I am trying to figure out a way to have extended context conversation for things like brainstorming. Cognitively I retain and learn information best by speaking, and itās better if someone is there to react or respond to what I am saying. It could be my Voice PEs instead of any adult that shows up at my house.
1 Like
both of those are piper fwiw thx for fixing links!
Speech-to-Phrase is awesome! It makes VPE actually usable locally. Great job!
2 Likes
Iām very happy with the speech-to-phrase, finally reliable local speech-to-text. This is HUGE!
Iāve a request. Is it possible to:
- enable the wake word ping
- disable the talk back when the action was succesful performed
- enable the talk back when the action was unsuccesful (e.g. ācould not find a device named blablablaā)
Reason: iād like to reduce unneeded audio; my home is not a funfair.
3 Likes
Thank you for all the great hard work!! 
My biggest two issues right now are:
-
The order that the assistant answers requests, currently it will first do the action and than reply, when this is happening if I ask the assistant to play music (LLM) it will start the music and than reply which cause the music to stop and wont continue (when the device speaker is used for both replying and hearing music) which is very frustrating, if it was the opposite than it will reply and only than start the music.
-
At least with the Atom echo, there are tons of false positives to the wake words (no matter which one is chosen which makes me think it isnt even an actual the wake word itself)
Thank you
This is all amazing!
If I could suggest another area of potential focus that would really round out the audio pipeline for me: bring snapcast support to the voice PE/esphome. Then responses could be sent back to cast groups and the aux port on the voice PE would be much more useful.
It would also be wonderful to be able to customize the behavior of the dial on the voice PE!
Awesome work on all of this, what an impressive community we have here 
2 Likes
Is STP is meant to replace local STT in the future (whisper, ā¦) ? I guess not, so as others wrote, it would be great to be able to define custom workflows. Iām thinking such as :
- STP
- No intent recognized => local STT + local agent
- No intent recognized => cloud STT (Nabu, Google etc.) + local agent
- No intent recognized => LLM agent (openai, ollama)
Is someone is working on a āFabricā AI tooling / AI agent integration for Home Assistant?
fabric (from Daniel Miessler) is an open-source framework for augmenting humans using AI. It provides a modular framework for solving specific problems using a crowdsourced set of AI prompts that can be used anywhere.
Would be awesome to have integration(s) for Fabricās āPatternsā (currated crownsoured promts) as AI agents / tooling that can use Fabric pattern promts as tools can use from Home Assistant.
fabric intro video:
(Keep in mind that many of these were recorded when Fabric was Python-based, so remember to use the current install instructions)
Also check out this fabric origin story:
Although STP is probably awesome at recognizing those straightforward home control commands like āsome light on/offā, āclose curtainsā etc. it is to restricted to serve all voice assistance needs as Rhasspy also was. Using Rhasspy soon turned frustrating as the limits kept chaving.
Having the possibility for āfall throughā to a whisper (or LLM) powered speech-to-text service would be perfect and (for me at least) essential to want to try STP.
2 Likes
I just ordered one of theseā¦ want to try it outā¦ I have one based on anotherās hardware but it doesnāt always do what I want so hopefully this one will be more integrated with HA
Is anything required to setup to get text streaming to start working? I have openAI configured and running but when typing in a question I get the response all at once.
1 Like
A post was split to a new topic: Voice assistant help