SeiAI Voicebot Project

Overview

This project is designed to create a voicebot that leverages various technologies to provide assistance with Wise financial transactions and payments. The voicebot integrates with LiveKit for real-time communication, Deepgram for speech-to-text, OpenAI for language models, and Cartesia for text-to-speech. The system also includes scripts to crawl the data from websites and store the same into ChromaDB, manage inbound SIP trunks, dispatch rules, and handle FAQs using vector embeddings.

Sample Video

sample_video.mp4

Setup

Prerequisites

• Python 3.9+
• LiveKit account
• Deepgram account
• OpenAI account
• Cartesia account
• Twilio account (for SIP trunks)

Installation

1. Clone the repository and move to the folder:

   cd SeiAI

2. Install the required Python packages:

   pip install -r requirements.txt

3. Set up the environment variables:

   • Create a .env file in the project root.
   • Copy the contents of .env.example into .env.
   • Fill in the required environment variables with your account details.
   • Ensure that the .env file contains the following variables:
     • LIVEKIT_URL
     • LIVEKIT_API_KEY
     • LIVEKIT_API_SECRET
     • DEEPGRAM_API_KEY
     • OPENAI_API_KEY
     • CARTESIA_API_KEY

Running the Voicebot

1. Create the inbound SIP trunk:

• Create an inbound SIP trunk using Livekit's API in order to handle the incoming calls.

  python inbound_trunk.py

• Alternatively, you can use Livekit's CLI to make the inbound SIP trunk.
• Make sure to configure the inbound_trunk.json file with the correct details.

  lk sip in create inbound_trunk.json

2. Set up the dispatch rule:

• Create a dispatch rule using Livekit's API in order to route the incoming calls to the voicebot.

  python dispatch_rule.py

• Alternatively, you can use Livekit's CLI to make the dispatch rule.
• Make sure to configure the dispatch_rule.json file with the correct details.

  lk sip dispatch create dispatch_rule.json

3. Store FAQ data:

• store_data.py script is used to crawl the data from the website links and then store that crawled data in the database.
• Make sure to recheck the links in the file before running it.

  python store_data.py

  • As we are using Craw4AI to crawl the data from websites, it will throw an dependency error.
  • Make sure to downgrade the psutill package to version 5.9.3 and the protobuf package to version 5.29.3. After you run the store_data.py script.

4. Start the voicebot:

• Run the sei_simple_voicebot.py script to start the voicebot.

  python sei_simple_voicebot.py dev

JSON Configuration Files

• inbound_trunk.json: Configure the inbound SIP trunk details.
• dispatch_rule.json: Configure the dispatch rule for routing calls.