I could see this being a use case for a NixOS deployment where your company manages the configuration file and versioning of the system, as well as providing support. Over time, I’d you’re diligent about building documentation based off of each support request, you’ll end up with a personalized guide. And if your customer decides take a break or quit entirely, they have a configured system that doesn’t lock them in into something too esoteric.
Disclaimer: I only know of Nix, never used it because I just don’t manage that many machines to be worth my while to learn it.
Some people have different requirements. Some have been burned by a program becoming obsolete or migrating between programs and finding out getting their data was (at least initially) beyond their capabilities. Some don’t see the tradeoff of having true rich text as worth it.
I’m not in that camp, but I can see the appeal.
Not to mention fixing it in Immich isn’t possible. What that donut fails to see is that the only way this is acceptable to OSM is because of the proxy to lower the demand on their servers. Unless they’re suggesting reimplementing a proxy inside Immich, which is way beyond the scope of a “simple patch”.
I’m really curious about the motivation behind such permissive licensing too.
The project is yours, you can always distribute a commercially licensed version of it regardless. GPL or any other license will never be a hindrance.
The only thing a permissive license does is allowing everyone else to issue commercial versions of it without even sharing source code.
The requirements for a media server mesh well with a NAS and *arr suite and other light loads. Low CPU demand, some RAM demand, integrated GPU if you need transcoding and that’s it.
They are wildly different from generative AI. For good performance, you’ll want a decent GPU with loads of VRAM or brute force with raw CPU power and RAM. If you care about power draw at all, you don’t want this on 24/7/365. Why not build a cool gaming rig and use it for AI? As a bonus, now you have a cool gaming rig with your AI machine!
I am just starting so take this not as a recommendation but as an option. I am familiar with Linux but do not work in IT.
I got myself a used desktop as a starting point. It can handle 2x 3.5” drives, one 2.5”, plus an NVMe. You could buy an adaptor and change the DVD drive for another 2.5” caddy, but more on that later. It came with 8GB of RAM, but it can handle 64. I spent something like $250 including cables, bolts, caddies, but not drives.
If you watched the video, you’ll notice the CPU has video transcoding acceleration and encryption acceleration too. It comes out ahead of modern N100 CPUs being widely used for home NAS these days, and draws a minuscule amount of power while idle. Indeed, most of the idle power draw for my machine comes from the drives.
So pros:
Cons:
For software, I’m using TrueNAS scale. It’s easy to install and configure, there’s good documentation and a support forum, can run docker containers and VMs. Lots of administration quality of life tools built in that you don’t need to build. Plus it’s Linux and I can tinker with it if the need arises.
To get to what you want, you could install an M.2 A+E to SATA adaptor and a slim DVD to 2.5” caddy to come up to 4 drives, add memory, a multiport multigigabit NIC, an NVMe and 4 drives and you’d be set. VMs for your firewall, VPN, pihole, dockers for the rest.
500GB HDDs? If you don’t need hot swap, with your budget, buy new ones. Transcoding is solved by using any modern-ish Intel CPU. I built one for $230 using a used office desktop and three 4TB HDDs plus a small used NVMe for TrueNAS.
Look at the 3 minute mark for transcoding accelerated CPUs: https://youtu.be/WCDmHljsinY
Live fast and die young. Don’t ever slow down. If the account locks, then it locks.