trent-local-voice-server-prod

name	trent-local-voice-server-prod
description	Deploy the local voice development server with ngrok for outbound calling. Use when the user asks to test voice locally, start the voice server, or debug voice calls.

Local Voice Server Deployment (Production)

Deploy the local voice development server with ngrok tunnel for outbound calling. This runs:

1. ngrok (to expose voice server to internet)
2. Voice server (port 8080, with ngrok URL)
3. Backend server (port 8000)
4. Frontend (port 3000)

Manual Steps (if running without tmux automation)

1. Start ngrok: ngrok http 8080
2. Get URL: curl -s http://localhost:4040/api/tunnels | jq -r '.tunnels[0].public_url'
3. Voice server: DECAGON_ENV=prod VOICE_OUTBOUND_BASE_URL=<ngrok-url> uvicorn backend.server_cs.voice_server:app --reload --port 8080
4. Backend: DECAGON_ENV=prod uvicorn backend.server_cs.server:app --reload --port 8000
5. Frontend: cd frontend && DECAGON_ENV=prod yarn run dev --webpack

Full Script (Automated 2x2 Layout)

Creates 4 NEW panes (leaves Claude Code pane untouched):

[Claude] | [ngrok]   | [backend]
         | [voice]   | [frontend]

# Create 4 NEW panes - don't touch the original Claude Code pane
# 1. Split right for ngrok
NGROK_PANE=$(tmux split-window -h -P -F '#{pane_id}')

# 2. Split ngrok down for voice
VOICE_PANE=$(tmux split-window -v -t "$NGROK_PANE" -P -F '#{pane_id}')

# 3. Split ngrok right for backend
BACKEND_PANE=$(tmux split-window -h -t "$NGROK_PANE" -P -F '#{pane_id}')

# 4. Split voice right for frontend
FRONTEND_PANE=$(tmux split-window -h -t "$VOICE_PANE" -P -F '#{pane_id}')

# Layout is now:
# [Claude] | [ngrok]   | [backend]
#          | [voice]   | [frontend]

# 5. Start ngrok
tmux send-keys -t "$NGROK_PANE" 'ngrok http 8080' Enter

# 6. Wait for ngrok to initialize
sleep 4

# 7. Get ngrok URL (retry if needed)
for i in 1 2 3 4 5; do
  NGROK_URL=$(curl -s http://localhost:4040/api/tunnels 2>/dev/null | jq -r '.tunnels[0].public_url' 2>/dev/null)
  if [ -n "$NGROK_URL" ] && [ "$NGROK_URL" != "null" ]; then
    break
  fi
  sleep 2
done

if [ -z "$NGROK_URL" ] || [ "$NGROK_URL" = "null" ]; then
  echo "ERROR: Could not get ngrok URL. Check if ngrok is running."
  exit 1
fi

echo "ngrok URL: $NGROK_URL"

# 8. Start voice server (with ngrok URL) - use 'env' for fish shell compatibility
tmux send-keys -t "$VOICE_PANE" "env DECAGON_ENV=prod VOICE_OUTBOUND_BASE_URL=$NGROK_URL uvicorn backend.server_cs.voice_server:app --reload --port 8080" Enter

# 9. Start backend - use 'env' for fish shell compatibility
tmux send-keys -t "$BACKEND_PANE" 'env DECAGON_ENV=prod uvicorn backend.server_cs.server:app --reload --port 8000' Enter

# 10. Start frontend - use 'env' for fish shell compatibility
tmux send-keys -t "$FRONTEND_PANE" 'cd frontend && env DECAGON_ENV=prod yarn run dev --webpack' Enter

echo "Voice development environment started!"
echo "ngrok URL: $NGROK_URL"
echo ""
echo "Layout:"
echo "  [Claude] | [ngrok]   | [backend]"
echo "           | [voice]   | [frontend]"
echo ""
echo "To test: Go to http://localhost:3000 -> Preview -> Voice -> Enter phone # and select agent"

Prerequisites

• ngrok installed and authenticated (ngrok config add-authtoken <token>)
• tmux running
• jq installed (for parsing ngrok API response)

Expected Result

• ngrok tunnel running, exposing port 8080
• Voice server at http://localhost:8080 (with VOICE_OUTBOUND_BASE_URL set)
• Backend at http://localhost:8000
• Frontend at http://localhost:3000
• Test via: Preview -> Voice -> Enter phone # & select voice agent

Notes

• This is for outbound calling without Twilio webhook setup
• For inbound calling, you need to update the webhook URL in Twilio
• The ngrok URL changes each time unless you have a paid ngrok plan with reserved domains