Twilio AI Call Center

Overview

This project is a FastAPI-based service designed to handle voice interactions through Twilio with an AI bot.

Prerequisites

• Python 3.10+
• Twilio Account with a phone number
• Azure Speech Services Subscription
• Ngrok for local development and testing

Installation

1. Install the required dependencies:

   pip install -r requirements.txt

2. Set up environment variables: Create a .env file in the project root and add the necessary environment variables:

   TWILIO_ACCOUNT_SID=<your_twilio_account_sid>
   TWILIO_AUTH_TOKEN=<your_twilio_auth_token>
   OPENAI_API_KEY=<your_openai_api_key>
   SPEECH_KEY=<your_azure_speech_key>
   SPEECH_REGION=<your_azure_speech_region>
   NGROK_AUTHTOKEN=<your_ngrok_auth_token>

Running the Service

1. Start the FastAPI application and starting the ngrok tunnel:

   python main.py

2. Update the ngrok domain from the ngrok website:

   listener = ngrok.forward(f"http://localhost:{port}", authtoken_from_env=True, domain="your_ngrok_domain.ngrok-free.app")

Example Usage

1. Make a call to your Twilio phone number.
2. The call details are processed, and an audio stream is initiated.
3. Speech recognition processes the audio stream in real-time.
4. Recognized speech is sent to the bot agent, which generates a response.
5. The response is sent back through the WebSocket and played to the caller.

License

This project is licensed under the Apache License 2.0.

Contact

For any inquiries or issues, please contact [[email protected]].