Inject a message that interrupts an in-progress turn, plus full LLM, TTS, and end-to-end latency breakdowns emitted automatically. Build barge-in voice UX and diagnose latency without extra instrumentation.
Stay on top of what shipped in AI. Weekly update.