Simple project leveraging Whisper(Voice Audio -> Text), ChatGPT(Text -> commands) and my local running instance of Home-Assitant A.I to create a sort of voice interface home assistant.
Interesting thing here is that it can convert commands from many different languages (English, Spanish, Japanese tested so far) into commands.
- Have the home-assistant.io server up and running
- Make sure to set the following env-vars in an .env file
TOKEN=<HOME_ASSISTANT_TOKEN>
OPENAI_API_KEY=<OPENAI_API_KEY>
OPENAI_API_ORG=<OPEN_API_ORG>
- Create a virtual environment:
python3 -m virtualenv .venva) activate itsource .venv/bin/activate pip install -r requirements a) You may have some issues installing pyaudio, to fix that on debian install the following packagesportaudio19-dev python3-all-dev` with your favorite package managerpython jarvis.py