- My Forums
- Tiger Rant
- LSU Recruiting
- SEC Rant
- Saints Talk
- Pelicans Talk
- More Sports Board
- Fantasy Sports
- Golf Board
- Soccer Board
- O-T Lounge
- Tech Board
- Home/Garden Board
- Outdoor Board
- Health/Fitness Board
- Movie/TV Board
- Book Board
- Music Board
- Political Talk
- Money Talk
- Fark Board
- Gaming Board
- Travel Board
- Food/Drink Board
- Ticket Exchange
- TD Help Board
Customize My Forums- View All Forums
- Topic Sort Options
- Trending Topics
- Recent Topics
- Active Topics
Dallaswho
| Favorite team: | Missouri  |
| Location: | Texas |
| Biography: | 5.5x Message Board Genius “I do think that Dallaswho MIGHT BE Drinkwitz, posting on SECRANT, not kidding.” -HRV |
| Interests: | Football, cars, building things. Failing at garden/landscape. |
| Occupation: | Oil/Tech |
| Number of Posts: | 3489 |
| Registered on: | 12/4/2023 |
| Online Status: | Not Online |
Recent Posts
Message
I’m dumb so I tie my actions/automations to virtual buttons/switches (called helpers in home assistant)
That actually may be the right way to do it but I don’t know. With helpers you can do anything.
That actually may be the right way to do it but I don’t know. With helpers you can do anything.
1
re: Anyone using Home Assistant AI agent?
Posted by Dallaswho on 3/2/26 at 6:37 pm to LemmyLives
Qwen3-30b locally but you could probably get with something smaller. Openrouter has a bunch of free or super low cost options.
Used extended OpenAI integration for LLM api and Wyoming for TTS and STT.
Setup is pretty easy but function has to be pretty much perfect and prompt should explain what you want it to do.just use chat.qwen.ai for that. It fixed everything for me when Gemini couldn’t.
Apparently if you use an esp32 device for STT there are options for speech cut off sensitivity, but not if you’re using the iOS app. They just need to go to a button or something if they can’t get it right.
Used extended OpenAI integration for LLM api and Wyoming for TTS and STT.
Setup is pretty easy but function has to be pretty much perfect and prompt should explain what you want it to do.just use chat.qwen.ai for that. It fixed everything for me when Gemini couldn’t.
Apparently if you use an esp32 device for STT there are options for speech cut off sensitivity, but not if you’re using the iOS app. They just need to go to a button or something if they can’t get it right.
Anyone using Home Assistant AI agent?
Posted by Dallaswho on 3/2/26 at 6:22 pm
This thing is so damn awesome but there is a HUGE problem. The speech to text listener, at least on the iOS app is completely useless unless you’re in a quiet room by yourself. It’s not that most attached STT models can’t easily cancel background noise but the app itself just keeps listening and listening until things are dead silent. That’s never the case around me. Completely useless. Still the agent is so good it’s worth typing.
re: Best cell phone and provider to receive unbiased feedback
Posted by Dallaswho on 3/2/26 at 6:05 pm to LemmyLives
DDG may help a little with privacy but it uses Bing and certainly won’t help with finding things.
re: Best cell phone and provider to receive unbiased feedback
Posted by Dallaswho on 3/2/26 at 5:50 pm to Royalfishing
It’s has been exactly 19 years(Google IPO) since you could search the web in an information-first biased manner. Ever since then, search has been purposefully designed to give you ads and poor results so that you have to stay on the search site longer to find anything remotely relevant. This is not a secret.
You won’t find a better platform than Apple for non-bias and privacy unless you go full open source and with a non-certified phone it will be very difficult to watch media and what not.
Things may be getting better though. You can try Kagi(paid) or Brave to search.
You won’t find a better platform than Apple for non-bias and privacy unless you go full open source and with a non-certified phone it will be very difficult to watch media and what not.
Things may be getting better though. You can try Kagi(paid) or Brave to search.
Akuvox VoIP endpoint shows weather. I guess that counts.
If you want something to follow music, there are a million LED things that do that and are way more immersive than a screen but the ones that work best for local control are mostly just RGBWW and no tricks.
If you want something to follow music, there are a million LED things that do that and are way more immersive than a screen but the ones that work best for local control are mostly just RGBWW and no tricks.
Not sure, precision are bigger hungrier. Might need 100w.
My 5430 runs fine w/ a little 30w charger all day but needs more than that to charge faster.
Here’s a 65w for $14
LINK
Here’s a 65w for $14
LINK
re: what have you done with AI today?
Posted by Dallaswho on 2/26/26 at 4:49 pm to TigahsOnTop
I gave up after a couple weekends. Works really good for messaging apps but most things I want automated are structured enough that openclaw is an asinine waste of resources. 90% of the time, even with opus, it just talked to itself.
quote:
s it python or something?
No but most containerized apps have both images available and the “data” folders are universal so moving the data over and changing the image name in the compose file completes the transition.
I don’t use any of that except jellyfin and dispatcharr.
For me, it’s just not having to reserve ram for a process and have it locked up without ballooning. Most my stuff is extremely busty. Any process might need 0-16G depending on what it’s doing.
Worst thing about proxmox for me was I built a lot of services (HAOS, apache2, nginx, ddclient, mqtt, and others) all on VMs without docker and even though it was easy to move those VMs to an old NUC8, I still either need to keep proxmox or run janky KVMs on Debian. If they were Docker to begin with, they’d just fit nicely onto any of my more modern NUCs and I could sell the old nuc8 or at least have it stop sucking power and making heat before summer. It’s just a Proxmox graveyard at this point.
For me, it’s just not having to reserve ram for a process and have it locked up without ballooning. Most my stuff is extremely busty. Any process might need 0-16G depending on what it’s doing.
Worst thing about proxmox for me was I built a lot of services (HAOS, apache2, nginx, ddclient, mqtt, and others) all on VMs without docker and even though it was easy to move those VMs to an old NUC8, I still either need to keep proxmox or run janky KVMs on Debian. If they were Docker to begin with, they’d just fit nicely onto any of my more modern NUCs and I could sell the old nuc8 or at least have it stop sucking power and making heat before summer. It’s just a Proxmox graveyard at this point.
re: anyone used AI to help set up radarr/sonarr/overseerr..
Posted by Dallaswho on 2/25/26 at 7:25 pm to notsince98
quote:
I enter one command and everything updates from repositories automatically. Seems much more simple. Differences in sizes are not trivial. The few apps I tested were under 20MB on their own but the containers were more than 5x larger.
I put off docker WAY too long in favor of VMs. I was so wrong. Docker is infinitely more resource efficient and an extremely organized way to keep software. Now for each machine it’s just /srv/[list,of,app-groups]/[group-list]-data and that’s my essentially entire directory tree that I ever use. Each app group gets its own docker-compose.yml and each individual app has one or two binds for data. Want to find a config? no /etc searching it’s just right there in that data folder.
Want to change machines?, just rsync the data directory binds and copy and paste (or rsync) the compose and you’re done. No dependencies except host drivers and docker itself.
You want to move a VM? sure it’s one file, but you’re stuck with a hypervisor forever.
Edit: docker containers can also be moved from x86 to arm64 with ease. Just replace the image path in docker-compose.yml.
Networking can get complicated but you’ll mostly just bind access ports 1:1. All AI chat bots know docker very well but none will correctly configure arrs or other self-hosted stuff unless fed the correct documentation.
re: Transferring Everything from Laptop to Laptop?
Posted by Dallaswho on 2/25/26 at 6:34 am to RetiredSaintsLsuFan
scp, rsync.
Any will be fine, don’t forget gluetun in your compose.
Also prefer chat.qwen.ai for this kind of work because you can point the bot directly at the official documentation and it will ingest several thousand relevant tokens to do things the documented way. This is rare for chat bots. Most hit safeguards or other policies and force the backend model to start hallucinating. If you use a lot of tokens too quickly it will defiantly degrade though.
Also prefer chat.qwen.ai for this kind of work because you can point the bot directly at the official documentation and it will ingest several thousand relevant tokens to do things the documented way. This is rare for chat bots. Most hit safeguards or other policies and force the backend model to start hallucinating. If you use a lot of tokens too quickly it will defiantly degrade though.
re: Starlink deal to jump on now! Deep dive don't read if you have attention deficit disorder.
Posted by Dallaswho on 2/24/26 at 3:31 pm to idlewatcher
All I heard was high latency and CGNAT. Not sure if he actually wrote that, but it’s what I heard.
Finally made a home assistant bot. Used qwen3-30b and extended OpenAI integration. It handles complex instructions WAY better than Alexa once you get prompt and function structure correct. Took me a couple hours to figure out default “execute_service” function sucked. Now I can ask it to rotate colors and make 20% brightness on 20 bulbs all at once or say make the red ones green. Alexa screws things like that up almost every time.
PSA: IPT is exact same service as mybunny and private iptv access .com.(pia) Dogshit for at least 2-3 hours per day. Otherwise best selection on earth. Just don’t count on it being a TV replacement if you have a wife.
re: Re-setting old work laptop
Posted by Dallaswho on 2/18/26 at 3:29 pm to GeauxldMember
Couple things:
You don’t need to buy w11 before installing and it will never force you to enter a key.
Admin password usually refers to a password in the UEFI that prevents changes. (Like booting from a thumb drive). The only way around modern ones is to reflash the bios chip which often involves soldering.
Before modifying, need to make sure you ca enter system menu and enable boot from external drive. If you can do that, then you can install whatever you want.
If you’re locked out of changing the boot order, then you might be able to get a system board for like $40 or so on eBay assuming it was a w10 PC. Thats how I got my current laptop and am in it for like $100 for a perfect 12th gen i5 latitude.
You don’t need to buy w11 before installing and it will never force you to enter a key.
Admin password usually refers to a password in the UEFI that prevents changes. (Like booting from a thumb drive). The only way around modern ones is to reflash the bios chip which often involves soldering.
Before modifying, need to make sure you ca enter system menu and enable boot from external drive. If you can do that, then you can install whatever you want.
If you’re locked out of changing the boot order, then you might be able to get a system board for like $40 or so on eBay assuming it was a w10 PC. Thats how I got my current laptop and am in it for like $100 for a perfect 12th gen i5 latitude.
quote:
I know it seems extreme to the average user, but I would never buy a laptop/prebuilt to not then immediately follow up with a clean OS install. Way too much bloatware comes standard.
You may be out of luck. Microsoft new driver policies mean if you don’t have a desktop or business laptop, then your non-certified drivers won’t be in the OS image. If you have a modern budget laptop, your options are spend hours to get the NVMe to work or use the OEM image installer from the machine’s UEFI.
So my vision experiment went worse than just using yolo-world. Still might work as a verification step though. Anything to avoid tagging images manually.
That’s the thing about doing novel things with AI: It either fails, or there’s a GitHub with a way better solution within a few days. If you couldn’t find a good tool for something a week ago, it’s time to look again.
That’s the thing about doing novel things with AI: It either fails, or there’s a GitHub with a way better solution within a few days. If you couldn’t find a good tool for something a week ago, it’s time to look again.
re: Help me buy a central HVAC system - best price / reputable source for drop shipped system
Posted by Dallaswho on 2/12/26 at 10:52 am to DownshiftAndFloorIt
eBay another option. Under $4k after tax. Just make sure get 80-81% furnace if it vents through a metal pipe. The 90+ ones need special exhaust.
Popular
17
