My recommendation would be to use Logseq.

It’s similar to Obsidian (“Second Brain”/ PKM), but with the journal function as backbone.

It relies heavily on crosslinking, is markdown-based, very efficient and a joy to use once you “got” it, and supports a hell lot of features, including TODO, plugins, a knowledge network (“graph view”) and much more.

I use it for everything (external brain) and pretty much never loved a piece of software this much!
It sounds like it is THE tool you’re searching for!

I made a similar post a few weeks ago.
I will try uBlue core and give you all a small update about it.

I feel similar about Debian. It’s a good distro for sure and I don’t have any issues with it for server use, but somehow, I still don’t like it somehow. RPM-/ OSTree based distros are more my taste, and I don’t even know why.

Because containers (Distrobox, Flatpak, etc.) are bae.
You can read my post I made a while ago for more information: https://feddit.de/post/8234416

Once you “get” image based distros, you probably never want to go back. Traditional distros just feel… off now for me.
Containerisation is the biggest strength in Linux, we use it all the time on servers, so why not on the desktop?
Atomic OSs just make more sense for me, not only because of security/ bug/ whatever reasons, no, also because they feel simpler and are pretty convenient and robust.

There’s a big shift happening right now, you’re right on that.
Traditionally, ARM is not as capable in solving complex issues, but more efficient.

That’s why it has always been used on smartphones for example. You want a lot of battery and don’t need to do highly complex stuff on that, that’s what you have your PC for.

The big focus in the last years has always been to top the competitor in terms of performance, and only right now, people begin to question if the computing power they have right now isn’t enough and if they rather wouldn’t like to have a device that’s more efficient.
The tradeoff is, you’re more limited to this specific architecture. Apple solved this by making a compatibility layer for x86 apps, but that of course comes with a performance hit.

I’m no expert in that topic tho, so take all I said with a lil grain of salt.

Right now, I think you’re better off with x86, because your server will definitely run on some sort of Linux, and we don’t have any compatibility layer or something like that yet.

Where I live, electricity is also very expensive. I monitor every watt.

I asked the same question half a year ago, here’s what I’ve learnt: RPis tend to be less reliable and aren’t that energy efficient. They’re great for small appliances, but for servers (e.g. NAS) not as much.

Get an used Thinclient/ mini PC. They cost something between 50-150€ and give you a huge performance boost, more ports, a x86 architecture, are better repairable (still often bad) and more.

Mine uses about 10-15 W on normal use, and 20 rarely when my cloud is under heavy use.

Dude… It’s the hundredth time you’ve posted this copypasta.
Image-based OSs aren’t locked down and also don’t depend on proprietary services.

You can just read my post I made about immutable systems, maybe we can discuss it there.

But, I wouldn’t choose a image based OS right now too for servers. At least yet.
I’m just afraid about compatibility, because many installers and services might rely on access to the root file system for now. Debian is right now the best choice as server OS, but that might change in the future.

Alright, interesting… As I said, I’m no expert or anything and this was just my noob optinion.

Thank you for the correction and further resources!

I don’t know what’s your intention.
I’m no expert or highly qualified in any way, so please correct me, but I don’t know if that’s the right way.

LLMs usually need lots of computing power, optimally in form of a GPU.
I use GPT4All, and when I send a prompt, I notice the temps/ fan speed and usage of my GPU turning up instantly to almost 100%. If it’s a longer one, my PC sounds like a helicopter 😁

In terms of hosting a server, you want something barely good enough for your service, e.g. running your cloud. This results in way less power draw, which is what you want, since it runs 24/7. Something powerful enough to run LLMs comfortably would likely draw lots of power, even an Apple Silicon.

I think, you’re better off just using GPT4All on your gaming PC if you need it.

I hope I’m wrong, and that M1s draw barely any power, especially in idle.
And even if I am, they (almost) can only run MacOS, which wouldn’t be a good server OS.

It sounds like your configuration is not sufficiently backed up.

It is backed up. 1x per auto-backup (Borg, included in the AIO) and 2x on different external drives.
The setup isn’t complicated too, basically barebones Debian with docker.

It’s just that setting everything up (once) again is annoying and highly inconvenient.

But, thank you for your tips on the LUKS-stuff, I will consider it! 🙂

I quite like your idea of just moving my data (not the other stuff) to another drive.

How can I decrypt my drive without a GUI, monitor or keyboard?

Even if firearms were legal in my country, I wouldn’t murder someone because of some embarrassing photos where I discovered Snapchat filters when I was 15.

Also, if they were legal, they wouldn’t help, since I’m not 24/7 at home

I asked the same question a few months ago on a german community.

Most people advised against a SBC (RPI, …). They’re not that much more energy efficient than Mini-PCs (especially with an Intel NUC or Celeron), are more modular + repairable, and use the more common x86-architecture.

You can get an used ThinClient for less than a RPI3, not even to mention a 4. This, and that you don’t contribute to more E-waste, is great.

I use a Fujitsu Esprimo Q920 with an Intel i5-4590T processor, 8 gb RAM and only SSDs.
It draws about 11W under normal load, a RPI3 draws about 5-7, including hard drives, the 4 even more with the active cooling and more performant CPU.

The RPI isn’t that more energy efficient, even with the enormous german energy prices, the thin client costs only a few bucks a year.
The RPI is also more prone to break, especially the SD-card.

I’m pretty happy with my current setup, would recommend.

Getting used to a UNIX shell and to UNIX philosophy can take some time, but it’s very rewarding in making everything more simple (thus more efficient).

Yeah, and that’s the problem for me. See my comment above. Nextcloud and those services are “bloated”, yes, but very convenient. I never worked in an IT-environment, so I’m a total noob.

But stuff like NC AIO give me a whole pre-set-up LAMP stack without needing to know how everything works, and that’s unbelievable for me.

My least favourite (and only) method for me to cut costs is reduce my energy consumption.

I already have a super cheap setup (used Mini-PC for 50 bucks, old SSDs I had lying around, etc.), so reducing the hardware costs more isn’t possible.

But, without tweaks, this setup would eat 15W (idle) and 25W (under load) electricity. At least, thats the case atm.

I just started selfhosting to be fair, and I didn’t have time to throttle the server. I use it mainly as NAS, so speed isn’t as important in this case as for other services like webapps, where reactivity is needed.

The CPU isn’t too bad, so, even when reducing the performance to 50%, it should still work.

Also, I will try to change the active cooling fan to a passive heat sink, that might reduce the bill further.

What mainly eats resources like crazy is my Nextcloud AIO. I try to follow the UNIX-principle as good as I can, and NC doesn’t follow it well, at least for my use case.

I only need a file server, and NC is pretty “bloated” with talk, calender, and so on. So I disabled all of that.

But, I’m not capable enough to set up an Online-FTP-Server and secure it enough without ever working in that environment. NC AIO provides a lot of comfort and “just works”. So, I’m fine with that.

Here in Germany, especially thanks to the energy crisis, electricity is absurdly expensive, and even reducing the TDP by a few % will save me much money over a year.

So, I try to reduce the load and increase efficiency wherever I can.

Still, even now, with an increased energy consumption + paid domain, it’s still cheaper than using OneDrive or something like that, even when ignoring storage size.

Maybe you don’t even need that, at least for accessibility.

Windows for example now has exactly this feature, which is a speech-to-text-transformer powered by some “AI”.

But, in contrast to the Bing chat, this works (afaik) offline by some FOSS-backend, which I don’t know the name of anymore (maybe someone else will?) You could use that tool for live transcription. That is supposed to work extremely well!

Please correct my if I’m wrong, I don’t use Windows anymore personally, and at work we have a business edition that doesn’t ship this brand new feature yet.

(Side note: as strongly as I hate Windows, this feature is absolutely godsend for hearing-impaired people and should be adopted by every other OS!)

If you want to transcript movies and thereof in bulk by uploading them, I can’t give you any information, sorry.

But I believe there are some sites that give you the “subtitle file” for download freely, which you can add manually for each movie in Plex/ Jellyfin.