Fedegenerate

Sum total of my hardware:

Ugreen dxp 4800, the Pentium one. 32gb ram. My main box: jellyfin, arrs, immich, pihole, nginx, etc… It can’t go down. I don’t think I want an LLM here.

Beelink n100 16gb ram. Local back up, redundant pihole, immich machine learning… Generally under utilised, I’d like to move some services around, does proxmox have an auto balancer?!

Spare no name n100. 8, or 16gb, I can’t remember. Abandoned box, It’s what I would put the LLM on, I did have it reserved for a remote back up.

I think it would need a ram upgrade, see corporate AI pricing me out of personal computing. Currently Amazon has a Crucial 32gb ddr5 sodimm module for £280. Which is too high a price for what I’d use it for.

Oh, and I have an abandoned gaming rig with a gtx970, and some rPi0/3s

I’ve put Ollama on an n100 before. It obviously ate all of that box and made everything on it chug, and it was too slow for human use. But if it’s just generating logs, and resetting containers then I wouldn’t mind how slow it is.

I’ve been dreaming of local AI for a hot minute. Sadly corporate AI has priced me out of personal computing, and current hardware isn’t up to it.

Maybe my n100 could, I don’t really care how slow it’s generating the tokens if it’s just resetting containers and logging errors.

Does your setup handle automagic updates, how do you handle prompt injection if so? Or, just error correction/logging?

Also, 1500 years ago, destroyed every transpohbe’s “define a woman” nonesense.

A hosting service for games. Feed it games and it allows you to download them from your server, manage saves, emulate in browser… A bunch of stuff. It’s a good project.

I went rpi4>n100> a couple n100s and that pi> the dxp4800, I think it’s a pentium, and those n100s. I think I’m ok here, I have networking, compute + local backup, and storage all in their own box.