AI-Powered Car Dealership Assistant

This project consists of two main components: a voice-based AI assistant for handling customer inquiries and a dashboard for monitoring and analyzing call data.

Tech Stack

Caller Agent

• Node.js
• Express.js
• WebSocket (express-ws)
• OpenAI GPT for natural language processing
• Deepgram for speech-to-text transcription
• Twilio for handling phone calls
• MongoDB for data storage

Dashboard

• Next.js 14
• React
• TypeScript
• Tailwind CSS
• Chart.js and react-chartjs-2 for data visualization
• MongoDB for data retrieval

Caller Agent

The caller agent is built on top of the call-gpt project by Twilio Labs, which provides a foundation for creating AI-powered voice applications using Twilio Media Streams. We've extended and customized this project to fit the specific needs of our car dealership assistant.

Key Features:

1. Voice-to-text transcription using Deepgram
2. Natural language processing with OpenAI's GPT model
3. Text-to-speech conversion for AI responses
4. Integration with Twilio for handling phone calls
5. MongoDB integration for storing call logs and transcriptions
6. Real-time audio streaming and processing

Main Components:

• app.js: The main application file that sets up the Express server and WebSocket connections
• gpt-service.js: Handles interactions with the OpenAI API
• tts-service.js: Manages text-to-speech conversion
• transcription-service.js: Handles speech-to-text conversion
• db-service.js: Manages database operations with MongoDB
• recording-service.js: Handles call recording functionality
• stream-service.js: Manages audio streaming

Function Calls:

The AI assistant can perform various functions, such as:

• Checking car availability
• Getting car prices
• Scheduling test drives
• Scheduling maintenance appointments
• Scheduling oil changes

These functions are defined in the functions directory and are dynamically loaded based on the function-manifest.js file.

Dashboard

The dashboard is a Next.js application that provides a user interface for viewing and analyzing call data.

Key Features:

1. Overview of call statistics
2. Detailed call logs
3. Transcription viewer
4. Charts for data visualization
5. Appointment management

Main Components:

• app/dashboard/page.tsx: The main dashboard page
• components/CallLogDetail.tsx: Displays detailed information for a specific call
• components/CallLogsList.tsx: Lists all call logs
• components/charts.tsx: Renders various charts for data visualization
• components/AppointmentsList.tsx: Manages and displays appointments

UI Components:

The dashboard uses custom UI components built with Tailwind CSS and Radix UI primitives, including:

• Cards
• Tables
• Tabs
• Badges
• Charts

Setup and Installation

1. Clone the repository

2. Install dependencies for both the caller agent and dashboard:

   cd caller-agent && npm install
   cd ../dashboard && npm install
   

3. Set up environment variables:

   • For the caller agent, copy .env.example to .env and fill in the required values
   • For the dashboard, set up the necessary environment variables, including the MongoDB connection string

4. Start the caller agent:

   cd caller-agent && npm run dev
   

5. Start the dashboard:

   cd dashboard && npm run dev
   

Testing

The caller agent includes Jest tests for various functions. Run tests using:

npm run test

Deployment

Caller Agent

The caller agent can be deployed using Fly.io. Modify the fly.toml file with your app name and use the following commands:

fly launch
fly deploy
fly secrets import < .env

Dashboard

The dashboard can be deployed to your preferred Next.js hosting platform, such as Vercel or Netlify. Follow the platform-specific deployment instructions.

Development

• The project uses ESLint for code linting
• Tailwind CSS is used for styling in the dashboard
• TypeScript is used in the dashboard for type safety

License

This project is licensed under the MIT License. See the LICENSE file for details.