Call Center powered by Twilio, AmiVoice API (Speech-to-Text) and ChatGPT (3.5)

🌸 Overview

This is a application that uses Twilio Programmable Voice, AmiVoice API (Speech-to-Text) and ChatGPT (3.5) to create a call center that can respond to customer inquiries. ⚠️ AmiVoice API is a voice recognition for Japanese language only. So this application only works for Japanese conversation.

📚 Technologies

• Node.js
• Websocket
• Twilio Programmable Voice with Twilio SDK for Node.js
• AmiVoice API with AmiVoice Websocket API
• ChatGPT with OpenAI API library for Node.js

👌 Prerequisites

You need to set the following information in .env to run this application:

AMIVOICE_API_KEY=[Obtainable from AmiVoice Cloud Platform]
OPENAI_API_KEY=[Obtainable from OpenAI API]
TWILIO_ACCOUNT_SID=[Obtainable from Twilio Console]
TWILIO_AUTH_TOKEN=[Obtainable from Twilio Console]
TWILIO_PHONE_NUMBER=[Obtainable from Twilio Console]
NGROK_DOMAIN=[Obtainable from ngrok]

Also, for running this application locally and make it accessible from the Internet, you need to install ngrok and run it with the following command. You will get a temporary domain name that can be accessed from the Internet. You will need to set this domain name (with https protocol and /twiml path) in the Twilio Console as the webhook URL for your phone number (example: https://54de-126-36-198-149.ngrok-free.app/twiml) You will need to set it inside .env as NGROK_DOMAIN.

ngrok http 3000

💻 Installation

npm install (only once)

ngrok http 3000
node ./server.js

📞 Usage

Call the Twilio phone number associated with the function and enjoy the conversation with a Japanese AI bot 🤖