feat: add macOS say as alternative TTS engine

[hydrohelix]

Summary

• Adds macOS built-in say command as an alternative TTS engine alongside ElevenLabs
• Configurable via settings.json (voice.tts_engine: "macos-say") or TTS_ENGINE env var
• Auto-falls back to macOS say when no ElevenLabs API key is present
• Adds voice CLI script for easy engine switching and server management
• install.sh symlinks the CLI to /usr/local/bin/voice for universal access

Changes

• VoiceServer/server.ts — TTS engine abstraction: speakWithSay() function, engine branching in sendNotification(), updated /health endpoint and startup logs
• VoiceServer/voice — New CLI: voice say [name] [rate], voice 11labs, voice status, voice test, voice restart, voice log, voice voices
• VoiceServer/install.sh — Symlinks voice CLI to /usr/local/bin with permission fallback
• VoiceServer/uninstall.sh — Removes the symlink on uninstall
• .gitignore — Adds .local-backup/ for user customizations

Configuration

{
  "voice": {
    "tts_engine": "macos-say",
    "macos_say": {
      "voice": "Samantha",
      "rate": 200
    }
  }
}

Or via env var: TTS_ENGINE=macos-say

Test plan

• voice say switches engine and restarts server
• voice 11labs switches back to ElevenLabs
• voice status shows correct engine and server health
• voice test produces audible speech with both engines
• Server auto-falls back to macOS say when ELEVENLABS_API_KEY is missing
• curl localhost:8888/health reflects active engine in response
• Pronunciation rules apply to both engines
• Volume control works with macOS say (renders AIFF, plays via afplay)
• install.sh creates /usr/local/bin/voice symlink
• uninstall.sh removes the symlink

🤖 Generated with Claude Code