AI Call MVP

A minimal Flask service for handling AI-powered voice calls via Twilio.

Project Structure

app/
  __init__.py  # create_app() factory
  backend/
    __init__.py
    ngrok_control.py   # Utility functions for managing ngrok server
    routes.py          # health + /voice endpoints
    rag.py             # retrieval + Gemini integration
    convo.py           # ConversationController
    scraper.py         # Web scraper for online knowledge base
  frontend/
    ...        # React app files
cli.py         # helper commands
run.py         # Flask app launcher

Setup

Build from source

1. Install dependencies and set up the project:

pip install -e .
cd app/frontend && npm install

2. Run the Flask application:

# Option 1: Using run.py
python run.py

# Option 2: Using Flask CLI
flask --app app run

# Option 3: With debug mode
flask --app app run --debug

3. Start React app from inside the frontend folder

npm run dev

Docker container

docker compose up --build

Endpoints

• GET /health - Health check endpoint
• POST /voice - Twilio webhook for voice calls
• POST /ngrok/start - Start a new ngrok server
• POST /ngrok/stop - Stop the current ngrok server
• GET /ngrok/status - Get ngrok server status

CLI Commands

# Test call service
python cli.py calltest

# Scrape a website
python cli.py scrape <url>

# Ingest a document
python cli.py ingest

# Start ngrok service
python cli.py ngrok

# Show all configurations
python cli.py --help

Web App Demo

Environment Variables

Create a .env file with the necessary API keys:

GOOGLE_API_KEY=<your-gemini-api-key>
TWILIO_ACCOUNT_SID=<your-twilio-account-sid>
TWILIO_AUTH_TOKEN=<your-twilio-auth-token>
TWILIO_PHONE_SID=<your-twilio-number-sid>
NGROK_AUTH_TOKEN=<your-ngrok-auth-token>