I thought it was pretty weird, so I tried again a while later; same result. I checked through issues and couldn’t see anything, I figured it was just me. I tried because NPM was just full of error logs and was having some sort of shitfit, so I blew it out and rebuilt from scratch, now all is fine. But the NPM+ defeated me. I might have to try again just because.

NPM+ didn’t work worth a damn for me. No proxies would forward, I have no clue why and couldn’t figure it out. It was like I was turning knobs that weren’t connected to anything. But YMMV.

Long ago I learned to stop convincing people to stop shooting themselves in the foot.

It needs to be running on postgres and redis. The AIO is your best bet

Maybe you should just count yourself lucky that you get to control your own fertility. In civilized countries at least.

Thank you. Body odor is gross, but it doesn’t give me an insta-headache like the chemicals people douse themselves in every day.

Just the 5-6 raid modes are shit. And its weird willingness to let you boot a failed raid without letting you know a drive is borked.

I can buy it with no scent. I wish to fuck other people did. If you put this junk on yourself every day, and put more and more on until you can actually smell it after your olfactory nerves have been burned out by years of such abuse, think about what that’s doing to people that are actually sensitive to this shit.

I’d rather smell unwashed gamer than the overpowering perfume in these bloody deodorants.

In the docker-compose.yml for the AIO, you can specify the port that the Apache server runs on with APACHE_PORT=XXXX

Then you set your reverse proxy to proxy to that port. If you need some pointers about how to use a reverse proxy, ask away.

That’s what I was wondering too.

What’s your general self-hosting setup and what machines are you building for that? I’d like to have HA Proxmox running all the time on three nodes with a low power bill and lots of memory available (like 256GB) but space for memory seems to be difficult to find in a reasonable priced consumer board.

10k charge cycles isn’t revolutionary. LFP do 8k and even then they just drop down to 80% of original capacity.

Like a real Scotsman.

Be me, build a Dockerfile to do all that shit in a nice CI package, still can’t run it.

You’re not my supervisor.

Isn’t she the same spinny bitch that wanted to sell mouse subscriptions?

Good to know, thanks for the PSA.

OpenWebUI is pretty much exactly what you’re looking for. It can start up an ollama instance that you can use for your other applications over the network, and chat with it as you see fit. If you have an API key from an outside subscription like OpenRouter or Anthropic, you can enter it and use the models avaialable there if the local ones you’ve downloaded aren’t up to the task.

Somewhere, sometime, ChatGPT is telling the stupidest person in the world “Yes, you are absolutely correct!”