A local-first, privacy-focused speech documentation system for multilingual teams. It allows users to speak in their native language, automatically translates to the team's language, tracks tasks from voice reports, and securely attaches photos. Designed for teams where members speak different languages but need to collaborate seamlessly. No uploading of sensitive data to the cloud.

• Audio Recording PWA: Progressive Web App with real-time waveform visualization
• Speech-to-Text: OpenAI Whisper Small (multilingual) for general purpose dictation
• Automatic Classification: AI-based categorization of tasks, notes, and updates (identifying custom categories)
• Multilingual Support: Speak in ANY language, get documentation in your configured language
• GDPR Compliant: Complete on-premise deployment with AES-256 encryption
• Multi-Room Support: Handle multiple treatment rooms per practice
• Export Integration: Zip (CSV + photos) export for creating reports, like billing

┌─────────────────────────────────────────────────────────────┐
│                    Frontend (PWA)                           │
│  React + TypeScript + TailwindCSS + Vite                   │
│  - Audio recording with MediaRecorder API                   │
│  - Real-time waveform visualization                         │
│  - Offline-capable via Service Worker                       │
└─────────────────────────────────────────────────────────────┘
                              │
                              ▼
┌─────────────────────────────────────────────────────────────┐
│                    Backend (API)                            │
│  FastAPI + SQLAlchemy + Whisper                            │
│  - Audio upload and encryption                              │
│  - Batch processing queue                                   │
│  - Transcription with vocabulary boost                      │
│  - Rule-based classification                                │
└─────────────────────────────────────────────────────────────┘
                              │
                              ▼
┌─────────────────────────────────────────────────────────────┐
│                    Storage Layer                            │
│  SQLite + Encrypted File Storage                           │
│  - AES-256-GCM file encryption                             │
│  - 90-day retention policy                                  │
│  - Secure deletion                                          │
└─────────────────────────────────────────────────────────────┘

• Docker and Docker Compose
• Node.js 20+ (for development)
• Python 3.11+ (for development)

1. Clone the repository:

   git clone <repository-url>
   cd VoxDocs

2. Configure environment:

   cp .env.example .env
   # Edit .env with secure keys

3. Start services:

   docker-compose up -d

4. Access the application at http://localhost:3000

1. Start development services:

   docker-compose -f docker-compose.dev.yml up

2. Or run services locally:

   Backend:

   cd backend
   python -m venv venv
   source venv/bin/activate  # or `venv\Scripts\activate` on Windows
   pip install -r requirements.txt
   uvicorn app.main:app --reload

   Frontend:

   cd frontend
   npm install
   npm run dev

VoxDocs/
├── backend/                 # FastAPI Backend
│   ├── app/
│   │   ├── api/            # API endpoints
│   │   ├── core/           # Configuration, database
│   │   ├── models/         # SQLAlchemy models
│   │   ├── services/       # Business logic
│   │   └── main.py         # Application entry
│   └── requirements.txt
├── frontend/                # React PWA
│   ├── src/
│   │   ├── components/     # React components
│   │   ├── hooks/          # Custom hooks
│   │   ├── pages/          # Page components
│   │   ├── services/       # API client
│   │   └── stores/         # State management
│   └── package.json
├── docker/                  # Docker configuration
│   ├── Dockerfile.backend
│   ├── Dockerfile.frontend
│   └── nginx.conf
├── ml/                      # ML training scripts
├── docs/                    # Documentation
└── docker-compose.yml

• POST /api/audio/upload - Upload audio recording
• GET /api/audio/{uuid} - Get recording details
• GET /api/audio/ - List recordings
• DELETE /api/audio/{uuid} - Delete recording

• GET /api/transcription/{uuid} - Get transcription
• POST /api/transcription/{uuid}/process - Process recording
• POST /api/transcription/{uuid}/correct - Submit correction

• GET /api/classification/{uuid} - Get classifications
• POST /api/classification/classify-text - Classify text

• POST /api/export/generate - Generate export file

The system automatically classifies your voice notes into categories. These can be customized, but default to:

• Encryption: AES-256-GCM for all audio files at rest
• On-Premise: No cloud dependencies, complete local deployment
• Data Retention: Automatic deletion after 90 days
• Secure Delete: Multi-pass overwrite for file deletion
• Audit Logging: All data access is logged

Key environment variables:

• Audio processing: 30-60 seconds per minute of audio (CPU)
• Batch processing: Overnight for non-urgent recordings
• Immediate processing: Available for urgent cases

MIT License - see LICENSE for details

For support inquiries, please contact the development team.