HYCHAT APEX / VOICE + LIVE DIAGNOSTICS

Audits the active Chat V16 pipeline: realtime transcription, the existing personalized model/tool pipeline, incremental Hypatia PCM, and first-audio playback timing.
User-Agent:
Secure Context:
Protocol:
↗ Open Production Chat
Real-provider text-to-audio test only; microphone transcription, real barge-in and full conversation are NOT checked by this button.
🌐EnvironmentPENDING
🔐Auth (whoami)PENDING
⚡EL Agent PrewarmPENDING
🎤Microphone AccessPENDING
🔈Fallback TTS OutputPENDING
📡EL WebSocket TransportPENDING
⏱️Startup Deadline & RecoveryPENDING
🧰EL Tools ContractPENDING
🔐Signed Production SessionPENDING
⚡Voice Engine Pre-warmPENDING
🪪Voice Sheet / Dock UIPENDING
🎮GPU / Canvas HealthPENDING
🔇Echo Gate (AEC)PENDING
🍎Safari Mic / Audio SuspensionPENDING
♻️iOS Lifecycle RecoveryPENDING
⬜Safari White-Screen GuardPENDING
🔆Wake Lock (Screen-on)PENDING
🧠Personality InjectionPENDING
📦Context Items (profile + memory + modules)PENDING
📡 Live Client Telemetry — real sessions (last 80)
Events the chat voice engine persisted to state/voice/client-events.jsonl (hv_session / hv_openai / hv_tts / hv_lipsync / hv_error). This is what to copy when reporting a voice problem.
🎙️ Production Live-Avatar Controller — Direct Probe
This uses the same HypatiaLiveAvatar controller as production chat.html: signed ElevenLabs WebSocket session, one microphone owner, still avatar by default, optional renderer. Use it for a real browser/microphone transport check after readiness passes.
Not running
Transportidle
Micidle
Rendererstill
📟 Event Log (this test run)