Clinical Decision Support
Clinical decision support powered by AI
Cedar is a clinical support application that augments medical practitioners with AI-powered document analysis, second opinions on lab reports, imaging studies, and healthcare documents. Built with Electron, React, and Python, Cedar provides intelligent document processing, semantic search, and contextual AI assistance for healthcare professionals.
| Feature | Description |
|---|---|
| 📄 Multi-Format Documents | PDF, DOCX, XLSX, PPTX, images, and more |
| 🖼️ Medical Image Analysis | Multi-angle analysis using MedGemma |
| 🔍 Semantic Search | Vector-based similarity search across documents |
| 💬 AI-Powered Chat | Contextual conversations with RAG pipeline |
| 🎤 Voice Input | OpenAI gpt-4o-transcribe for hands-free operation |
| 📎 Attachments | Upload supplementary files during chat |
| 📋 Summarization | AI-generated document summaries |
| 📝 Notes | Professional notes linked to documents |
- React 18 + TypeScript
- Tailwind CSS + shadcn/ui
- Zustand for state management
- Vite build tool
- Electron desktop runtime
- SQLite with better-sqlite3
- hnswlib-node for vector similarity search
- gpt-4o-transcribe for speech-to-text
- FastAPI with Granian ASGI server
- MarkItDown for document extraction
- OpenAI SDK for embeddings and transcription
- Google Vertex AI – MedGemma-1-27B IT for chat completions and image analysis
- OpenAI API – text-embedding-3-small for semantic embeddings, gpt-4o-transcribe for speech-to-text
┌─────────────────────────────────────────────────────────────────┐
│ CEDAR APPLICATION │
├─────────────────────────────────────────────────────────────────┤
│ │
│ ┌──────────────────┐ ┌──────────────────────────────────┐ │
│ │ ELECTRON UI │ │ PYTHON API │ │
│ │ (Renderer) │ │ (FastAPI + Granian) │ │
│ │ │ │ │ │
│ │ React + │◄──►│ - Chat Completions │ │
│ │ TypeScript │ │ - Image Analysis (MedGemma) │ │
│ │ Tailwind CSS │ │ - Document Extraction │ │
│ │ Zustand Store │ │ - Embeddings (OpenAI) │ │
│ │ Lucide Icons │ │ - Speech Transcription │ │
│ └──────────────────┘ │ - Summarization │ │
│ │ └──────────────────────────────────┘ │
│ │ │ │
│ ▼ ▼ │
│ ┌──────────────────────────────────────────────────────────┐ │
│ │ MAIN PROCESS (Electron) │ │
│ │ │ │
│ │ - IPC Bridge (preload.js) │ │
│ │ - File System Access │ │
│ │ - SQLite Database (metadata, FTS5) │ │
│ │ - HNSW Vector Store (embeddings) │ │
│ │ - Processing Queue │ │
│ └──────────────────────────────────────────────────────────┘ │
│ │ │
│ ▼ │
│ ┌──────────────────────────────────────────────────────────┐ │
│ │ CLOUD SERVICES │ │
│ │ │ │
│ │ Google Vertex AI (MedGemma-1-27B IT) │ │
│ │ ├── Chat completions with streaming │ │
│ │ └── Multi-angle image analysis │ │
│ │ │ │
│ │ OpenAI API (text-embedding-3-small, gpt-4o-transcribe) │ │
│ │ └── 768-dimension semantic embeddings, speech-to-text │ │
│ └──────────────────────────────────────────────────────────┘ │
│ │
└─────────────────────────────────────────────────────────────────┘
- Node.js 20+ and npm or pnpm
- Python 3.12+
- Google Cloud credentials for Vertex AI access
- OpenAI API key for embeddings and transcription
# Clone the repository
git clone https://github.com/nwaughachukwuma/cedar.git
cd cedar
# Install dependencies
pnpm install
# Install Python dependencies
cd api && uv pip install -e . && cd ..
# Configure environment variables
cp .env.example .env
# Edit .env with your API keys
# Start development servers
npm run dev# Build the application
npm run build
npm run packagecedar/
├── api/ # Python FastAPI backend
│ ├── src/
│ │ ├── main.py # FastAPI application entry
│ │ ├── _env.py # Environment configuration
│ │ ├── constants.py # API constants and endpoints
│ │ ├── logger.py # Logging utilities
│ │ ├── chat_agent/ # Chat completions
│ │ │ ├── base.py
│ │ │ ├── no_streaming.py
│ │ │ ├── use_openai.py
│ │ │ └── with_streaming.py
│ │ ├── embeddings/ # Text embeddings
│ │ │ ├── base.py
│ │ │ ├── nvidia_nim.py
│ │ │ └── utils.py
│ │ ├── image_agent/ # Medical image analysis
│ │ │ ├── base.py
│ │ │ ├── image_analysis.py
│ │ │ ├── use_openai.py
│ │ │ ├── utils.py
│ │ │ └── nvidia_nim/
│ │ ├── speech_transcription/ # Speech-to-text
│ │ │ ├── base.py
│ │ │ ├── nvidia_nim.py
│ │ │ └── use_openai.py
│ │ ├── summarization/ # Text summarization
│ │ │ ├── base.py
│ │ │ ├── use_openai.py
│ │ │ └── utils.py
│ │ └── utils/ # Shared utilities
│ │ ├── medgemma_response.py
│ │ ├── nim_sdk.py
│ │ ├── openai_sdk.py
│ │ ├── parse_document.py
│ │ └── vertext_auth.py
│ ├── pyproject.toml
│ └── install.sh
│
├── electron/ # Electron main process
│ ├── main.ts # Application entry point
│ ├── preload.ts # IPC context bridge
│ ├── defineEnv.ts # Environment type definitions
│ ├── ipc/
│ │ ├── index.ts # IPC handler registration
│ │ ├── chatHandlers.ts # Chat streaming handlers
│ │ ├── fileHandlers.ts # File upload/management
│ │ ├── noteHandlers.ts # Notes CRUD
│ │ ├── searchHandlers.ts # Search handlers
│ │ ├── debugHandler.ts # Debug utilities
│ │ └── helpers/ # IPC helper modules
│ ├── services/
│ │ ├── database.ts # SQLite database operations
│ │ ├── vectorStore.ts # HNSW vector storage
│ │ ├── modelService.ts # Model orchestration
│ │ ├── searchService.ts # Search coordination
│ │ ├── speechToText.ts # OpenAI gpt-4o-transcribe
│ │ ├── documentProcessor.ts # Document processing
│ │ └── imageProcessor.ts # Image processing
│ ├── workers/
│ │ ├── workerManager.ts # Background worker management
│ │ └── worker.js # Worker script
│ └── utils/
│ ├── fileProcessingQueue.ts # Async processing queue
│ ├── runMigrations.ts # Database migrations
│ └── ...
│
├── src/ # React frontend
│ ├── App.tsx # Main application component
│ ├── main.tsx # React entry point
│ ├── components/ # React components
│ │ ├── FileList.tsx # Document browser
│ │ ├── FileDetailView.tsx # Document viewer + chat
│ │ ├── FilePreview.tsx # Document viewer
│ │ ├── ChatPanel.tsx # Chat interface
│ │ ├── ChatInput.tsx # Message input component
│ │ ├── SearchBar.tsx # Semantic search bar
│ │ └── ...
│ ├── stores/ # Zustand state stores
│ │ ├── fileStore.ts
│ │ ├── filesStore.ts
│ │ ├── chatStore.ts
│ │ ├── messageStore.ts
│ │ └── ...
│ ├── lib/ # Utilities and hooks
│ └── types/ # TypeScript type definitions
│
├── migrations/ # SQLite migrations
├── tests/ # Test suite
├── docs/ # Documentation
├── dist-electron/ # Built Electron application
├── package.json
└── README.md
# API Configuration
OPENAI_API_KEY=sk-...
GRANIAN_WORKERS=4
APP_ENV=dev
# Google Cloud (Vertex AI)
# Ensure gcloud CLI is authenticated| Table | Description |
|---|---|
files |
Uploaded documents and images |
chunks |
Document content segments for RAG |
embeddings |
Vector embedding metadata |
annotations |
Image analysis results |
chat_sessions |
Chat conversation sessions |
chat_messages |
Individual chat messages |
notes |
Professional notes |
processing_queue |
Async job tracking |
See migrations/ for complete schema.
| Endpoint | Method | Description |
|---|---|---|
/health-check |
GET | Health check |
/analyze-image |
POST | MedGemma image analysis |
/extract |
POST | Document content extraction |
/chat |
POST | Chat completions |
/get-embeddings |
POST | Generate text embeddings |
/get-transcription |
POST | Speech-to-text transcription |
/summarize-text |
POST | Text summarization |
# Run development server
pnpm dev
# Run tests
pnpm test
# Lint and format
pnpm lint
pnpm format
# Type check
pnpm typecheck- Fork the repository
- Create a feature branch (
git checkout -b feature/amazing-feature) - Commit your changes (
git commit -m 'Add amazing feature') - Push to the branch (
git push origin feature/amazing-feature) - Open a Pull Request
This project is licensed under the MIT License – see the LICENSE file for details.
Cedar is a clinical decision support tool. It is not intended to replace professional medical judgment or diagnosis. Always verify AI-generated insights with qualified healthcare providers.
Cedar – Empowering healthcare professionals with AI