Unfortunately, nothing is standard. So I would say, across all the configs you looked at, which had a file and module structure that you understood? I’d follow that then.

My config has a users and hosts dir, to distinguish home manager and nixos configs. Inside each is the list of users and hosts configuration files. In addition, there is a modules folder that holds modules that are common among different users/hosts.

I think this is good idea. If the modules/options you are writing are for internal use, and not expected to be shared with the wider community, then this is great. I should incorporate this in my own config, but I dont know if this is common practice.

Wow thats a crazy trackball! What brand/model is it?

Okay this looks big as far as data, but is there any information here that isnt (wasnt) already public? This looks like profile information. Isnt all this already available?

That looks like the same idea for a different set of tools. That’s a great reference, thank you!

Pff, if pandas gets me numpy that works that may not be a bad hack. I’ll try this! Sorry I dont know how to fix qt!

I’ve had the same problem running numpy. Shockingly with a library so popular I havent found a way to make an environment with it work. I also had the most success with poetry, so I think you’re on the right track.

That’s right, “text-generation-webui”. At least its unambiguous lol. Thanks for sharing.

What GPU are you using to run it? And what UI are you using to interface with it? (I know of gpt4all and the generic sounding ui-text-generation program or something)

I think it overinflates the click rate, which means despite having more clicks on an ad, that doesnt mean that more people bought some product. This devalues click rate which might make the ad service less valuble to advertisers, so they dont spend as much on Google’s ad service.

And in general I think makes any training data for a model more muddy, since adnauseum isnt behaving like a human. So it could make it more difficult to train models that do targeted advertising.

As a community, I do think we get hungup on distros. Most of them, as you mentioned, are just different defaults of the same packages.

But at the maintainer level, I do think theres a lot of work distributions do at making sure the software they choose as defaults are up to date, secure, and work with one another. I dont enounter it often, but relying on maintainers to prevent mismatched depencies ending up in the day-to-day linux user has to be worth something. And every set of defaults needs that level of assurance, I would think. Im not a maintainer, I could be off here.

What database client do you use? Maybe a plain database is enough with the right client.

Doesn’t need to. That’s a plus though. I think the features I like the most are dropdowns for foreign keys and more specific column types. For instance, a date type gives me a calendar picker, and an image type lets me upload and image and then see it as I browse the data.

Yes everyone would need a client (probably?) but after having recently set it up the first time, its incredibly simple.

You can also use p2p mesh vpn services like zerotier or tailscale to establish a direct connection without opening any port in the router at all.

deleted by creator

This is a very specific project. This is cool to see. Im curious if anybody else would use this.

Ive been playing with the nixified.ai project, which packages two web interfaces for LLMs and image generation. Im also looking into Tabby.ml for code assistant as well. I haven’t gotten deep, but these all look like promising options for utilitizing a server’s hardware but offering the functionality across the network.

Thanks for sharing this codec wiki. Looks like an incredible project.

Futhark is another language with the same goals, executed differently.