Just a stranger trying things.
There is no guarantee headscale can keep working the way it does or that it is allowed to keep existing.
Edit: FYI headscale is not at all at feature parity with what tailscale offers.
Congrats! Amazing project, exciting interface and you went the extra mile on the integration side with third parties. Kudos!
Edit: I’ll definitely have to try it out!
Perhaps give Ramalama a try?
Indeed, Ollama is going a shady route. https://github.com/ggml-org/llama.cpp/pull/11016#issuecomment-2599740463
I started playing with Ramalama (the name is a mouthful) and it works great. There is one or two more steps in the setup but I’ve achieved great performance and the project is making good use of standards (OCI, jinja, unmodified llama.cpp, from what I understand).
Go and check it out, they are compatible with models from HF and Ollama too.
I will always remember my first experience using MacOS: I am comfortable with computers and a relative needed help with their recently purchased macbook. I had plugged in a USB stick to transfer some files and was done and wanted to eject it. I spent way too much time than I care to admit, trying all possible options, right-clicks, settings, everything imaginable, to eject the damn thing.
It was impossible to me to find the simplest operation with a USB stick, something required to operate it. I capitulated and looked online. The solution? I had to drag and drop the USB stick icon into the trashcan!?!?!?
To this day, I will never understand the absolute ridicule of this and I will never comprehend how anyone is expected to figure it out on their own. And this is from the OS touted as the most user friendly and intuitive. Go figure.
Edit: this was a long while back, no idea how it is nowadays.
Don’t give them ideas!
Running on prem is certainly possible, but requires a dedicated sysadmin team for anything serious. It is very important to be able to have availability guarantees and some expert you can count on to solve your problem with a phone call.
Related: Uber recognized how battery levels affect customer psychology in pricing but claims it doesn’t use it to hike prices for low battery devices:
Same, I rocked a second hand GTx 680 from 2012-2013, which I upgraded to a second hand RTX 3060 12GB for a fantastic price, in 2024. Still rocking a DDR3 platform with the intel i7 4400K. And that’s more than enough for most games with nice graphics on 1680x1050 :) (display probably 15 years old too). Eventually, I will be looking for some other second hand components to upgrade the rest of the system, but it does everything more than well enough.
Would you be able to share more info? I remember reading their issues with docker, but I don’t recall reading about whether or what they switched to. What is it now?
It seems Signal has already pushed out a fix for this, which was abusing the QR codes to actually link a device when it was presenting itself as a way to join a group.
Paywalled: https://www.wired.com/story/russia-signal-qr-code-phishing-attack/
I hear you, I always see this problem being solved by the link being in the description and the host saying “link in the description”. I hadn’t come across a situation where an audio only format was accessible and there was no way to interact with the content but in some corner cases it does make sense.
I don’t understand in what circumstances anyone would like to use link shorteners? I can only find reasons why not to use them:
One thing which I find useful is to be able to turn installation/setup instructions into ansible roles and tasks. If you’re unfamiliar, ansible is a tool for automated configuration for large scale server infrastructures. In my case I only manage two servers but it is useful to parse instructions and convert them to ansible, helping me learn and understand ansible at the same time.
Here is an example of instructions which I find interesting: how to setup docker for alpine Linux: https://wiki.alpinelinux.org/wiki/Docker
Results are actually quite good even for smaller 14B self-hosted models like the distilled versions of DeepSeek, though I’m sure there are other usable models too.
To assist you in programming (both to execute and learn) I find it helpful too.
I would not rely on it for factual information, but usually it does a decent job at pointing in the right direction. Another use i have is helpint with spell-checking in a foreign language.
Regarding photos, and videos specifically:
I know you said you are starting with selfhosting so your question was focusing on that, but I would like to also share my experience with ente which has been working beautifully for my family, partner and myself. They are truly end to end encrypted, with the source code available on github.
They have reasonable prices. If you feel adventurous you can actually also host it yourself. They have advanced search features and face recognition which all run on device (since they can’t access your data) and it works very well. They have great sharing and collaborating features and don’t lock features behind accounts so you can actually gather memories from people on your quota by just sharing a link. You can also have a shared family plan.
Ollama, latest version. I have it setup with Open-WebUI (though that shouldn’t matter). The 14B is around 9GB, which easily fits in the 12GB.
I’m repeating the 28 t/s from memory, but even if I’m wrong it’s easily above 20.
Specifically, I’m running this model: https://ollama.com/library/deepseek-r1:14b-qwen-distill-q4_K_M
Edit: I confirmed I do get 27.9 t/s, using default ollama settings.
You can. I’m running a 14B deepseek model on mine. It achieves 28 t/s.
Ollama is very useful but also rather barebones. I recommend installing Open-Webui to manage models and conversations. It will also be useful if you want to tweak more advanced settings like system prompts, seed, temperature and others.
You can install open-webui using docker or just pip, which is enough if you only care about serving yourself.
Edit: open-webui also renders markdown, which makes formatting and reading much more appealing and useful.
Edit2: you can also plug ollama into continue.dev, an extension to vscode which brings the LLM capabilities to your IDE.
I’m not convinced having a specific ruling targeting just tiktok would be a good move. I think I understand and side with all the reasons they are doing it: be it privacy, security and more. But they should not uniquely apply to tiktok, but to all apps. If suddenly there’s a Russian app, or a Nigerian app or a swiss app, any which for whatever reasons has security and privacy concerns, they should all be held to the same standards. Me no like double standards.
I feel you are a bit out of touch when the topic is specifically enshittification and that it is based on the history of companies turning against their users, showing little good faith. It is also not something which is sparing open source projects (remember bitwarden’s attempt?). So sure, I’m not going to deny that I’m making assumptions and that I am concerned it may one day happen. But it is grounded in reality, not some tinfoil hat stuff.
Edit: and the fact that bitwarden did not eventually go through with it does not counter the fact that they intended to and tried. Sometimes companies back off and play the long game and try to be more subtle about it.