🍸 Outback Bar

AI-Powered Bartender Experience

---

Welcome to Outback Bar! Powered by SignalWire's AI Agent framework, Max, your virtual bartender at Outback Bar, takes drink orders, manages tabs, and provides personalized service through natural voice conversations.

🌟 Features

Core Functionality

• Natural Voice Interaction: Talk to Max like a real bartender
• Full Drink Menu: Cocktails, beer, wine, and non-alcoholic options
• Tab Management: Real-time tracking of orders and running total
• Happy Hour: Automatic 20% discount on cocktails (4-7 PM)
• Drink Customization: Order doubles, specify ice preferences, etc.
• Responsible Service: Built-in monitoring and service limits
• Real-time Updates: Live UI updates as orders are placed

Menu Categories

• Cocktails ($10-13): Margarita, Old Fashioned, Mojito, Martini, and more
• Beer ($6-8): IPA, Lager, Stout, Wheat Beer, Pale Ale
• Wine ($9-11): House Red/White, Prosecco, Pinot Noir, Sauvignon Blanc
• Non-Alcoholic ($0-6): Mocktails, sodas, juices, water

🚀 Quick Start

Prerequisites

• Python 3.8+
• SignalWire account with AI Agent access
• Node.js (for frontend development)

Installation

1. Clone the repository:

cd bartender

2. Install Python dependencies:

pip install -r requirements.txt

3. Configure environment:

cp .env.example .env
# Edit .env with your SignalWire credentials

4. Update the SignalWire token: Edit web/app.js and replace the token:

const STATIC_TOKEN = 'your_signalwire_token_here';

Running the Agent

python bartender_agent.py

The server will start on port 3030 by default.

Access the Web UI

Open your browser to: http://localhost:3030

🔧 Configuration

Environment Variables (.env)

# SignalWire Configuration
SIGNALWIRE_PROJECT_ID=your_project_id
SIGNALWIRE_TOKEN=your_token
SIGNALWIRE_SPACE=your_space.signalwire.com

# Server Configuration
HOST=0.0.0.0
PORT=3030

# Bar Settings
HAPPY_HOUR_START=16  # 4 PM
HAPPY_HOUR_END=19    # 7 PM
HAPPY_HOUR_DISCOUNT=0.20  # 20% off

# Service Limits
MAX_DRINKS_PER_HOUR=2
MAX_TOTAL_DRINKS=5

💬 Usage Examples

Customer Interactions

Ordering Drinks:

• "I'll have a margarita"
• "Can I get a double old fashioned?"
• "What beers do you have on tap?"
• "I'd like something fruity and refreshing"

Tab Management:

• "What's my tab?"
• "Can you review my order?"
• "I'm ready to close out"
• "Add a 20% tip"

Happy Hour:

• "Is it happy hour?"
• "What's on special?"

🏗️ Architecture

Backend (bartender_agent.py)

• Built on SignalWire's AgentBase class
• Implements SWAIG functions for drink operations
• Uses TF-IDF for fuzzy drink matching (optional)
• Manages conversation flow through state contexts
• Sends real-time events to frontend

Frontend (web/)

• Single-page application with real-time updates
• WebRTC connection for voice communication
• Responsive design for mobile and desktop
• Visual feedback for all operations
• Bar-themed UI with elegant styling

Key Components

SWAIG Functions

• add_drink: Add drinks to customer's tab
• remove_drink: Remove items from tab
• review_tab: Display current tab status
• check_happy_hour: Check happy hour status
• close_tab: Process payment and close tab

Conversation States

1. greeting: Welcome and initial order
2. taking_order: Main ordering phase
3. closing_tab: Payment processing
4. tab_closed: Post-payment state

User Events

Events sent from backend to frontend for UI updates:

• drink_added: New drink added to tab
• drink_removed: Item removed from tab
• tab_review: Full tab sync
• tab_closed: Payment completed
• happy_hour_status: Happy hour status update

🎨 Customization

Adding New Drinks

Edit the DRINKS dictionary in bartender_agent.py:

"C011": {
    "name": "Espresso Martini",
    "price": 14.00,
    "description": "Vodka, coffee liqueur, espresso",
    "abv": 20,
    "category": "cocktail"
}

Modifying Happy Hour

Adjust times and discount in .env:

HAPPY_HOUR_START=15  # 3 PM
HAPPY_HOUR_END=20    # 8 PM
HAPPY_HOUR_DISCOUNT=0.25  # 25% off

Changing the Bartender's Personality

Edit the personality section in bartender_agent.py:

self.prompt_add_section(
    "Personality",
    "Your custom personality description here..."
)

🔊 Audio Feedback

The agent uses audio files for ambient feedback:

• /pouring.mp3 - When adding drinks
• /clearing.mp3 - When removing items
• /calculating.mp3 - When reviewing tab
• /payment.mp3 - When processing payment

Place these files in your SignalWire assets directory.

📊 Responsible Service Features

Built-in safety features include:

• Maximum 5 drinks per customer session
• Maximum 2 drinks per hour monitoring
• Automatic water suggestions after 3 alcoholic drinks
• Service refusal when limits are reached

🐛 Troubleshooting

Connection Issues

• Verify SignalWire credentials in .env
• Check token in web/app.js is valid
• Ensure port 3030 is not in use
• Check firewall settings

No Audio

• Allow microphone permissions in browser
• Check SignalWire project has voice enabled
• Verify WebRTC connection established

Events Not Updating UI

• Check browser console for errors
• Verify user events are being sent (check server logs)
• Ensure WebSocket connection is active

📚 API Endpoints

• GET / - Main web UI
• GET /api/menu - Returns drink menu JSON
• GET /api/happy-hour - Current happy hour status
• GET /api/info - System information
• POST /swml - SignalWire webhook endpoint
• POST /swml/swaig - SWAIG function endpoint
• GET /health - Health check endpoint

🤝 Contributing

Contributions are welcome! Areas for improvement:

• Additional drink recipes
• Payment integration (Stripe/Square)
• Loyalty program features
• Drink recommendation engine
• Multi-language support

📝 License

This project is based on the Holy Guacamole architecture and is provided for educational and demonstration purposes.

🙏 Acknowledgments

• Built with SignalWire AI Agents
• Inspired by the Holy Guacamole implementation
• Uses SignalWire's WebRTC for real-time communication

---

Need Help? Check the SignalWire Documentation or open an issue.