reyna-cli/backup/macmini-automation-baseline-2026-08-03/macmini-mcp/WatchHermes

Hermes Watch PTT Foundation

These files are intended to be added to a standalone watchOS SwiftUI app target:

• AudioCaptureManager.swift: captures microphone audio with AVAudioEngine, converts it to 16 kHz, 16-bit, mono PCM, and emits raw Data chunks.
• AudioStreamManager.swift: opens a URLSessionWebSocketTask, sends binary PCM chunks, receives binary response audio, and plays either complete WAV blobs or raw 16 kHz PCM chunks.
• ContentView.swift: minimal push-to-talk UI using DragGesture(minimumDistance: 0) for press/release.
• InfoPlistAdditions.xml: exact plist entries for microphone, LAN access, and background audio.

The referenced ESP32 demo posts a complete WAV file to http://192.168.68.126:8642/api/esp32/voice with 16 kHz, 16-bit, mono PCM and receives WAV audio back. The watch implementation here follows your requested WebSocket model: it streams raw PCM chunks while the button is held, sends simple {"type":"start"} and {"type":"stop"} text control messages, and accepts either WAV or raw PCM binary responses.

Update the endpoint in ContentView.init():

AudioStreamManager(endpoint: URL(string: "ws://YOUR_LOCAL_IP:PORT/stream")!)

If your Hermes WebSocket endpoint expects authentication headers, create a URLRequest, set the headers, and pass that request into session.webSocketTask(with:) in AudioStreamManager.connect().