[cpill]

I think the best example I've seen of the pipeline approach: ASR(STT) -> LLM -> TTS is still from about 10 months ago in HN https://github.com/KoljaB/RealtimeVoiceChat

Just a pity the code is undecipherable :(