I don’t know if the content is of particular interest to everyone, but I did want to follow the topic as it was a source of disagreement between FUTO and Rossman and I wanted to see whether they had come to terms.

Do you have a copy of the now deleted/removed video, or are you referring to a videoviseo which is now publicly available? If it is now publicly available, is it at the same link as my edited original post or something else? Thanks!

Perhaps it would be worth checking with the !homelab@lemmy.ml community and see if they have any advice? They are more likely to have dealt with such situations, by nature.

Good luck on your defense! Goodspeed

It’s pretty new but if you are interested in degoogling and looking for ways to use android auto, there is apparently a working solution with GrapheneOS but I have not tested so can not verify it.

https://grapheneos.org/features#android-auto

Edit: typo

Is your family unhappy because you use Linux and Lemmy?

Can you elaborate on why it is a bad security practice? It’s the first time I’m reading about it and I’d like to read more about it. Thanks!

You are right and I would add that this is even a privacy and security measure, to make use of wildcard certificates. The reason is, those subdomains will be public because of websites like crt.sh which show all subdomains which have their dedicated certificate. Obfuscation can be helpful in not disclosing which are some services or naming schemes you use for yourself even if it is only meant to be for internal use.

What are your actual requirements? Are you looking for something specific?

Have you considered second hand? I’m not talking second hand specifically for framework or other mentioned brands, but just in general.

I feel it is not yet normalized to consider second hand for electronics, yet you can find quite some good deals. Not everything needs to be bought new, especially if you are price sensitive. One generation or 2 older hardware bought second hand can be better and cheaper than new.

Fedora is still pretty frequently and recently up to date with respect to packages and kernel, not sure you’d be losing much over arch.

But the debate to me is also not that important, I’ve been running fedora and have at some few occasions gotten some instabilities due to updates (mostly Nvidia with Wayland) so I can totally understand someone wanting stability and reliability over bleeding edge).

I have heard of shadowsocks for this purpose. I have not tried it myself but I recall having read it being used to hide VPN traffic behind the great firewall. A brief intro to it here:

https://errande.com/obfuscate-wireguard/

I tried truenas and was at first enamored but felt quite limited when it came to using docker compose. There are no official ways for using it. Additionally because it uses kubernetes under the hood it’s not as easy to go under the hood to make it work, as it requires knowledge of that in addition to docker. There is an unofficial docker compose package but there is almost no documentation that I could find and it seems to be largely unmaintained nowadays.

The other OS are good candidates, though through my research, if you are after reliability, people seem to prefer Debian with its more robust validation of updates.

I’m here mostly because I’m setting up my own homelab and I’m preparing for making use of Alpine Linux, which has no desktop environment which makes it particularly fast to boot. Additionally it is very robust, has decent documentation and allows for quite some customization but requires to learn a bit about how to make it work. With its barebone structure, it is more secure as it has a smaller attack surface and you only install what you need. Perhaps an alternative for you to consider?

Edit: to keep some sanity while needing to manage your homelab, make sure to put some weight on an os you are familiar with. I have juggled for months between OSs to find what does what I want but it has been a long battle of learning everything from scratch every time. Having tried proxmox, truenas, xcpng, and now hopefully settling with Alpine and zfs. Go the more sane way, any of those first three will work just fine.

With HoloISO? Is it compatible? I think?

https://github.com/HoloISO/holoiso

I think there’s some truth to it. But I imagine it will be more AOSP than what android is with google services. AOSP is really a great operating system with very good security and built in features. And with neural engines and high bandwidth emmc, it is mostly just lacking a large amount of storage to make it all complete, but the latest SOCs are most probably powerful enough. Is there something like docker for android? :)

Edit: I do recognize what someone else said, which is that one big challenge would be software updates. We are not that limited by hardware when we consider servers, they can easily run for years with regular software updates.

Thank you, I’ll see if this new information can help me pinpoint the video.

I also found this which could be of interest:

MLC-LLM, which “Enable everyone to develop, optimize and deploy AI models natively on everyone’s devices.”

Here used to deploy Llama-2-13B on the RX 7900 XTX:

https://blog.mlc.ai/2023/08/09/Making-AMD-GPUs-competitive-for-LLM-inference?ref=upstract.com

AMD has ROCm which is available on And Radeon Instinct GPUs (server GPUs) and some consumer GPUs. You’d need to double check whether your GPU supports ROCm.

It seems there is some discussion happening here on the use of ROCm with Whisper: https://github.com/openai/whisper/discussions/105 And here (showing it might be possible?): https://github.com/openai/whisper/discussions/55

OpenAI’s Whisper model is a really great one, supports many languages ans translationa and is available both as a pretrained model (https://github.com/openai/whisper) which can be selfhosted and as an API (https://openai.com/pricing).

I suppose you can split your content in 3 categories:

• text
• audio
• image

For text, you can use Langchain which allows to get embeddings from text (read more here: https://js.langchain.com/docs/modules/data_connection/text_embedding/).

For images, you can use CLIP (this model is open source, from OpenAI). You can read more about it here: https://github.com/openai/CLIP

For audio, I don’t know anything off the top of my head but you are likely to find something even open source similar to the above I mentioned.

I read somewhere that this required connecting to Microsoft’s cloud? Is that true?