I think this idea would be really cool, but if it isn't inline with the direction of this project, then I can close this and work on a personal fork.
Idea: The initial pipeline could be the Assist pipeline, then if the response is Sorry, I couldn't understand that then you could call a second OpenAI pipeline that could handle the input. Calling the Home assistant Assist pipeline first would allow actions to be executed, but still also have the ability to converse in natural language with ChatGPT. One way to enhance the performance is to have the first pipeline only be STT and Assist and the second pipeline only ChatGPT and TTS.
I think this idea would be really cool, but if it isn't inline with the direction of this project, then I can close this and work on a personal fork.
Idea: The initial pipeline could be the Assist pipeline, then if the response is
Sorry, I couldn't understand that
then you could call a second OpenAI pipeline that could handle the input. Calling the Home assistant Assist pipeline first would allow actions to be executed, but still also have the ability to converse in natural language with ChatGPT. One way to enhance the performance is to have the first pipeline only be STT and Assist and the second pipeline only ChatGPT and TTS.