Bandwidth + OpenAI Live SIP Integration - Python

Table of Contents

• Description
• Pre-Requisites
• Environmental Variables
• Running the Application
• Callback URLs
  • Ngrok

Description

This is a sample application that demonstrates how to use Bandwidth's Programmable Voice API with OpenAI's GPT-Live-1 model via the Live API SIP interface to create a real-time AI-powered voice assistant. Unlike the WebSocket integration, the Live SIP integration uses OpenAI's SIP Connector to handle media directly — your application only needs to handle webhooks and send commands via the sideband.

Pre-Requisites

In order to use this integration you need:

• A Bandwidth Universal Platform account with a Voice Configuration Package and SIP Connector enabled
• Your OpenAI Project ID — used to configure your Bandwidth trunk destination
• Your OpenAI API Key
• A publicly accessible server to host your webhook application (e.g., using ngrok)
• Docker (optional)

Your Bandwidth trunk must point to sip:$PROJECT_ID@sip.api.openai.com;transport=tls as the termination destination.

Environmental Variables

The sample app uses the below environmental variables.

OPENAI_API_KEY   # Your OpenAI API Key (must have access to the Live API)
REFER_TO         # The phone number to transfer calls to (E.164 format, e.g. +19195554321)
LOG_LEVEL        # The logging level for the application (e.g. INFO, DEBUG)
LOCAL_PORT       # (optional) The local port for the application (default: 3000)

Create a .env file in the root of the project:

OPENAI_API_KEY="your_openai_api_key_here"
REFER_TO="+19195554321"
LOG_LEVEL="INFO"
LOCAL_PORT=3000

Running the Application

This application is built using Python 3.13. You can use pip to install the required packages, or Docker Compose to run the application.

# Using Docker Compose
docker compose up --build

# Using Python
python -m venv .venv
source .venv/bin/activate
cd app
pip install -r requirements.txt
python main.py

A successful startup will log:

INFO:     Uvicorn running on http://0.0.0.0:3000 (Press CTRL+C to quit)
INFO:     Application startup complete.

Callback URLs

Below are the callback paths exposed by this application:

• /health
• /webhooks/openai/live/transport/inbound — receives live.transport.incoming events from OpenAI

Ngrok

A simple way to set up a local callback URL for testing is to use the free tool ngrok.

After you have downloaded and installed ngrok run the following command to open a public tunnel to your port ($LOCAL_PORT)

ngrok http $LOCAL_PORT

You can view your public URL at http://127.0.0.1:4040 after ngrok is running.

Configure your OpenAI project's webhook URL to https://<your-ngrok-url>/webhooks/openai/live/transport/inbound.