I used the llama 3.2 3b model for a while. It ran Okeyish enough on a laptop with gtx1050 (about 10s to 2 minute response time). I’ve personally opted to go without ollama for now though as the automations and build in functions in the voice preview edition are more than enough for me at the moment especially with recent updates.
I’ve only set this up with add-ons. I’ve tried the “nginx home assistant SSL proxy” and the “nginx proxy manager” add-ons and have the problem with both. With the nginx home assistant SSL proxy I don’t seem to be able to set a toggle for web socket support. With the nginx proxy manager I did specifically toggle the web socket support.
I really appreciate your help by the way :)
Yes, I did. I’ve added the IP address stated in the documentation and also added the localhost IP 127.0.0.1 This didn’t help any though
When connecting locally it isn’t a problem. In the logs there isn’t much that I can figure out as being the problem except that it somehow probably has to do with the http integration. My reverse proxy is setup with an HTTPS connection between external connection and proxy server and with http between proxyserver and home assistant. As far as I’m aware that should be good.
Logger: homeassistant.components.http.ban Source: components/http/ban.py:136 integration: HTTP (documentation, issues) First occurred: 16:58:44 (2 occurrences) Last logged: 16:58:44
Login attempt or request with invalid authentication from 178.230.xxx.xxx (178.230.xxx.xxx). Requested URL: ‘/api/calendars/calendar.afval_kalender?start=2025-05-25T22:00:00.000Z&end=2025-06-01T22:00:00.000Z’. (Mozilla/5.0 (Linux; Android 11; 2201117SY Build/RP1A.200720.011; wv) AppleWebKit/537.36 (KHTML, like Gecko) Version/4.0 Chrome/136.0.7103.125 Mobile Safari/537.36)
I’m running it on a pi5. It was able to handle everything including voice easily although the voice thing out of the box is quite limited. I was able to do a lot with it through automations with custom commands though. It was pretty much able to replace my Google assistant system.
I’ve now upgraded it with ollama and Whisper on a network device (old laptop with i7 CPU and GTX1050 mobile GPU) and that works perfectly with llama3.2:3b. Pretty accurate and works well enough. Especially useful coupled with the local home assistant command processing since that by passes the LLM.
If you do you should use speech-to-phrase instead of whisper. I’ve found it to be more reliable if you’re using it with automation commands instead of an LLM. In your automations you can setup phrases as a trigger. It has support for aliases as well and I believe that it also supports templating for sentences.