Strange, postgres is usually not the one to die. My server is permanently running out of space, so I have to keep deleting shit, and my PG did not die once. On the other hand, my other very old server (not updated in 3 years) has it’s PG die (and automatically restart) at least once a day.
Wow, I never thought about storing build data in an SQLite file. That’s quite clever.
Easily verified by creating another bunch of domains and using a browser that doesn’t do tracking - like waterfox
Metrics on what - how much beating can a server take before it commits ritual Sudoku and fries itself?
I use a similar setup with Traefik instead of Nginx PM, and Headscale instead of Tailscale. It is almost the same kind of setup.
You basically never want to expose your local network to the internet. The most secure and simple way are either Tailscale or WireGuard combined with a VPS that is exposed to the internet and takes all the beating.
Did it show the correct time? Did it tick correctly? From my observations, even clocks that looked okay had like wrong second speed, swapped hour and minute, or was rotated altogether.
Question mark says they are updated every minute, even gives you a prompt.
Watching for several minutes, none of then was good except Kimi K2. Sure, not every time, but solid third of them was actually working, while the others scored a perfect 0.
Also, as a Kimi K2 user (because Kagi), I approve. I don’t use Kimi K2 for coding, though, because JetBrains doesn’t offer that, but I use it in Kagi Assistant.
Also can be purchased via crypto for maximum privacy. They also offer something I eould call “AI Studio”, where you can easily switch models, and the chats are fully private and don’t share context. You can customize which domains to promote, demote, block, pin, and they have some nice bangs functionality.
I stopped paying for all streaming platforms, and Kagi is hands down the best bang for my buck as I already use it every day (in contrast of binging a show every two months).
It is there: No, No, Bleeding Edge, No
That’s what everyone calls GraphQL now lol
Oh, not sure about that honestly.
Super easy with Docker, and also quite portable. Usually is a copy paste and minor changes that irk me, but now I don’t have to explain my family what are torrents, how are torrents, what is a tracker, give them my credentials, and teach them to SSH to the server in order to copy a file let alone show them how to properly name it in order for Jellyfin to correctly recognize it.
All I have to do is log them in the app once, and tell them "If you want a movie, find it here, and it will probably be available in Jellyfin in and hour or two.
JetBrains is a company that, creates one of the most popular IDE for many programming languages. Although some of them are free, there is a paid option for 200€ for their full pack for a year (you can pay monthly, and you can choose a smaller pack or individual IDE). Also every year you pay the next one is cheaper.
They also have an AI agent Junie and an AI chat assostant, both currently running on Claude Sonnet 3.5 and 4 (can choose).
They also offer a free AI, which is running locally and can do very simple autocomplete and doesn’t support any chatting ability.
However, as you might know, AI usually needs some code to work with. This autocomplete AI can be enabled to run online as well, thus sensing your code to either JB or Claude.
Of course, both chat and agent require internet access (but all this online functionality can be disabled and everything can be connected to custom AI model running locally or elsewhere, except I think agent).
OP is implying that they want money for their IDEs, their AI, and gobble up code fragments.
CTRL+R for those unitiated
Docker runs abbhorently, instability and having to restart every week or so (tried Raspbian, Arch, Alpine, and all went unstable in a couple of days, even tried changing the SD and power supply).
Back on the day, sure, ARM was the thing you wanted if you went for maximum power saving, but I think there are comparable options that give more bank for your buck in the x86 space, especially if you want to repurpose it later.
In the end, go for whatever you want, you can always sell it later if you don’t like it. I’d just personally wouldn’t go for ARM anymore.
If I was buying a Pi, I’d go second hand.
Might be worth looking into ODroid H4+ (or any other really) especially into non-arm architecture (personally had more issues than I would’ve wanted).
Honestly can’t tell
For when setting up a kube cluster is too time consuming and you still need basic clustering.