m-p{3}

4GB of RAM: load a model into llama.cpp

Explodes

At that price point not really, and there is definitely going to be some tradeoffs required between power vs consumption vs price.

Not sure where you’re shopping, but I can’t find it below $200 for 16GB of RAM.

I don’t think there’s anything on the market that checks all the boxes for a $140 budget.

So it’s basically a local CDN for Steam and other platforms?

The best practice is that the device private key should never leave the device. One device, one key. A device is lost or compromised? Remove that device’s key from your list.

I also use ConnectBot and importing a private key generated elsewhere never worked for me, so I generate directly in ConnectBot itself and add its public key to my server, which is what I should do anyway.

You’ll have to convert your docs and sheets to another format if you migrate to something else, that’s kind of inevitable.

Which file formats do you mainly use? You can likely open them on mobile or desktop as long as you have a viewer for the filetypes without requiring a web-based viewer/editor, which would reduce the strain on the server as well.

It also makes the server setup way simpler, as it just deal with hosting and serving files, and the data processing / rendering is all done locally.

Easier to do backups too.

Meet -> Jitsi

For Drive, do you need the editor to be web-based? If not, you could just enable SSH access and access it through SFTP using keyfile authentication.