Lumiere: Clinical Intelligence & Identity Resolution

Lumiere is a high-performance clinical data synthesis platform designed to resolve patient identities across fragmented health systems. It combines deterministic record matching with an AI-powered conversational layer to provide clinicians with a unified "Golden Record."

---

🏗️ Architecture

Lumiere is built with a decoupled, tri-service architecture:

Service	Technology	Port	Description
Frontend	Next.js 14+ (App Router)	3000	Clinician Portal & Dashboard
Node Backend	Node.js / Express	3001	EHR, FHIR, and HL7 Data Services
Python Engine	FastAPI / Python 3.10+	8000	Lumiere Clinical Intelligence (Resolution & Chat)

---

🚀 Getting Started

1. Prerequisites

• Node.js: 18.x or higher
• Python: 3.10 or higher
• API Key: A GroqCloud API Key for conversational fallback.

2. Installation

Clone the repository and install dependencies for each service:

# 1. Install Node Backend dependencies
cd backend && npm install

# 2. Install Frontend dependencies
cd ../frontend && npm install

# 3. Install Python Engine dependencies
cd ../backend/lumiere_python
pip install -r requirements.txt

3. Environment Setup

Create a .env file in backend/lumiere_python/ and add your Groq key:

GROQ_API_KEY=your_key_here

---

🛠️ Running the Application

To run the full stack, you need to start all three services. Open three terminal windows:

Terminal 1: Node Data Server

cd backend
node server.js

Terminal 2: Lumiere Python Engine

cd backend/lumiere_python
uvicorn main:app --reload --port 8000

Terminal 3: Next.js Frontend

cd frontend
npm run dev

The application will be accessible at http://localhost:3000.

---

🔍 Core Features

• Lumiere Search: Deep-resolve patient identities by name, phone, or email using weighted confidence scoring.
• Identity Synthesis: Automatic merging of low-stakes duplicates and human-in-the-loop review for high-stakes cases.
• Clinical Chat: Ask natural language questions about the unified patient record.
• Interoperability: Built-in support for FHIR R4, HL7 v2, and OCR-extracted clinical documents.

---

🛡️ Security & Privacy

Lumiere implements a "PII-Blind" intelligence flow. Sensitive identifiers are processed locally via deterministic algorithms, while conversational queries are sanitized before being sent to external LLMs.

---

🤝 Contribution

1. Pull the latest code: git pull origin main
2. Create your feature branch: git checkout -b feature/new-logic
3. Commit changes: git commit -m "feat: updated resolution weights"
4. Push to main: git push origin main