AI Voice Receptionist & Autonomous Appointment Booking Agent

An enterprise AI-powered phone receptionist that answers calls 24/7, checks real-time availability, books appointments on Google Calendar, and sends instant SMS confirmations via Twilio — powered by Django, Claude AI (Anthropic), Vapi, and PostgreSQL.

📑 Table of Contents

• What It Does
• Architecture & Call Flow
• Tech Stack
• Quick Start Guide
  • 1. Clone and Configure
  • 2. Generate Secret Key
  • 3. Configure Environment Variables
  • 4. Google Calendar Setup
  • 5. Run with Docker Compose
  • 6. Expose Local Webhook (ngrok)
  • 7. Connect Vapi Voice Pipeline
• Project Layout & Endpoints
• Django Admin & Telemetry
• Production Deployment
• Author & Community
• License

🌟 What It Does

• 📞 24/7 Inbound Voice Calls: Answers phone calls with ultra-low latency, human-like voice synthesis.
• 📅 Real-Time Calendar Sync: Dynamically checks Google Calendar for conflict-free availability.
• ⚡ Autonomous Booking: Locks in appointments and auto-generates Google Calendar events.
• 📱 Instant SMS Verification: Dispatches booking confirmations and reminders via Twilio SMS.
• 📊 Audit & Call Logs: Persists full call transcripts, duration, summaries, and patient data into PostgreSQL.

🏛️ Architecture

Vapi is leveraged as the voice transport layer (Speech-to-Text and Text-to-Speech), while the Django backend acts as the autonomous decision brain driving Claude AI function calling and tool execution.

sequenceDiagram
    autonumber
    actor Caller as 👤 Caller
    participant Vapi as 🎙️ Vapi (STT/TTS)
    participant Django as 🐍 Django Backend
    participant Claude as 🤖 Claude AI (Anthropic)
    participant GCal as 📅 Google Calendar
    participant Twilio as 💬 Twilio SMS
    participant DB as 🗄️ PostgreSQL

    Caller->>Vapi: Speaks ("I want to book an appointment")
    Vapi->>Django: POST /api/vapi/chat/completions/ (Transcript)
    Django->>Claude: Prompt + Tool Definitions + History
    Claude-->>Django: Tool Call: check_calendar_availability
    Django->>GCal: Query busy/free slots
    GCal-->>Django: Available slots returned
    Django->>Claude: Tool Response (Slots)
    Claude-->>Django: Natural Voice Response
    Django->>DB: Log conversation turns
    Django-->>Vapi: Synthesized Text Reply
    Vapi-->>Caller: Speaks response back to caller
    Note over Caller,DB: Once confirmed: GCal Event Created & Twilio SMS Sent!

🛠️ Tech Stack

🚀 Quick Start

1. Clone and configure

git clone https://github.com/codewithmuh/ai-voice-agent.git
cd ai-voice-agent
cp .env.example .env

2. Generate a Django secret key

python3 -c "import secrets; print(secrets.token_urlsafe(50))"

Paste the output into .env as the SECRET_KEY value.

3. Add your API keys to .env

SECRET_KEY=your-generated-secret-key
DEBUG=True
ALLOWED_HOSTS=localhost,127.0.0.1,0.0.0.0,*

DB_NAME=ai_receptionist
DB_USER=postgres
DB_PASSWORD=postgres
DB_HOST=db
DB_PORT=5432

ANTHROPIC_API_KEY=sk-ant-...
VAPI_API_KEY=...
GOOGLE_CALENDAR_CREDENTIALS=google-calendar-credentials.json
[email protected]
TWILIO_ACCOUNT_SID=AC...
TWILIO_AUTH_TOKEN=...
TWILIO_PHONE_NUMBER=+1...

Where to get keys:

• Anthropic: console.anthropic.com
• Vapi: dashboard.vapi.ai
• Twilio: twilio.com/console
• Google Calendar: Follow setup below

4. Google Calendar setup

Step 1: Create a Google Cloud project

1. Go to Google Cloud Console.
2. Click Select a Project → New Project → Name it → Create.

Step 2: Enable the Calendar API

1. Go to APIs & Services → Library.
2. Search for "Google Calendar API" → Click Enable.

Step 3: Create Service Account Credentials

1. Go to APIs & Services → Credentials.
2. Click + Create Credentials → Service Account.
3. Name it (e.g. calendar-bot) → Create and Continue → Done.
4. Click the newly created service account → Keys tab → Add Key → Create New Key → JSON.
5. Save the file in the project root as google-calendar-credentials.json.

Step 4: Share Calendar with Service Account

1. Copy the client_email from google-calendar-credentials.json (e.g. [email protected]).
2. In Google Calendar, click your calendar settings → Share with specific people or groups.
3. Add the service account email and give permission: Make changes to events.

5. Start with Docker Compose

docker compose up --build

Django will start at http://localhost:8000 with PostgreSQL connected.

6. Create Superuser for Admin Dashboard

docker compose exec web python manage.py createsuperuser

7. Expose with ngrok

ngrok http 8000

Copy the forwarding HTTPS URL (https://xxxx.ngrok-free.app).

8. Connect Vapi

1. Go to Vapi Dashboard.
2. Create an Assistant → Set Model to Custom LLM:

   https://xxxx.ngrok-free.app/api/vapi/chat/completions/
   

3. Set Server URL (for post-call summary webhooks):

   https://xxxx.ngrok-free.app/api/vapi/webhook/
   

4. Assign a phone number and dial to test!

📂 Project Structure

ai-voice-agent/
├── receptionist_project/
│   ├── settings.py              # Django config
│   └── urls.py                  # Root URL routing
├── calls/
│   ├── models.py                # CallLog & Appointment schema
│   ├── views.py                 # Custom LLM & Vapi webhook endpoints
│   ├── urls.py                  # API route definitions
│   ├── admin.py                 # Django admin registration
│   ├── agent/
│   │   ├── receptionist.py      # Claude agentic loop orchestrator
│   │   ├── tools.py             # Tool schemas (Calendar, SMS, DB)
│   │   └── prompts.py           # Domain system prompt
│   └── services/
│       ├── calendar.py          # Google Calendar API integration
│       ├── sms.py               # Twilio SMS client
│       └── database.py          # Database operations
├── docker-compose.yml
├── Dockerfile
├── .env.example
└── requirements.txt

Key API Endpoints

🖥️ Django Admin Dashboard

Access the admin dashboard at http://localhost:8000/admin/:

• Call Records: Full transcripts, duration, caller phone numbers, and LLM sentiment.
• Appointments: View, modify, or cancel bookings synced with Google Calendar.
• Filters: Quickly filter calls and appointments by date, status, and patient name.

🚢 Deploy to Production

Deploy easily on platforms like Railway, AWS EC2, or Render:

# Example with Railway CLI
railway init
railway add --plugin postgresql
railway up

Set environment variables in your cloud provider's dashboard and update the Vapi webhook URLs accordingly.

👨‍💻 Author

Muhammad Rashid — AI & DevOps Engineer, AWS Community Builder.

• 🎥 YouTube: @codewithmuh
• 💼 LinkedIn: Muhammad Rashid
• 🌐 Website: codewithmuh.com · devopsbymuh.com

⭐ If you find this AI Voice Agent helpful, please give it a star!

📄 License

This project is licensed under the MIT License.