VaakSetu (వాక్సేతు) 🎙️

Automate outbound sales calls with AI. Built with Vapi, Python & Claude — handles greetings, objections, meeting booking & CRM logging out of the box.

---

What is VaakSetu?

VaakSetu (వాక్సేతు) means "Voice Bridge" in Telugu — a bridge between your business and every lead through intelligent, automated sales conversations.

VaakSetu is an AI-powered voice agent that autonomously handles outbound cold calls end-to-end — from the first greeting to booking a meeting or logging the outcome in your CRM. No human needed on your side.

---

✨ Features

• 🤖 AI Sales Agent — Powered by Claude (Anthropic) for natural, intelligent conversations
• 📞 Outbound Calling — Triggers real phone calls via Vapi + Twilio
• 🗣️ Natural Voice — ElevenLabs text-to-speech for human-like audio
• 🧠 Objection Handling — Trained to handle common sales objections
• 📅 Meeting Booking — Books demos and follow-ups automatically
• 📋 CRM Logging — Logs call outcomes, notes, and lead quality
• 🔁 Bulk Dialing — Parallel async dialing for high-volume campaigns
• 📊 Webhook Events — Real-time call event handling via FastAPI
• 📱 Voicemail Detection — Leaves custom voicemail messages automatically
• 🔒 Secure — All secrets managed via environment variables

---

🏗️ Architecture

Lead List (CSV/CRM)
      │
      ▼
  agent.py  ──►  Vapi API  ──►  Prospect's Phone
                    │
                    ▼
            Real-time Pipeline
         STT → Claude AI → TTS
                    │
                    ▼
              webhook.py
         (FastAPI webhook server)
                    │
            ┌───────┴────────┐
            ▼                ▼
        tools.py          CRM / Calendar
   (book_meeting,        (HubSpot, Calendly,
    log_outcome,          Google Calendar)
    transfer_call)

---

🗂️ Project Structure

vaaksetu/
├── agent.py          # Outbound call trigger + bulk dialer
├── assistant.py      # Claude system prompt + function tools definition
├── tools.py          # Tool handlers (booking, CRM logging, transfer)
├── webhook.py        # FastAPI server for Vapi event handling
├── test_vapi.py      # Connection test + call trigger debugger
├── .env.example      # Environment variable template
├── requirements.txt  # Python dependencies
└── README.md

---

⚙️ Prerequisites

• Python 3.10+
• Vapi account — $10 free credits on signup
• Anthropic API key
• Twilio account — for phone number
• ngrok — for local webhook tunnel (dev only)
• ElevenLabs account — for voice (optional, Vapi includes basic voices)

---

🚀 Quick Start

1. Clone the repository

git clone https://github.com/yourusername/vaaksetu.git
cd vaaksetu

2. Install dependencies

pip install -r requirements.txt

3. Set up environment variables

cp .env.example .env

Edit .env with your credentials:

VAPI_API_KEY=your_vapi_private_key
VAPI_PHONE_NUMBER_ID=your_vapi_phone_number_id
VAPI_ASSISTANT_ID=your_vapi_assistant_id
ANTHROPIC_API_KEY=your_anthropic_api_key
WEBHOOK_URL=https://your-ngrok-url.ngrok-free.app/webhook

4. Set up Vapi Assistant

1. Go to Vapi Dashboard
2. Create a new assistant → Blank Template
3. Set model to Anthropic → claude-sonnet-4-20250514
4. Paste the system prompt from assistant.py
5. Set voice to ElevenLabs → Adam
6. Add your webhook URL under Server URL
7. Add tools: book_meeting, log_outcome, transfer_call
8. Save and copy the Assistant ID

5. Start the webhook server

# Terminal 1 — Start ngrok
ngrok http 8000

# Terminal 2 — Start FastAPI webhook server
python -m uvicorn webhook:app --port 8000 --reload

6. Test the connection

python test_vapi.py

7. Make your first call

# Single call
python agent.py --phone +917013549646 --name "Test Lead"

# With prospect context
python agent.py --phone +917013549646 --name "Jane" --context "SaaS founder, 20 employees"

# Bulk dial all leads in SAMPLE_LEADS
python agent.py --bulk

---

📞 Call Flow

1. VaakSetu dials the prospect
2. Prospect picks up
3. Alex greets: "Hi, is this [Name]? I'm Alex from [Company]..."
4. Qualifying questions → Value pitch → Objection handling
5. Prospect agrees → book_meeting tool fires → Meeting booked ✅
6. Call ends → log_outcome tool fires → CRM updated ✅

---

🛠️ Customisation

Change the sales script

Edit SYSTEM_PROMPT in assistant.py — replace [Company] with your company name and tailor the pitch.

Connect your CRM

Edit tools.py → handle_log_outcome() — add HubSpot, Salesforce, or Airtable API calls.

Connect your calendar

Edit tools.py → handle_book_meeting() — add Calendly or Google Calendar integration.

Add more leads

Edit SAMPLE_LEADS in agent.py or load from a CSV:

import csv

with open("leads.csv") as f:
    leads = list(csv.DictReader(f))  # columns: phone, name, context

asyncio.run(dial_leads_parallel(leads, concurrency=10))

---

📦 Requirements

vapi-python
anthropic
fastapi
uvicorn
httpx
requests
python-dotenv

Install all:

pip install vapi-python anthropic fastapi uvicorn httpx requests python-dotenv

---

🔒 Security

• Never commit your .env file — it's in .gitignore
• Use Private Key from Vapi (not Public Key) for server-side calls
• Rotate API keys immediately if accidentally exposed
• Use environment variables — never hardcode secrets in code

---

🗺️ Roadmap

• CSV lead import
• Calendly integration
• HubSpot CRM integration
• Real-time dashboard for call monitoring
• A/B testing for opening messages
• Slack alerts on meeting bookings
• Retry logic for unanswered calls
• Call recording transcripts

---

🤝 Contributing

Pull requests are welcome! For major changes, please open an issue first to discuss what you would like to change.

---

📄 License

MIT License — feel free to use, modify, and distribute.

---

🙏 Built With

• Vapi — Voice AI infrastructure
• Anthropic Claude — AI brain
• ElevenLabs — Natural voice synthesis
• FastAPI — Webhook server
• Twilio — Telephony

---

Built with ❤️ from Hyderabad, India 🇮🇳
వాక్సేతు — Voice Bridge