feat(voice-server + installer): Google Cloud TTS + cross-platform audio

[fayerman-source]

Summary

• Google Cloud TTS as alternative provider alongside ElevenLabs — configurable via settings.json → daidentity.ttsProvider
• Cross-platform audio playback — detects afplay (macOS), mpv/ffplay/aplay (Linux/WSL2) at startup
• Cross-platform desktop notifications — osascript (macOS), notify-send (Linux) with silent fallback
• Backwards compatible — defaults to ElevenLabs when ttsProvider is not set

Why

• ElevenLabs free tier is 10K chars/month; Google Cloud free tier is 4M chars/month (Standard) or 1M (WaveNet/Neural2)
• Voice server used macOS-only afplay and osascript, making it non-functional on Linux/WSL2
• No new dependencies — uses Google's REST API directly via fetch

Configuration

Add to ~/.env:

GOOGLE_CLOUD_API_KEY=your_key_here
# or reuse existing: GOOGLE_API_KEY=your_key_here (both accepted)

Add to ~/.claude/settings.json:

{
  "daidentity": {
    "ttsProvider": "google-cloud",
    "googleCloudVoice": {
      "languageCode": "en-US",
      "voiceName": "en-US-Neural2-D",
      "voiceType": "NEURAL2",
      "speakingRate": 1.0,
      "pitch": 0.0
    }
  }
}

Or keep using ElevenLabs by not setting ttsProvider (or setting it to "elevenlabs").

Test plan

• Voice server starts with ttsProvider: "google-cloud" and logs correct provider
• Health endpoint shows voice_system: "google-cloud", google_cloud_configured: true, audio_player
• TTS generates and plays audio end-to-end on Linux/WSL2 via mpv
• Backwards compatible — omitting ttsProvider defaults to ElevenLabs
• Verify on macOS with afplay

Context

Re-implementation of #687 (closed during v4.0 restructuring) with additional Linux/WSL2 cross-platform support. Tested live on WSL2 with Google Cloud Neural2-D voice.

🤖 Generated with Claude Code