PrivateNurse

• 

  Homepage-Privacy First Multimodal Triage Interface

• 

  injury page-Injury Analysis with Severity Scoring

• 

  medication page-Medication Label Scan with Risk Detection

• 

  starting chat-Voice First Triage Interaction

• 

  responce-Model Generated Clinical Guidance with Urgency Score

Inspiration

Healthcare decisions are stressful, especially at 2am when your child has a fever or when chest tightness could be anxiety or something worse. Most people either panic and rush to the ER unnecessarily or dismiss something serious because they do not want to overreact. There is rarely a calm, knowledgeable voice in the room to help you decide which it is.

I built PrivateNurse to be that voice. Not a chatbot that dumps a Wikipedia article at you, but a nurse. Someone who listens, asks the right follow-up questions, and tells you clearly: this can wait, or go now.

What It Does

PrivateNurse is an AI-powered triage assistant you can talk to by voice or text.

You describe your symptoms. It responds like a nurse with 20+ years of clinical experience. It validates how you feel, asks focused follow-up questions, and gives you a clear action plan.

Every response includes a 1–10 clinical severity score that dynamically shifts the entire interface.

• Calm green for low severity
  
• Yellow for moderate concern
  
• Red for urgent
  

The background color, header, and severity bar transition visually, so urgency is felt, not just read.

Input Modes

• Typed text
  
• Live voice recording
  
• Prescription image scanning
  
• Injury camera capture
  
• Text-to-speech responses for hands-free interaction
  

How I Built It

Frontend

Single-page HTML, CSS, and JavaScript.
No framework. No build step.

All UI logic, audio handling, severity transitions, and chat state live in one file for speed and simplicity.

Backend

Node.js with Express acting as a secure proxy to AI APIs.
Audio files are converted using ffmpeg before speech-to-text processing.

AI Stack

Venice AI for:

• Large language model chat
  
• Speech-to-text
  
• Text-to-speech
  

Severity Extraction

The model appends a JSON token like:

{"severity": 7}

Built With

• api
• audio
• css3
• dotenv
• express.js
• html5
• javascript
• mediarecorder-api
• multer
• node.js
• vanilla-browser-js
• venice-ai-(llm-?-venice-uncensored)
• venice-ai-(speech-to-text)
• venice-ai-(text-to-speech)
• web
• web-speech-api-(speechrecognition)