🏥 CareRouter

CareRouter is an AI-powered mental health triage and navigation system designed to connect individuals in crisis with the right support at the right time. By streamlining the path to help, CareRouter reduces cognitive load during high-stress moments and provides actionable resources immediately.

🌟 Key Features Conversational Assessment: An empathetic, 6-question chatbot interface designed to minimize stress during a crisis.

AI-Powered Triage: Utilizes Gemini 2.5 Flash to classify severity (Levels 1-4), urgency, and issue type with personalized messaging.

Location-Based Matching: Integration with Google Maps API to identify nearby mental health facilities within a 5km radius.

Smart Filtering: A dual-LLM architecture that filters resources based on specific user constraints such as cost, transportation, and language.

Immediate Coping Tools: Access to three evidence-based exercises for immediate use while waiting for professional care.

Accessibility First: Features ElevenLabs Text-to-Speech integration for enhanced accessibility.

Safety-First Design: Includes automated crisis keyword detection that bypasses AI for immediate 911/988 routing.

Secure Authentication: JWT-based secure login to track assessment history safely.

🛠 Tech Stack Frontend Next.js 16: React framework utilizing TypeScript and App Router.

Tailwind CSS: For a responsive, utility-first UI design.

@vis.gl/react-google-maps: Interactive map integration.

Lucide React: Comprehensive icon library.

ElevenLabs API: AI-driven Text-to-Speech functionality.

Backend FastAPI: High-performance, asynchronous Python web framework.

Gemini 2.5 Flash: Advanced LLM for intelligent triage and resource filtering.

Google Places API: Powering location-based facility discovery.

SQLModel & SQLite: Pydantic-based SQL toolkit (PostgreSQL-ready).

Security: JWT Authentication with bcrypt password hashing via Python-Jose and Passlib.

⚙️ Setup and Installation

1. Backend Setup Bash

Navigate to the backend directory

cd backend

Create and activate a virtual environment

python -m venv venv

On Windows: .\venv\Scripts\activate | On macOS/Linux: source venv/bin/activate

Install dependencies

pip install -r requirements.txt Environment Variables (.env): Create a .env file in the backend directory:

Code snippet GEMINI_API_KEY=your_gemini_api_key_here GOOGLE_MAPS_API_KEY=your_google_maps_api_key_here SECRET_KEY=your_jwt_secret_key_here Note: Use an API key without HTTP referrer restrictions for the backend.

2. Frontend Setup Bash

Navigate to the frontend directory

cd frontend

Install dependencies

npm install Environment Variables (.env.local): Create a .env.local file in the frontend directory:

Code snippet NEXT_PUBLIC_GOOGLE_MAPS_API_KEY=your_frontend_google_maps_key NEXT_PUBLIC_ELEVENLABS_API_KEY=your_elevenlabs_key_optional 🚀 Running the Application Start Backend From the backend directory:

Bash uvicorn main:app --reload Server runs at: http://127.0.0.1:8000

Start Frontend From the frontend directory:

Bash npm run dev Application available at: http://localhost:3000

🛡 Safety & Privacy Crisis Detection: Specific keywords trigger immediate emergency resource displays, bypassing AI logic.

Conservative Bias: The system is engineered to err on the side of higher urgency when triage data is uncertain.

Graceful Degradation: Core crisis resources remain accessible even if third-party APIs fail.

PHIPA-Ready: Architecture designed with Ontario health privacy standards in mind.