I use the HA Voice Preview in two different rooms and got rid of my Alexa Dots. I’ve been trying both speech-to-phrase and whisper with medium.en running on the GPU for STT, tried llama3.2 and granite4 for the LLM with local command handling

I’ve been trying to get it working better, but it’s been a struggle. The wake word responds to me, but not my girlfriend’s voice. I try setting timers, and it says done, but never triggers the timer.

I’d love to improve operating performance of my assistant, but want to know what options work well for others. I’ve been experimenting with an intermediary STT proxy to send it to both whisper and speech-to-phrase to see which one has more confidence.

I’d love for my HA Voice Preview to be sufficient to replace my Alexa/Google devices. I even unplugged my Alexa devices. However, it’s been rough going for me. It never responds to my girlfriend speaking the wake word and doesn’t set timers. There’s a number of knobs that define how well it works including the physical hardware (there’s obviously the Voice Preview, but also some community made versions with better mics,) wake word model, conservation LLM model and the speech to text model (whisper vs speech to phrase). If it works well for you, can you share your configuration you’re using?

That’s the option to publish it. I was curious about the aggregated results.

Is there a place to view the device database yet or is that coming soon?

EDIT: Found it here

I use it to play music from Jellyfin to my Sonos speakers. It won’t fix a Jellyfin library that has bad data, but it can pull in music from multiple different sources and push to different players.

It works well enough. Some issues where songs get interrupted, but I think that’s issue with the Music Assistant/Sonos integration.

Just an update. Firefox 146 just dropped with:

• Firefox now natively supports fractional scaled displays on Linux (Wayland), making rendering more effective.

After upgrading to 146 and natively using Wayland, it feels faster. Some fade animations are still choppier, but on average it’s at least tolerable.

Thanks for the suggestions everyone.

Interesting. I played around with X11 vs Wayland settings just to see what different configurations give me

• MOZ_ENABLE_WAYLAND=1 /snap/bin/firefox - Exhibits low FPS issue
• MOZ_ENABLE_WAYLAND=0 DISABLE_WAYLAND=1 /snap/bin/firefox - Actually feels fast like it should be. Most animations feel faster, some are still choppy though. It’s hard to tell.

It seems like running with X11 sort of the problem? Which seems unexpected and concerns me since I know distros are starting to default to Wayland.

Yep, both are plugged into the graphics card. Other programs and games are a lot faster.

If the app is just a WebView wrapper around the application, then the challenge page would load and try to be evaluated.

If it’s a native Android/iOS app, then it probably wouldn’t work because the app would try to make HTTP API calls and get back something unexpected.

Unless you’re running VLANs, in which case the inter VLAN is normally handled by the router. I also expose my home lab services over BGP so all my traffic hits the router then comes back to my lab services.

https://forum.syncthing.net/t/discontinuing-syncthing-android/23002

According to this post, it was partly that and lack of maintainers. Given there’s maintainers for a fork, I’m curious why they didn’t bring them into the main project.

Reason is a combination of Google making Play publishing something between hard and impossible and no active maintenance. The app saw no significant development for a long time and without Play releases I do no longer see enough benefit and/or have enough motivation to keep up the ongoing maintenance an app requires even without doing much, if any, changes.

We’re sort of in this situation because the official project decided not to continue providing an official Android app, yet people want to use it on Android forcing unofficial versions to be created and maintained.

I get that they don’t want to deal with Google Play anymore, but somebody has to deal with it and them not owning the app is putting users at risk.

I tried that with my KeePass database and then I ended up with different conflict versions if a device didn’t sync to my server before I updated on another device. Then I ended up with the conflict versions and old versions I had manually resolve. The Android Syncthing app could get auto killed by the Android OS for memory/battery saving, so I had to go back to OneDrive style syncing where the KeePass Android app would internally merge and resolve conflicts. It couldn’t do that using the Syncthing style syncing.

How would that work? The use case is for previews for pull requests. Somebody submits a change to the website. This creates a preview domain that reviewers and authors can see their proposed changes in a clean environment.

CloudFlare pages gives this behavior out of the box.

It is for pull requests. A user makes a change to the documentation, they want to be able to see the changes on a web page.

If you don’t have them on the open web, developers and pull request authors can’t see the previews.

The issue they had was being marked as phishing, not the SSL certificate warning page.

The MinisForum B550 is what I use for my own setup and it works pretty well.

I was surprised to see this. I encouraged friends to get the Yellow because of the PoE support that the Green didn’t have IIRC.

For what it’s worth, they reversed that policy

That probably makes sense. Most people don’t have the risk of a malicious actor sniffing their TPM bus and those that do can opt in. Everybody else can still benefit from a TPM

That’s a safe assumption given the post is about entries in the logbook and that depends on the recorder.