curbstickle
@curbstickle@anarchist.nexus
- Comment on [AIP] Modern offline-first self-hosted open-source TickTick/Todoist alternative. Will Be Done v0.15 released, now with calendar, mindmap and openapi support 3 hours ago:
Its part of our rules here - take a look at rule 8.
It was recommended for !opensource@lemmy.ml not that long ago iirc, though I don’t know that they’ve adopted it.
- Comment on Update all your stuff, NOW - again. 4 days ago:
it goes back a while
It goes back at least to my teen years, which is…. not remotely recent.
- Comment on The docker group is root access. An agent used it to write to /etc 4 days ago:
One of the reasons my pref is unpriveleged lxc.
- Comment on Looking to offload Plex from my gaming PC 1 week ago:
Definitely suggested then, you’re also going to get drastically more features. If you remember a time when Plex plugins were available - its that. And more flexible.
Any intel-based chipset past 6th gen or so will be ok, 7+ recommended, 10th+ for AV1. If you aren’t doing any transcoding, just about anything will work include a pi.
- Comment on NUC recommendations please 1 week ago:
This is what I’d do, but youre going to be limited to 2.5” + m.2 in the ultra small form factors (USFF, more NUC sized), while the SFF size you can get one (sometimes two) 3.5”. So make sure to check the form factor of the machine!
Tons more computing power in either case, and the most reasonable price point IMO.
- Comment on It's self hosting Sunday! What's up? 1 week ago:
Only a little as far as animal health goes, my wife knows a lot though, and she knows quite a few vets who have let me abuse them with questions :) Majority is as I go along though for the non-tech side.
- Comment on It's self hosting Sunday! What's up? 1 week ago:
Similar but not the whole thing. I’m identifying more broadly (deer, horse, chicken, etc) but more focused on their movement and condition. Are their coats are healthy, their movements normal, so on.
Classification then changes which CV model is used to evaluate the rest.
Right now they are too big to be fast enough, and going smaller is breaking the model. N9t because there aren’t unnecessary bits to trim out, but because I’m not doing a good enough job at minimizing what’s actually needed.
I’ve been working on it for a while now, but still have a long way to go unfortunately.
- Comment on It's self hosting Sunday! What's up? 1 week ago:
In terms of my setup, there was basically no impact. I use Debian (servers and some workstations) and Arch (couple of laptops and a couple of miscellaneous task-oriented devices I want more recent builds on than Debian will give me)
Work? Different story, so much manual updating!
Otherwise just regular stuff this week like backups (monthly incl since yesterday was the 31st), reviewing changelogs for minor updates, etc.
Hit a bit of a wall in my animal identification tools I’m making, so I’m going back to do some more reading to see if I can find a way to improve efficiency more and make it smaller while maintaining accuracy.
- Comment on [AIT] Creepy crawlies (AI crawlers on git.kernel.org) 1 week ago:
Client dependent as to whether it shows the community its cross posted from.
Mostly because that crowd likes to come in and downvote anything they notice ai-related in any way, shape, or form, or they post comments I’ll need to remove. Despite tags being added so they could choose to filter those posts out instead.
Its less about the post and more about the general behavior of some people in that comm.
- Comment on [AIT] Creepy crawlies (AI crawlers on git.kernel.org) 1 week ago:
I’d prefer a new post there over a cross-post from here, personally.
- Comment on [AIP] Reitti v5.8.1: Complete Statistics rework, Intervals.icu fitness sync, and security hardening 1 week ago:
I love seeing these updates Daniel! The new statistics page looks great, I’m going to give my local instance an update and check it out
- Comment on Figuring out what to do with an old Mac Mini (Base M1) Linux? 2 weeks ago:
Asahi is great, you can just run whatever.
Mine is mostly a host for my self- created small scale models, but its also a great jellyfin server. The M series is pretty solid for that, or really anything that has GPU loads is good (immich f9r example).
The m -series is great in terms of low power and solid output, highly recommended.
- Comment on Tempus v4.31.0 android subsonic client release 2 weeks ago:
Since your last post, a few rules have been added, specifically adding the appropriate tag in the title, and an ai disclosure if required (and with Claude as a contributor, it is).
Please take a look at rules 7 & 8 and update accordingly, then send me a message.
- Comment on Reloop: a self-hosted CRM that builds itself from your mailbox (AGPL, Docker) 3 weeks ago:
Please update the title with [AIP]
- Comment on HomeAssist Kiosks 3 weeks ago:
With my particular unit? Its not going to get any updates ever, the only way to install is over ADB, and changing the running app requires script changes and a reboot (or changing things to allow the on-screen navigation buttons, but that comes with other annoyances).
So sdk version isn’t changing, which is a potential compatibility issue going forward.
- Submitted 3 weeks ago to selfhosted@lemmy.world | 0 comments
- Comment on Is anyone self-hosting their own AI? 3 weeks ago:
As a service? No, not really.
I use llms for specific tasks, usually small models, for things like identification and tagging (immich, paperless), translation or transcription (whisper), etc. I have a bunch of larger models on a Mac mini and my work laptop for various testing purposes, and have used them to make summaries and other such things.
They aren’t great for search IMO, and search summaries can be quite bad.
I do find them useful to interrogate an unknown rs232 protocol or similar, though, which is where larger models come in handy.
Corporate LLMs are generally garbage to me not only because they are just data farming, but also because they really aren’t good at everything. They might claim it, but the reality is a tailored model is often the better approach.
Using openwebui as a front end for llama.cpp is a decent strategy, combined with sub-agent routing (so generic model hands off to tailored models then returns the right info). This can be useful for some scenarios.
A caveat on coding - IMO, they are mostly bad at it. If you are using it to generate a small part of something that needs to be repeated they are handy, but at scale they aren’t great, meaning for the entire project (or even an entire plugin or module, depending on complexity). Maintainability becomes a serious issue. While they have improved somewhat, there are limits to what they can do.
Macs with M series chips are good options for a hardware base since the memory is shared, and IMO is one of the most cost effective options over a dgpu with a ton of memory. A dgpu with a decent bit of memory (say 8-12GB) is a solid choice for a model thats larger (27b with 3 or 4b active as a mixture-of-experts model) especially if you have an iGPU for regular graphics tasks. Since its all shared on the M series macs, just get something that fits entirely in memory there.
- Comment on PSA to Home Assistant users 3 weeks ago:
No to both.
Its a long-standing and documented feature. Included in that documentation is how to turn it off, which was added a few years back iirc.
Its brought up every once in a while by people.who don’t understand cryptographic hashes or the hibp tools. Usually with a bunch of posts about how passwords are being leaked (they aren’t).
- Comment on HomeAssist Kiosks 3 weeks ago:
Generally speaking, I use the app.
I do, however, have miscellaneous tablets that are in rooms to show information and have a designated account just for that room’s controls. I also have a wall-mounted touch panel that was from a pro control system that I picked up cheap that I rooted and set HA to open at boot, but I don’t recommend this route unless you happen to get/see one cheap and know exactly what can/can’t be done with it.
- Comment on PSA to Home Assistant users 3 weeks ago:
any credentials in your config have been hoovered up
Anyone reading this would interpret it exactly as @frongt@lemmy.zip did. Its well passed “inflammatory” and hovering around “misinformation”.
- Comment on Framework expands funding support for CachyOS Linux 3 weeks ago:
@mntmn@mastodon.social for Lucie of MNTRe, regularly posts what’s being worked on too
- Comment on Framework expands funding support for CachyOS Linux 3 weeks ago:
They have no intention of stopping their support, so I have no intention of supporting them.
For anyone looking for an open design, MNT Research seems the best to me, and they are on the fediverse. If anyone has other options that are on the same level, I’d love to hear about those too.
- Comment on Framework expands funding support for CachyOS Linux 3 weeks ago:
Have they stopped supporting omarchy and hyprland though?
As of their last update in March… Nope, still supporting them.
- Comment on A thumbnail generator for your backend. 4 weeks ago:
Much appreciated!
- Comment on A thumbnail generator for your backend. 4 weeks ago:
Just add an [AIP] tag to the title of the post and you’re all set! :)
Considering its the only part that is, its covered by the “miscellaneous” approach which is what you’ve got in the disclosure, so just a tag to go.
- Comment on A thumbnail generator for your backend. 4 weeks ago:
Please take a look at rules 7&8; tag and disclose as required
- Comment on Lidarr not finding release / group that exists in MusicBrains? 4 weeks ago:
Unfortunate, but also happens sometimes
Sorry I don’t have a better answer!
- Comment on Lidarr not finding release / group that exists in MusicBrains? 4 weeks ago:
So thats the caching portion not being 100% - using the ID for the search, try it twice, that has worked where the first fails but the second query shows up, specifically for the id you want
- Comment on Lidarr not finding release / group that exists in MusicBrains? 4 weeks ago:
https://github.com/Lidarr/Lidarr/issues/5498
AFAIK still not a closed out issue.
- Comment on How do you monitor your servers? 4 weeks ago:
Well I used to use VGA for my monitors, now its mostly displayport.
Seriously though, prom+Loki+alloy+a few other things to Grafana, with alerting in grafana.