Tracks and reports voice turn latency stages for diagnostics.
# Source: https://github.com/emqnuele/projectBEA | License: MIT | Copyright 2026 Emanuele Faraci # Provided via justclaw.it.com; keep this notice with the code. """Where the milliseconds of a voice turn go. One number decides whether any of the voice work was worth doing: how long between someone stopping talking and Bea's first sound. Everything else — a streaming TTS, a push channel, a reflex — is a bet on moving that number, and a bet nobody can read is a bet nobody can settle. So the turn carries a stopwatch from the moment audio lands to the moment the first byte leaves for the call, and the breakdown says which stage to blame. """ import time from dataclasses import dataclass, field from typing import Callable, Dict, List, Optional from src.core.events import EventCategory from s ... (truncated -- full source via MCP)
See the full source, get the GitHub permalink, and search for more like it.
Get a free API key