Offline AI Assistant for Medication Management using IBM Granite
PillboxCheck is a multimodal, offline AI-powered desktop application designed to assist elderly individuals and others managing complex medication schedules. It utilizes IBM's Granite 3.2 large language model and integrates voice, OCR, and scheduling to improve adherence, prevent errors, and support patient autonomy — all while keeping user data entirely on-device.
- Offline AI Chatbot: Powered by Granite 3.2 and a custom RAG pipeline.
- Voice Assistant: "Hello PillboxCheck" to activate and interact naturally.
- OCR Label Scanning: Recognizes prescription data from medication packaging.
- Dashboard: View current inventory, medication history, and scheduled doses.
- Smart Reminders: Alerts users to take their medication and notifies if doses are missed.
- PDF Summaries: Export user health reports for sharing or printing.
- Secure Authentication: Password-protected access with recovery options.
- Multilingual Support (beta): Interact in English and selected other languages.
- Windows 11 (tested)
- .NET 8 SDK
- Visual Studio 2022
- 16 GB RAM recommended for offline AI inference