backend or a Google Cloud Function for the AI calls.

Please set it up so I can run it locally with a simple install and start flow, and include clear places to plug in my Gemini or Google Cloud credentials. If anything depends on current Google setup details, look up the latest docs online and wire it up in the simplest working way. Keep the UI clean and focused on showing the artwork and captions in the gallery.