Implement OpenAI completion streaming to post-process AI responses faster. This includes function calling aka command execution and TTS output.