Voice chatbot for realistic AI cold calls
Top 94.9% on sourcepulse
This project provides a voice chatbot for practicing realistic cold calls using AI. It targets sales professionals and anyone looking to improve their phone communication skills by simulating conversations with an AI bot powered by Twilio, Deepgram, and OpenAI.
How It Works
The application leverages Twilio for phone call management, Deepgram for speech-to-text, and OpenAI's GPT-4 for natural language understanding and response generation. A Python script acts as the backend, receiving incoming calls via a Twilio webhook. This webhook is exposed to the internet using ngrok, which forwards requests to the local Python server. The script processes audio, sends it to Deepgram for transcription, feeds the transcript to GPT-4 for response generation, and then uses Twilio's TwiML to speak the AI's response back to the caller.
Quick Start & Requirements
pip install -r requirements.txt
..env
file. Set up a Twilio number and configure its webhook to point to an ngrok URL. Run ngrok http --hostname=coldcallgenius.ngrok.io 8080
and python3 start.py
. Systemd services are recommended for resilience.Highlighted Details
Maintenance & Community
The repository is maintained by kevingduck. No specific community channels or roadmap details are provided in the README.
Licensing & Compatibility
The README does not explicitly state a license. It is a demonstration project, and commercial use would require careful review of the underlying service terms (Twilio, Deepgram, OpenAI) and the project's own licensing if specified elsewhere.
Limitations & Caveats
The setup relies heavily on external services (Twilio, Deepgram, OpenAI, ngrok) and requires careful configuration of API keys and webhooks. The use of ngrok without a paid subscription means the public URL is temporary, necessitating a paid plan or alternative hosting for persistent availability. The project is presented as a demo, and production-readiness is not guaranteed.
1 year ago
1 day