Clinical Scribe for macOS

A native macOS app for generating clinical notes from doctor-patient conversations using local LLMs via Ollama.

Features

• Split Panel UI: Conversation input on the left, clinical note output on the right
• Real-time Streaming: Watch the clinical note generate word-by-word
• Multiple Model Support: Use gpt-oss-20b, Mistral, LLaMA, or any Ollama model
• Privacy First: All processing happens locally on your Mac
• Copy to Clipboard: One-click copy of generated notes

Screenshot

┌─────────────────────────────────────────────────────────────────────┐
│  Clinical Scribe                                                    │
├──────────────────────────────┬──────────────────────────────────────┤
│ Doctor-Patient Conversation  │ Clinical Note          [Model ▼]    │
│                              │                                      │
│ Doctor: What brings you in?  │ **CHIEF COMPLAINT**                 │
│ Patient: I've had a cough... │ Persistent cough x 2 weeks          │
│                              │                                      │
│                              │ **HISTORY OF PRESENT ILLNESS**      │
│                              │ 45-year-old presents with...        │
│                              │                                      │
│ [Load Sample]                │              [Generate] [Copy]      │
└──────────────────────────────┴──────────────────────────────────────┘

Requirements

• macOS 13.0+ (Ventura or later)
• Ollama installed and running
• A language model (gpt-oss-20b recommended)

Quick Start

1. Install Ollama

brew install ollama

2. Start Ollama Server

ollama serve

3. Pull a Model

# Recommended: OpenAI's OSS 20B model
ollama pull gpt-oss-20b

# Alternative: Faster, smaller models
ollama pull mistral:7b
ollama pull llama3.2:latest

4. Build and Run the App

# Clone the repo
git clone https://github.com/justlab-ai/oss-20b-med-macos.git
cd oss-20b-med-macos

# Open in Xcode
open Package.swift

# Build and run (Cmd+R in Xcode)

Or build from command line:

swift build
swift run ClinicalScribe

Usage

1. Paste or type a doctor-patient conversation in the left panel
2. Select a model from the dropdown (gpt-oss-20b recommended)
3. Click Generate to create the clinical note
4. Watch the note stream in real-time on the right panel
5. Click Copy to copy the note to clipboard

Models Comparison

Model	Size	Speed	Quality	Best For
gpt-oss-20b	20B (3.6B active)	~28s	Best	Production use
mistral:7b	7B	~5s	Good	Quick drafts
llama3.2:latest	8B	~6s	Good	General use

Tech Stack

• SwiftUI - Native macOS UI
• ollama-swift - Swift client for Ollama API
• AsyncThrowingStream - Real-time streaming responses

Project Structure

oss-20b-med-macos/
├── Package.swift              # Swift package manifest
├── Sources/ClinicalScribe/
│   ├── ClinicalScribeApp.swift  # App entry point
│   ├── ContentView.swift        # Main split-panel UI
│   └── OllamaService.swift      # Ollama API client
└── README.md

Related Projects

• oss-20b-aci-bench - ACI-Bench evaluation of OSS models
• ollama-swift - Swift client library
• Ollama - Local LLM runtime

License

MIT License - See LICENSE for details.

---

Built with ollama-swift by @mattt