Fair enough. I guess I imagined someone hosting all the selfhosted web apps that get posted to this forum, when most people likely just host only the few they need on the go, so it isn't really that burdensome.
Edit: Forgot to add: I always though that it could be useful to just set up Apache Guacamole, so that instead of the hosted services, my family members could just use remote desktop apps but I never got around to it.
Yep my thoughts. New selfhosters think the hard part of selfhosting is command line but that's "kinda" like thinking that the hard part of math is writing numbers on paper. Terminal is just the medium, not the complex part. Navigating filesystem and editing files is easier on the desktop but changing permissions and managing services would be be extremely difficult for a newbie without using the terminal because (almost) every online tutorial uses terminal. OP would have to learn how to translate the terminal command to its desktop counterpart at which point they might as well use the terminal.
OP also has an XY problem. They asked for a system which does not require terminal usage but they should have actually asked for an easy to set up system. People are recommending things like Yunohost though, so it's fine in the end.
It's nice there's a front end for all these tools but I kinda don't get why is everything "hosted". This could have just been a desktop app. I guess it can be useful when you want to convert something on your phone but to me it just seems like unnecessary server maintenance burden.
Dunno whether they are transcoding or not nor why they have such a bizarre setup. But I would hope 16C/32T CPU from 2017 could handle software transcoding. Also peak power consumption while playing a movie does not really matter compared to idle power consumption. What matters more is that the motherboard is probably packed with pcie slots that consume a lot of power. But to OP it probably does not matter if they use a threadripper.
I don't recommend it unless you just want it for storage or whatever else it does out of the box. It's basically impossible to tinker with it because it has so many layers of abstraction. At least that was my impression when I tried to edit their nginx config. It had like 2000 lines so I just gave up.
If you want a server that runs services that you download from the internet, don't buy it. Look at it as a box that does the thing that it promises to do, not as a computer. If you want it to do a different thing, buy a different box that does that. Kinda like a TV. It's technically a computer that runs some kind of linux but to the user it's a monitor that also shows videos from the internet.
Also it's perfectly fine to buy a "NAS black box" but maybe not something I'd buy if I wanted to get into selfhosting. I'd buy it if I wanted to have a NAS running at home with the least amount of "self" in "selfhosting" that's feasable.
Proxmox is nice for beginners. This is a nice tutorial: https://youtube.com/playlist?list=PLT98CRl2KxKHnlbYhtABg6cF50bYa8Ulo
Proxmox has nice UI for managing Linux Containers (LXC). They act like a computer inside a computer with the advantage that you can clone them. So you can basically save and load them whenever you succeed or fail at something. Proxmox also allows you to install Turnkey Linux containers which have the software you want to run preconfigured in them so that's also good for beginners.
Only downside is that this is not declarative so it won't be as scalable as docker or nix. It might be more worth it to learn docker from the beginning but that would also be less friendly for a beginner.
Yep any core i3 is fine even for desktop given an SSD and enough RAM. Once you delve into the core2 era, you start having problems because it lacks the compression and encryption instructions necessary for the day to day smoothness. In a server you might get away with core 2 duo as long as you don't use full disk encryption and get an SSD or at least use ram for caching. Though that would be kinda a bizarre setup on a computer with 512 MB of ram.
Ryzen is not recommended for transcoding because the Radeon integrated GPU's encoding accelerator is not as fast as in intel iGPUs. But this does not come into play if you A) have 16 cores and B) don't even have an integrated GPU.
And about idle power consumption: I don't think it's a point of interest if you are using a workstation class computer.
I used to selfhost on a core 2 duo thinkpad R60i. It had a broken fan so I had to hide it into a storage room otherwise it would wake up people from sleep during the night making weird noises. It was pretty damn slow. Even opening proxmox UI in the remotely took time. KrISS feed worked pretty well tho.
I have since upgraded to... well, nothing. The fan is KO now and the laptop won't boot. It's a shame because not having access to radicale is making my life more difficult than it should be. I use CalDAV from disroot.org but it would be nice to share a calendar with my family too.
You simply use a different packaging format as I said in the previous comment.
I'm not arguing whether snap or flatpak is better. Flatpak is better.
But your arguments are going against each other. You disagree that FSF should tell you what software you can use but then you want to tell other users what software they can use. If you use flatpak despite of FSF's opinions, you should let people use snap despite of your opinion.
You can tell Open Interpreter to run commands based on you human-language input. If you want local only LLM, you can pair it with Ollama. It works for "interactive" use where you're asked for confirmation before a command is run.
I set this up in a VM because I wanted a full automatic coding "agent" which can run commands without my intervention and I did not want it to blow up main system. It did not really work though because as far as I know Open Interpreter does not have a way to "pipe" a command's output back into the LLM so that it could create feedback with linters and stuff.
Another issue was that Starcoder2, which is the only LLM trained on permissive licensed code I could find, only has a 15B "human-like" model. The smaller models only speak code so I don't know how that would work for agentic usage and the 15B is really slow running on DDR4 CPU. I think agents are cool though so I would like to try Aider which is a supposedly good open source agent and unlike Open Interpreter is not abandonware.
Thanks for coming to my blabering talk, hope this might be useful for someone.