Customization

Before executing, check for user customizations at: ~/.claude/skills/PAI/USER/SKILLCUSTOMIZATIONS/VoiceServer/

If this directory exists, load and apply any PREFERENCES.md, configurations, or resources found there. These override default behavior. If the directory does not exist, proceed with skill defaults.

VoiceServer Skill

Domain: Voice notification system using ElevenLabs TTS.

Algorithm: ~/.claude/skills/PAI/SKILL.md

Voice Architecture (ElevenLabs)

The voice server uses ElevenLabs API for high-quality TTS. Voice characteristics are defined by:

1. Voice ID: ElevenLabs voice identifier (from settings.json → daidentity.voiceId)
2. Prosody Settings: stability, similarity_boost, style, speed, use_speaker_boost

Prosody Configuration (from settings.json → daidentity.voice):

{
  "stability": 0.35,
  "similarity_boost": 0.80,
  "style": 0.90,
  "speed": 1.1,
  "use_speaker_boost": true
}

API Endpoints

Primary: Voice Notification

curl -X POST http://localhost:8888/notify \
  -H "Content-Type: application/json" \
  -d '{
    "message": "Your message here",
    "title": "Kai says",
    "voice_enabled": true,
    "voice_id": "s3TPKV1kjDlVtZbl4Ksh",
    "voice_settings": {
      "stability": 0.35,
      "similarity_boost": 0.80,
      "style": 0.90,
      "speed": 1.1,
      "use_speaker_boost": true
    }
  }'

Health Check

curl http://localhost:8888/health

Management

Scripts: ~/.claude/VoiceServerV1/{start,stop,status,restart}.sh

Check Status:

~/.claude/VoiceServerV1/status.sh

Start Server:

cd ~/.claude/VoiceServerV1 && bun run server.ts

Voice Output Format

All algorithm completions include a voice line that gets spoken:

🗣️ Kai: [16 words max - THIS IS SPOKEN ALOUD]

This line is extracted by hooks and sent to the voice server automatically.

Infrastructure

• Engine: ElevenLabs API (eleven_turbo_v2_5)
• Port: 8888
• Voice Config: ~/.claude/settings.json → daidentity.voiceId and daidentity.voice
• Server Code: ~/.claude/VoiceServerV1/server.ts