Scribble

AI-powered clinical documentation for home healthcare.

Scribble helps clinicians capture visit notes faster — type or record audio, then let AI transform raw observations into structured SOAP notes. Every note links to a patient record, and the entire workflow lives in a single web app.

Live demo: scribble-alpha-gules.vercel.app

---

Videos

https://jam.dev/c/3f5efeda-d7dd-429e-b585-abb9fbbf8c51

Features

Clinical Documentation

• SOAP note generation — write raw visit notes and enhance them into structured Subjective / Objective / Assessment / Plan sections with one click
• Audio transcription — record directly in the browser or upload an audio file; transcribed via OpenAI Whisper and converted to SOAP
• Structured SOAP editor — four color-coded textareas (one per section) for editing each part independently; toggle between plain text and SOAP view
• Clinical templates — pre-filled wound assessment, medication review, vital signs, and general visit prompts to jumpstart documentation
• Undo / redo AI — every AI enhancement is reversible; step back to the previous version at any time

Patient Management

• Patient records — create and manage patient profiles with MRN, date of birth, and gender
• Patient-linked notes — every note is associated with one or more patients
• Patient detail view — see all notes for a specific patient in one place

App Experience

• Dashboard — practice-wide stats (total patients, notes today/this week) and recent activity feed
• Note viewer — click any note to view rendered SOAP sections, edit inline, enhance with AI, or delete
• Quick actions — floating button to create a note or patient from any page
• Landing page — product overview with animated SOAP mock and feature highlights
• Demo mode — one-click login with pre-seeded admin account

---

Screenshots

Dashboard	Notes

New Note	Audio Recording

Enhanced SOAP Note	Patients

Patient Detail	New Patient

Login

---

Tech Stack

Layer	Technology
Frontend	React 18, TypeScript, Vite, Tailwind CSS v4, shadcn/ui
Backend	Node.js 20, Express, TypeScript, Zod
Database	PostgreSQL 16, Prisma ORM
AI	OpenAI Whisper (transcription), GPT-4o-mini (SOAP generation)
Auth	JWT, bcryptjs
Testing	Vitest, Supertest
CI	GitHub Actions

Deployment

Service	Platform
Frontend	Vercel
API Server	Railway
Database	Neon (serverless PostgreSQL)

---

Local Development

Prerequisites

• Docker and Docker Compose
• OpenAI API key (whisper-1 + gpt-4o-mini)

Setup

git clone <repo-url>
cd lime
cp .env.example .env

Set your OpenAI key in .env:

OPENAI_API_KEY=sk-...

Start

First run — seed the database with demo patients and admin account:

SEED=on docker compose up -d --build

Subsequent runs:

docker compose up -d

Access

Service	URL
Web UI	http://localhost:8080
API	http://localhost:4000

Login credentials:

Email:    admin@scribe.local
Password: admin123

Hot-reload (dev mode)

SEED=on docker compose --profile dev up -d --build

Starts Vite dev server at http://localhost:5173 with HMR and API proxy.

---

Project Structure

├── server/
│   ├── prisma/          # Schema, migrations, seed data
│   ├── src/
│   │   ├── routes/      # Express route handlers
│   │   ├── services/    # Business logic (auth, notes, AI)
│   │   ├── middleware/  # JWT auth, error handling
│   │   └── storage/     # Audio file storage
│   └── tests/           # Integration tests
├── web/
│   ├── src/
│   │   ├── auth/        # Login, session management
│   │   ├── api/         # API client
│   │   ├── components/  # Shared UI (shadcn/ui, modals)
│   │   ├── features/    # Dashboard, patients, notes, layout
│   │   └── pages/       # Landing page
│   └── public/
└── docker-compose.yml

---

Notes

• Single user — one admin account, no registration or multi-user support
• Synchronous AI — transcription and enhancement run inline; production would use a job queue
• 25 MB audio limit — matches the OpenAI Whisper API constraint
• JWT in localStorage — sufficient for demo; production should use HttpOnly cookies