Tonight was testing out a fan controller, Lutron aurora, zigbee remote all working together in a room. I was not expecting Lutron aurora to have issues. The major issue was not seeing button events on zha. Changed the device to use a hue hub with no further issues.
Not happy with such repair being done on my refrigerator when a new led module is $8.39 online with 10.99 shipping which is cheap for appliance parts. The connector ripped out even more will be neeed. Also mad more of my time will be spent fighting for job to be done right
wasn't expecting a electronic lock I researched to have issues. Was attempting avoid issues with existing electronic locks I own. Spent hours Tues night installing/debuging only to end up revert to a older lock late in the night.
I really wish google assistant would not do "by the way" responses. Really don't like the by way is telling me I could use a button on the steering wheel to start a conversation when steering wheel button started the conversation to begin with.
slightly annoyed my mom thought I was messing with her phone with audio emoji. She was accidentally touching the emoji buttons with her face by her during phone calls. Audio emoji doesn't make sense to begin with so turned it off on her phone, my dads phone and my phone.
been looking at how Bluetooth lower energy area tracking works with 5 ESPresense nodes and 4 bermeuda scanners. Working well enough I think it might out well as a layer for presence detection when toggling lights. Since PIR and mmwave can have issues with sleep.
Interesting only the oldest sub I have showed as impacted. I don't see the bits showing after doing what was asked. Oh well, At least doing a cancel then sub did not seem to send anything to chat like I thought it could.
Still leaps and bounds better when years I ago attempted the same thing with the mycroft mark II years ago. Unfortunately the mark II hard bricked otherwise I might of attempted a convert to a wyoming voice satellite.
can't more find tuning for voice PE. Doing light commands as backup is there. Full out replacing current voice assistance isn't there. Hopefully that changes. Cause I don't ever want to hear "I am having trouble connecting to internet" when my parents use voice assistance.
Wow Ollama was much easier to setup for HomeAssistant. Response time did get longer on the 1080. Did find saying lights sometimes does more lights than the satellite area. So guess prompt research is next.
over the weekend put together a dedicated wyoming piper and wyoming whisper pc with a gtx 1080. Good response times still. Family wide test failed to understand "set the lights to x%". Going to see if I can get a llm to translate the statement to "set brightness to x%".
Got more time to work on the crashing issue. Setting acompile_process_limit set to two seems fixed things. Getting closer to full adoption/acceptance of fully local voice assistant.
having ESPHome Device Builder updating voice PE device brick my whole home assistant PC just to add different common wake word doesn't seem worth it. Will probably wait for openwakeword running outside a voice PE device to hopefully be supported in the config options.
I was able to get wyoming-whisper-gpu and wyoming-piper-gpu docker images running on my windows pc. Finding dockge helped greatly. Happy with the voice assistance speed responses in testing . Need to do more hardware testing to determine if dedicated hardware is a must.