I literally just sent some friends a docker image and told them to point Luna at it, for which they were set up in ~20minutes and $3 of tokens, so I'm skeptical.

My homelab is also currently ~40 nix service VM's plus a handful of Ubuntu, etc long tail, and some external DO droplets for NAT, and a more robust vlan config than I'd ever been able to put time towards before... which I haven't had to do more than a dozen hours of manual setup on IN TOTAL thanks to Fable, Sol and local Qwen.

There's truth in this post, but I doubt it's longevity.

▲

KnightHawk3 2 days ago | parent [-]

Why split the nix into so many VMs instead of just systemd units?

▲

glaslong 2 days ago | parent [-]

They're not all tooootally atomic, I have clustered things like libation+audiobookshelf, ollama+ds4+litellm, but I like that I can manage things almost entirely at the network and host level, generally atomically per service, and especially between services (eg homeassistant and the local ai).

It's DEFINITIVELY overkill, and fills up fast, but nix making everything near fully declarative to roll fwd/back or shuffle between proxmox hosts at any time makes for a clean and easy mental model. Both for me and the AI, because it can infinitely retry a nix flake config until it gets it perfect.

▲

lee_ars 2 days ago | parent [-]

Sounds like a poster child for LXC containerization with Proxmox — all the host-level management of a non-ephemeral VM (including snapshots!), without the every-machine-has-its-own-kernel overhead.

	▲	glaslong 2 days ago \| parent [-]
		I do have a couple LXC for the things that need tight GPU control. Ollama and DS4 permanently, and an ml-workbench that steals the 3090 from my Cachy gaming VM when I get it in my head that I could try implementing some cool gsplats paper haha. I could definitely slim things down without disko, virtio, etc, and especially whole separate kernels.. but it's also not such a huge diff that I mind. I'm not adding new services all that fast anymore now that I've got my basics + wishlist set up, so I'll grab another box from the local pc recycler if I ever need space for another ~30 VMs some day rather than when I need space for another ~40-50 LXCs. And in the meantime if OpenAI publishes a scary brag about how GPT-7 popped a shared kernel on accident I can be relieved that my private bookmarks repo isn't vulnerable while the internet burns down.