SafeCall: Emergency Dispatch Simulator

SafeCall is a web-based, real-time voice simulator designed for training and testing emergency dispatchers. It uses the ElevenLabs Conversational AI API to create dynamic, voice-driven emergency scenarios (like fire emergencies) and automatically evaluates the dispatcher's performance based on their responses.

Features

• Real-time Voice Conversation: Uses ElevenLabs Conversational AI over WebRTC for low-latency, natural voice interactions with an AI "caller".
• Dynamic Scenarios: Inject dynamic variables (e.g., caller name, emergency type, location) into the AI agent to instantly switch between different training scenarios (Fire, Medical, Robbery, etc.).
• Live Transcript & Audio Visualization: View the conversation transcript in real time, alongside an audio visualizer for mic input/output levels.
• Automated Grading System: Parses STT (Speech-to-Text) transcripts to grade dispatchers based on a custom rubric:
  • Did they ask for the location?
  • Did they check for building occupancy?
  • Did they assess medical injuries?
  • Did they give proper safety/evacuation instructions?
  • Did they communicate clearly (measuring turns and verbosity)?
• Geospatial Pinpointing (Google Maps Integration): Listen to the caller's location, drop a pin on a Google Map, and receive a score based on the Haversine distance to the actual target. Includes specific behaviors for "Training" (multiple attempts with best score tracking) and "Testing" (single locked guess).

Tech Stack

• Frontend: React.js, TypeScript, Vite
• Styling: Tailwind CSS v4, Lucide React (icons)
• AI / Voice: @elevenlabs/react (ElevenLabs Conversational AI)
• Maps / Location: @vis.gl/react-google-maps (Google Maps Platform)

Prerequisites

To run this application, you need the following API keys:

1. ElevenLabs Agent ID: Create a Conversational AI agent in the ElevenLabs Dashboard.
2. Google Maps API Key: Create a project in the Google Cloud Console and enable the Maps JavaScript API.

Getting Started

1. Clone the repository:

   git clone https://github.com/yourusername/elevenlabs-agent-app.git
   cd elevenlabs-agent-app

2. Install dependencies:

   npm install

3. Environment Setup: Create a .env file in the root directory and add your API keys:

   VITE_ELEVENLABS_AGENT_ID=your_elevenlabs_agent_id_here
   VITE_GOOGLE_MAPS_API_KEY=your_google_maps_api_key_here

4. Start the development server:

   npm run dev

5. Open the app: Navigate to http://localhost:5173 in your browser.

Project Structure

• src/pages/start/ - Landing page.
• src/pages/training/ - Training mode (review transcripts, practice map pinpointing).
• src/pages/testing/ - Testing mode (live conversation with AI, automated scoring, single-guess map).
• src/components/MapPanel.tsx - Google Maps integration for distance calculation and UI.
• src/components/ScorePanel.tsx - Evaluation engine scoring the transcript.
• src/components/TranscriptPanel.tsx - Live transcript UI.
• src/scenarios.ts - Scenario configurations and target locations.

License

This project is open-source and available under the MIT License.