Quickstart¶
Get voice-agent running in 5 minutes.
Prerequisites¶
- A Telegram bot token (from @BotFather)
- A running whisper-server
- Claude CLI installed and authenticated (
claude login)
Step 1: Clone and Configure¶
git clone https://github.com/paolino/voice-agent
cd voice-agent
# Create .env file
cat > .env << EOF
TELEGRAM_BOT_TOKEN=your-bot-token
WHISPER_URL=http://localhost:8080/transcribe
ALLOWED_CHAT_IDS=your-chat-id
EOF
Step 2: Run the Bot¶
# Using Nix
nix run
# Or using Python directly
python -m voice_agent
Step 3: Test It¶
- Open Telegram and find your bot
- Send
/startto see the welcome message - Send a voice message saying "list files"
- The bot will transcribe your message and show Claude's response
Basic Voice Commands¶
| Say this | What happens |
|---|---|
| "list files" | Lists files in current directory |
| "status" | Shows session status |
| "new session" | Starts a fresh session |
| "yes" / "approve" | Approves a pending permission |
| "no" / "reject" | Rejects a pending permission |
Next Steps¶
- Configure projects for easy directory switching
- Learn all voice commands
- Set up deployment for production use