MedVoice

🧠 Medical Speech Transcription System (Bachelor thesis)

A privacy-first, offline-ready system for medical professionals that transcribes and summarizes patient consultations using local AI models. Designed to reduce documentation time and improve accuracy in medical records.

⚙️ System Overview

This system consists of a client-server architecture, where:

• Backend (Python + FastAPI) serves only as a bridge to the local database (PostgreSQL).
• Frontend (NuxtJS + TailwindCSS + Tauri) handles the voice capture, ASR transcription (Whisper-rs), and LLM summarization (Ollama-rs) entirely on the client side.

🧩 Key Features

• 🎤 Real-time audio transcription with Whisper-rs
• ✍️ LLM-based summarization of transcripts via Ollama-rs
• 🧩 Modular architecture using Service–Repository pattern
• 🗄️ Secure local database for appointment records
• 📦 Cross-platform GUI app with Tauri and NuxtJS
• 🧪 Integrated unit and integration tests with Pytest
• 🔁 CI/CD via GitHub Actions

📦 Technologies Used

📝 License

This project is licensed under the MIT License.