From a theory perspective alone, ignorant of Lemmy specifics: a database query can be made to list all cached images including a unique identifier for each image. Use this list to find each cached image.

Look at your cached image list and decide how you want to prune it. The most likely pattern for this system is FIFO, so prune the oldest cached data until you drop below the target disk usage.

In practice, you’ll likely use somebody else’s solution. Be sure to read the contents of their solution carefully to ensure it doesn’t move sensitive data to an externally accessible location or exfiltrate data directly.

I’ve never had a problem side loading an entire folder into Firefox. Maybe it was an addon rather than extension though?

JVM can run on any platform that supports Java, as you’re building Kotlin directly into Java bytecode.

Multiplatform is for building native applications while using a single backend logic. You’ll have to write separate handlers for everything unique to the platform, according to documentation.

It’s pretty easy to just not put the AI tag on things, or to strip such things away from an image.