effect-uai voice-loop

A voice assistant is three streams composed with Effect:

mic audio -> Transcriber -> final utterances -> LanguageModel -> text deltas -> SpeechSynthesizer -> audio

Reach for this when the user says any of:

• "Build a voice assistant / talk to my agent"
• "Pipe live STT into an LLM and speak the answer"
• "Support stop/cancel while the assistant is talking"
• "Queue follow-up questions spoken during an answer"

Pipeline shape

import { Stream } from "effect"