Software Demo Real-Time Voice Agent

🚀 Features

• Real-Time Interaction: Instant responses and actions based on voice input
• Customizable Experience: Easily adapt to your SaaS application
• Seamless Browser Automation: Powered by Playwright for reliable web interactions

🏗️ How It Works

This Voice Agent Architecture is split into 3 different layers:

1. Speech-to-Text (STT) Layer 🎤

• Provider: Deepgram
• Converts user's spoken words into text in real-time
• Handles various accents and speech patterns with high accuracy
• Voice Activity Detection: Uses SileroVADAnalyzer for intelligent endpointing and turn detection

2. Large Language Model (LLM) Layer 🧠

• Provider: OpenAI (GPT-4)
• The brain of the operation where all the logic happens
• Processes user requests, understands context, and decides on actions
• Makes intelligent tool calls to navigate and interact with the application

3. Text-to-Speech (TTS) Layer 🔊

• Provider: Cartesia
• Converts AI responses back into natural-sounding speech
• Uses advanced voice synthesis for a professional, engaging experience

Browser Automation 🌐

• Technology: Playwright
• Handles all web interactions, navigation, and UI manipulations
• Includes visual cursor animations for a polished demo experience

🛠️ Setup

Prerequisites

• Python 3.8+
• API keys from the following services:
  • Deepgram for speech-to-text
  • OpenAI for the language model
  • Cartesia for text-to-speech

Installation

1. Clone the repository:

   git clone https://github.com/adriablancafort/software-demo-realtime-voice-agent.git
   cd software-demo-realtime-voice-agent

2. Install dependencies:

   pip install -r requirements.txt

3. Set up environment variables: Create a .env file in the root directory with your API keys:

   DEEPGRAM_API_KEY=your_deepgram_api_key_here
   OPENAI_API_KEY=your_openai_api_key_here
   CARTESIA_API_KEY=your_cartesia_api_key_here
   

4. Run the application:

   python main.py

5. Access the demo:

   • Open your browser and navigate to http://localhost:7860/client/
   • Click "Connect" to start the voice interaction
   • Enjoy your personalized software demo! 🎉

🎨 Customization for Your SaaS

To adapt this agent for your own SaaS application:

1. Update Prompts: Modify the prompts in custom/prompts.py to match your application's context and use cases
2. Configure Selectors: Update the CSS selectors in custom/selectors.py to target the specific elements in your web application

The agent uses tool calls to perform actions like:

• Navigating between pages
• Clicking on elements
• Typing into input fields
• Scrolling the page

Simply update the selectors and prompts to point to your application's UI elements and workflows.