in the past few days I've build a voice based interaction with OpenAI's DaVinci model while keeping conversation context.

Give it a try and let me know your thoughts: https://talkto.chat