Triage.OS - Advanced Clinical Intelligence & Ward Management

Triage.OS is a next-generation healthcare operating system designed to bridge the gap between high-volume clinical data and actionable insights. By integrating high-fidelity vital sign simulation with Large Language Models (LLMs), it reduces administrative burden and enhances patient safety through automated documentation and real-time monitoring.

🎯 Problem Statement & Core Ideas

In modern clinical settings, healthcare professionals are overwhelmed by:

• Documentation Tax: Spending excessive time on manual SOAP note entry.
• Data Fragmentation: Difficulty in synthesizing real-time vitals with patient history.
• Cognitive Overload: Managing high-occupancy wards with varying patient urgency.

Triage.OS solves this by:

1. AI-Driven Synthesis: Automating the conversion of unstructured clinical observations into structured medical records.
2. Predictive Triage: Using Machine Learning to identify deteriorating patients before they reach a critical state.
3. Unified Control Plane: Providing a centralized, role-based interface for the entire clinical team.

🚀 Key Features

1. AI-Powered Clinical Pipeline (SOAP.AI)

• Medical NER (Named Entity Recognition): Automatically extracts clinical entities (Symptoms, Drugs, Measurements, Conditions) from raw medical notes using specialized Transformer models.
• Automated SOAP Documentation: Transforms unstructured clinical text into structured Subjective, Objective, Assessment, and Plan (SOAP) formats using LLMs (via OpenRouter).
• Clinical Urgency Classification: Heuristic and ML-based classification of patient urgency levels to prioritize care.
• Risk Scoring: Predictive mortality and deterioration risk scoring based on vitals and historical data (XGBoost/Shap).

2. Real-Time Patient Monitoring

• Vitals Simulator: Background processes simulate heart rate, SpO2, blood pressure, and temperature for testing and training.
• Dynamic Dashboards: Real-time visualization of patient vitals using Recharts with instant history tracking.
• Alert System: Instant notifications for critical vital sign deviations or high-risk scores.

3. Ward Management

• Clinical Kanban Board: Drag-and-drop interface for managing patient flow (Waiting → In Progress → Done) using @hello-pangea/dnd.
• Role-Based Access: Specialized dashboards for Doctors, Nurses, and Administrators with tailored metric views.
• Shift Management: Dedicated panels for nurse shift swaps, schedule tracking, and ward-wide announcements.
• Ward Overview: High-level visualization of bed occupancy, patient distribution, and active clinical alerts.

4. Intel-Assisted Chat (RAG)

• Clinical Knowledge Base: Searchable medical documentation indexed using FAISS for lightning-fast retrieval.
• RAG Chatbot: Conversational AI that answers clinical queries based on the internal knowledge base to assist in clinical decision making.

---

🛠 Tech Stack

Backend

• Framework: FastAPI (Python)
• Database: PostgreSQL (Supabase) & SQLite (Local testing)
• ORM: SQLAlchemy
• AI/ML Libraries: transformers, torch, spacy, scikit-learn, xgboost, faiss-cpu, sentence-transformers
• LLM Integration: OpenRouter (NVIDIA Nemotron, etc.)

Frontend

• Framework: React 19 (Vite)
• State Management: React Context API
• UI Components: Modern CSS & Lucide-React Icons
• Charts: Recharts
• Drag-and-Drop: Hello Pangea DnD

Infrastructure

• Authentication: Supabase Auth
• Real-time: Supabase Real-time subscriptions
• Storage: Supabase Storage (for clinical documents)

---

📂 Project Structure

Triage.OS/
├── Backend/                # FastAPI Application
│   ├── ai/                 # Core AI Models (NER, Risk Scorer, SOAP Pipeline)
│   │   ├── rag/            # Retrieval Augmented Generation system
│   │   ├── ner_model/      # Medical NER implementation
│   │   └── vitals_simulator.py # Background vitals generation
│   ├── routes/             # API Endpoints (Chat, Patients, SOAP, Vitals, Kanban)
│   ├── models.py           # SQLAlchemy Database Models
│   └── main.py             # Server Entry Point
├── frontend/               # React (Vite) Application
│   ├── src/
│   │   ├── api/            # Supabase & API Service Layer
│   │   ├── components/     # UI Components (Kanban, Dashboards, SOAP View)
│   │   ├── contexts/       # Auth & Notification Contexts
│   │   └── hooks/          # Custom Hooks (Vitals simulation)
├── supabase/               # SQL Migrations and Seeding
└── SETUP_INSTRUCTIONS.txt  # Comprehensive model setup guide

---

⚙️ Installation & Setup

Prerequisites

• Python 3.9+
• Node.js 18+
• Supabase Account (for live data features)

Backend Setup

1. Model Extraction: Extract triage_models.zip into Backend/ai/. Ensure the path exists: Backend/ai/ner_model/...
2. Environment Variables: Create a .env in Backend/:

   OPENROUTER_API_KEY=your_key
   SUPABASE_URL=your_url
   SUPABASE_KEY=your_key
   DATABASE_URL=sqlite:///./triage.db

3. Install Dependencies:

   cd Backend
   pip install -r requirements.txt

4. Run Server:

   python main.py

Frontend Setup

1. Install Packages:

   cd frontend
   npm install

2. Run Development Server:

   npm run dev

---

🧠 Clinical Workflow Breakdown

1. Patient Intake: Patient is added to the system via the Patients route.
2. Triage: Nurse enters raw clinical notes. The SOAP Pipeline extracts entities and suggests a SOAP note and urgency level.
3. Monitoring: Vitals Simulator generates live data; Risk Scorer calculates deterioration probability.
4. Coordination: Doctors use the Kanban Board to move patients through treatment phases.
5. Retrieval: Clinicians use the RAG Chat to verify treatment protocols against the knowledge base.

---

📄 License

Internal use only. Part of the Hello project suite.