- Video Walkthrough: Watch Demo
- Presentation Slides: View Slides
PreVizAI is a modern, HIPAA-conscious healthcare web application built with Next.js and Cedar OS Voice Capabilities, designed to streamline and transform pre-appointment patient intake through natural, voice-powered AI interactions.
The Problem
Traditional patient intake relies on long, static forms—slow and often incomplete. Doctors spend 15+ minutes on repetitive questions, risking misdiagnosis and frustrating patients.
Our Solution
PreVizAI replaces traditional forms with a natural, voice-powered patient intake experience. Using Cedar OS Voice Capabilities, patients can speak freely while the interface guides them through a conversational workflow. Responses are captured and structured into a medical report, giving doctors a clear, organized summary before the appointment. Some AI-powered follow-ups are handled by GPT-4, enhancing the conversation with smart, context-aware questions where needed.
| Feature | Description | Cedar OS / AI Technology | Impact for Healthcare Providers |
|---|---|---|---|
| Voice-Powered Intake | Patients describe symptoms naturally in a clean interface. | Cedar OS (real-time Speech-to-Text & AI voice synthesis) | Captures rich, natural data and improves accessibility. |
| AI-Generated Medical Reports | Converts conversations into structured reports with chief complaint, history, and key insights. | OpenAI GPT-4 API | Saves clinician time and provides ready-to-use clinical insight. |
| Doctor Dashboard | Central hub to send intake requests and review reports. | Next.js / TypeScript | Streamlines workflow and highlights high-priority patients. |
| Knowledge Graph Visualization (optional) | Visual graph of symptoms, conditions, and treatments. | Neo4j Graph DB | Reveals complex clinical connections for advanced diagnostics. |
| HIPAA-Mindful Design | Secure handling of sensitive data. | Flask Backend + Secure Architecture | Ensures compliance and builds patient trust. |
- Doctor Action 👨⚕️ – Doctor sends a pre-visit intake request from the
/doctordashboard. - Patient Interface 📲 – Patient opens
/patientand begins the Cedar OS-powered voice dialogue. - AI Conversation 🗣️ – Cedar OS transcribes speech; GPT-4 asks smart follow-ups (duration, severity, history, red flags).
- Report Generation 📝 – Flask backend turns the transcript into a structured Medical Report.
- Doctor Review ✅ – Doctor reviews the report before the appointment, reducing visit time and improving accuracy.
| Area | Technology | Reason |
|---|---|---|
| Frontend | Next.js 15, TypeScript, Tailwind CSS | Fast, server-rendered, type-safe web app. |
| Core UI | Cedar OS Components, Lucide React Icons | Accessible, professional healthcare UI. |
| Voice & AI | Cedar OS SDK, OpenAI GPT-4 API | High-fidelity voice interaction + world-class language processing. |
| Backend | Flask (Python API) | Lightweight and flexible for AI/ML logic & secure data processing. |
| Database (optional) | Neo4j | Ideal for representing complex symptom-condition relationships. |
- Node.js 18+ & npm
- Python 3.11+ & pip
- OpenAI API Key
- Git
- Neo4j (optional for Knowledge Graph)
# Clone the repository
git clone <repository-url>
cd PreVizAI # or the folder name where you cloned the repo
# Install frontend dependencies
npm install
# (Optional) Create Python virtual environment
python3 -m venv .venv
# macOS / Linux
source .venv/bin/activate
# Windows (PowerShell)
# .venv\Scripts\Activate.ps1
# Install backend dependencies
pip install -r backend/requirements.txtCreate the following files and add your credentials.
Frontend — .env.local (in the repo root):
NEXT_PUBLIC_OPENAI_API_KEY=your_openai_api_key_here
NEXT_PUBLIC_BACKEND_URL=http://localhost:5000Backend — backend/.env:
OPENAI_API_KEY=your_openai_api_key_here
# Neo4j settings are optional if you are not using the Knowledge Graph
NEO4J_URI=bolt://localhost:7687
NEO4J_USER=neo4j
NEO4J_PASSWORD=your_neo4j_password_here
FLASK_ENV=development
FLASK_DEBUG=TrueSecurity note: Never commit
.env.localorbackend/.env. Add them to.gitignore.
Open two terminals.
Backend
cd backend
# activate virtualenv if you created one
# source .venv/bin/activate (macOS/Linux)
python run_server.py
# Backend running at: http://localhost:5000Frontend
# from repo root (where package.json lives)
npm run dev
# Frontend running at: http://localhost:3000Visit: http://localhost:3000/doctor
- Send intake requests to patients.
- Review completed reports and explore the Knowledge Graph (if enabled).
Visit: http://localhost:3000/patient
- Click the microphone to start Cedar OS voice intake.
- Describe symptoms naturally and answer follow-up questions.
- Submit when finished — doctor receives the report.
The AI generates a structured document that includes:
- Chief Complaint – patient’s primary concern in their own words
- History of Present Illness (HPI) – duration, severity, onset, context
- Triggers & Relieving Factors – diagnostic clues
- Medical History & Medications
- Red Flag Symptoms – urgent warning signs
- Functional Impact – effect on daily life
- Recommended Follow-up Actions – suggested next steps
- Arjun Pun Magar – Frontend & Cedar OS Integration
- Vinh Pham – Backend & Flask API ( Neo4j Graph Visualization)
- Om Nepal – AI Prompt Engineering / Medical Report Design
- Jonathan Navarro – Project Manager
Special thanks to the Cedar OS team for providing the voice SDK and support during HackGT.
This project is licensed under the MIT License. See the LICENSE file for details.