qwwiwi Groq Voice Transcription
When you receive a message containing <media:audio> tag, the user sent a voice message.
MANDATORY: Transcribe BEFORE responding
When you see <media:audio>:
- Extract the .ogg file path from the
[media attached: /path/to/file.ogg]line - Run the transcription script:
bash $CLAUDE_SKILL_DIR/scripts/transcribe.sh "/path/to/file.ogg"
- Use the transcript text to understand what the user said
- Respond to the user's spoken message naturally
- Do NOT mention the transcription process unless asked
Important
- Always transcribe first, then respond to the content
- The script uses Groq Whisper API (fast, accurate, supports Russian and 50+ languages)
- If transcription fails, tell the user you couldn't process their voice message
- Do NOT say "I can't listen to audio" -- you CAN via this script
Setup
- Get a free API key at https://console.groq.com
- Save it:
echo 'your-key' > ~/.claude-lab/shared/secrets/groq-api-key - The script reads the key from that file automatically