AI Sales Agent

An AI-powered voice sales agent that makes outbound calls, engages users in natural conversation, and books follow-up demo calls. Built with Twilio for telephony, Deepgram for AI voice interaction (using OpenAI GPT-4o), and a FastAPI/React stack.

Features

• Outbound Calls: Users can request calls via a React frontend; Twilio initiates the call.
• Real-Time Voice Interaction: The AI agent "Emma" uses Deepgram to listen, think (via GPT-4o), and speak naturally.
• Admin Dashboard: Configure the company name and knowledge summary for the AI agent.
• Audio Streaming: Bidirectional WebSocket streaming between Twilio and Deepgram with audio resampling.

Tech Stack

• Backend: FastAPI (Python), WebSockets, numpy, soundfile, aiohttp
• Frontend: React, React Router
• Telephony: Twilio
• AI Voice: Deepgram Agent (OpenAI GPT-4o, Aura-Asteria TTS)
• Dependencies: See backend/requirements.txt and frontend/package.json

Setup Instructions

Prerequisites

• Python 3.9+
• Node.js 18+
• Twilio account (API credentials)
• Deepgram account (API key)
• Git installed

Backend Setup

1. Navigate to backend/:

   cd backend

2. Create a virtual environment and activate it:

   python -m venv venv
   venv\Scripts\activate

3. Install dependencies:

   pip install -r requirements.txt

4. Set environment variables in a .env file:

   TWILIO_ACCOUNT_SID=your_sid
   TWILIO_AUTH_TOKEN=your_token
   TWILIO_PHONE_NUMBER=your_twilio_number
   PUBLIC_BASE_URL=https://your-public-url
   DEEPGRAM_API_KEY=your_deepgram_key
   

5. Run the FastAPI server:

   uvicorn main:app --reload --host 0.0.0.0 --port 8000

Frontend Setup

1. Navigate to frontend/:

   cd frontend

2. Install dependencies:

   npm install

3. Start the development server:

   npm run dev

   • Access at http://localhost:5173.

Running the Agent

1. Start the backend server.
2. Run npx localtunnel --port 3000 --subdomain mycustomname on terminal to expose backend API to internet for Twilio WebSocket connectivity.
3. Start the frontend.
4. Log in as:
   • Admin: admin/pass at /company-login
   • User: user/pass at /
5. Configure company info and knowledge via the admin dashboard.
6. Request a call as a user with a valid phone number.

Limitations

• No database; data is stored in memory.
• Hardcoded frontend credentials.
• Conversation history endpoint is not yet implemented.
• Requires a public URL (e.g., localtunnel) for Twilio WebSocket connectivity.

Future Improvements

• Add a database (e.g., Mongo) for persistence.
• Implement secure authentication (JWT, OAuth).
• Store and retrieve conversation logs.
• Enhance error handling and logging.