coro1404/fal — reverse-engineered prompt

Build me a simple mobile friendly web app for fal.ai that I can run locally or in Docker. It should have a basic password login, then one clean screen where I can pick a model from a dropdown and either generate images from text or edit images I upload.

For editing, let me upload 1 to 3 images, type a prompt, and optionally choose things like aspect ratio or output size when the model supports it. I also want a quick selfie capture from the camera that can drop into an upload slot. Include the special tools for background or object removal and old photo restoration.

Please show the generated images nicely with a download button and a rough cost hint. Save the last 25 prompts on the server, save the last 3 uploaded images as reusable thumbnails, and show the last 9 requests in a small gallery that survives reloads and restarts. If possible, sync recent history from fal on login and use small cached thumbnails so the gallery loads fast. Also show 24 hour fal usage if the API key allows it. Look up current fal.ai docs if you need to.