🩺 HealthTriage AI — Intelligent Medical Triage

An intelligent triage system built with .NET and the Microsoft Agent Framework, simulating real-time medical evaluations powered by AI agents that assess risk, recommend specialists, and provide first-aid guidance.

---

🧠 About the Project

HealthTriage AI is a simulated healthcare triage platform where multiple AI agents collaborate in real time to evaluate a patient’s condition.

The goal is to demonstrate how agent orchestration can solve real-world healthcare scenarios without handling sensitive medical data.

Each triage request goes through a complete evaluation pipeline:

• Risk analysis — determining urgency based on symptoms and vitals.
• Specialist recommendation — matching symptoms to the correct medical specialty.
• AI advice — generating a short and safe orientation message.
• First-aid guide — providing immediate, non-diagnostic care instructions.

All this happens live in the UI through a SignalR hub, showing how the case evolves step by step — as if multiple digital assistants were collaborating like a real medical team.

---

⚙️ Core Stack

Layer	Technologies
Backend	.NET 10 + Microsoft Agent Framework + SignalR
Frontend	React + Fluent UI + Vite
AI Connectors	OpenAI (GPT)
Testing	xUnit + custom hub capture for SignalR calls
Architecture	Clean Architecture with service orchestration pattern

---

🧩 Project Structure

src/
 ├─ HealthTriageAI.ApiService/      → Core API + Agent Orchestration
 │                                   - Hosts SignalR hub (/hubs/triage)
 │                                   - Coordinates 5 agents (Symptom, Risk, Specialist, Advice, FirstAid)
 │                                   - Exposes endpoints for case reporting (/api/triage/report)
 │                                   - Uses Microsoft Agent Framework for LLM orchestration
 │
 ├─ HealthTriageAI.Web/             → Frontend (React + Fluent UI)
 │                                   - Real-time dashboard for triage cases
 │                                   - Displays live case flow (symptom → risk → specialist → advice)
 │                                   - Responsive layout with filtering and search
 │                                   - Connects directly to SignalR hub
 │
 └─ HealthTriageAI.Tests/           → Automated Tests (xUnit)
                                     - Unit tests for RiskAgent, SpecialistAgent, TriageCoordinator
                                     - Custom fake hub context for SignalR message capture

---

🌟 Key Features

• 🔄 Real-time triage pipeline via SignalR
• 🧩 Agent orchestration using Microsoft Agent Framework
• 💬 GPT integration for dynamic and contextual medical advice
• 🧪 Extensive testing without external API costs (mocked advice agents)
• 🩹 First-aid assistant with contextual pre-care messages
• 💻 Modern Fluent UI dashboard with responsive layout and filtering

---

💡 How It Works

1. The user submits a triage form (name, age, symptoms, vitals).
2. The backend triggers a TriageCoordinator that orchestrates five agents:
   • SymptomAgent → extracts key data from text.
   • RiskAgent → classifies the urgency level.
   • SpecialistAgent → selects the correct specialty.
   • AdviceAgent → generates safe instructions.
   • FirstAidAgent → produces immediate non-diagnostic care tips.
3. Each step emits real-time updates to the UI via SignalR.
4. The frontend displays the evolution of each case visually — from report to advice.

---

🧪 Testing Strategy

Unit and integration tests ensure system stability without incurring AI usage costs.

Test Area	Description
RiskAgentTests	Evaluates risk classification logic based on vitals.
SpecialistAgentTests	Maps symptom keywords to medical specialties.
TriageCoordinatorTests	Validates SignalR sequence and orchestration flow using a fake hub context.

AI agents are mocked in tests via FakeAdviceAgent and FakeFirstAidAgent to prevent token usage.

---

⚠️ Guarantees

Before starting, you have to create a new file in HealthTriageAI.ApiService project named "appsettings.json", and put this value inside:

{
  "Logging": {
    "LogLevel": {
      "Default": "Information",
      "Microsoft.AspNetCore": "Warning"
    }
  },
  "AllowedHosts": "*",
  "OpenAI": {
    "ApiKey": "<ApiKey>",
    "Model": "gpt-4o-mini"
  }
}

Please, make sure you go change "" for your real api key from OpenAI, follow site to get it: OpenAI Api keys

---

🚀 How to Run

Backend

cd HealthTriageAI.ApiService
dotnet run

Frontend

cd HealthTriageAI.Web
npm install
npm run dev

Tests

cd HealthTriageAI.Tests
dotnet test

---

🌐 Live Demo

🔗 https://healthtriage.mathszoke.com

---

📫 Contact

📧 Email: matheusszoke@gmail.com
💼 LinkedIn: linkedin.com/in/matheusszoke

---

_{Made with 💚 by Matheus Szoke — powered by the Microsoft Agent Framework}