[import_awesome]

I have half a MUD that uses TTS and speech-to-text as the interface with LLMs generating the content and a database keeping the context. It works surprisingly well.