Smart Cellar a mobile app for managing domestic cellars

Abstract

This thesis presents the design and development of Smart Cellar, an intelligent and offline- capable application for efficient wine cellar management. The project aims to minimize user effort in wine insertion and retrieval by combining multimodal interaction through images, text, and voice with AI-enhanced automation. The system requirements were defined using the Goal-Oriented Requirements (GoReq) method, which derives them from stakeholder goals and represents them with UML models of the domain and interactions. The architecture follows a client–server model, consisting of a Flutter-based mobile client and a Python Flask backend integrated with PostgreSQL (pgvector), Firebase Authentication, and other cloud services. Optical Character Recognition (OCR), image embeddings, and OpenAI GPT models are integrated to enable automated wine label recognition, cross-modal search using image and text embeddings, and intelligent recommendations. Wine label images are stored in Google Cloud Storage. The offline-first design ensures full usability without internet connectivity through local caching and automatic synchronization mechanisms once online. Comprehensive unit and integration testing validated the reliability of core components, AI services, and synchronization workflows. The results confirm that the Smart Cellar system achieves its goal of providing a robust, intelligent, and user-friendly solution that reduces manual effort while supporting seamless management of digital wine cellars in both online and offline modes.

This thesis presents the design and development of Smart Cellar, an intelligent and offline- capable application for efficient wine cellar management. The project aims to minimize user effort in wine insertion and retrieval by combining multimodal interaction through images, text, and voice with AI-enhanced automation. The system requirements were defined using the Goal-Oriented Requirements (GoReq) method, which derives them from stakeholder goals and represents them with UML models of the domain and interactions. The architecture follows a client–server model, consisting of a Flutter-based mobile client and a Python Flask backend integrated with PostgreSQL (pgvector), Firebase Authentication, and other cloud services. Optical Character Recognition (OCR), image embeddings, and OpenAI GPT models are integrated to enable automated wine label recognition, cross-modal search using image and text embeddings, and intelligent recommendations. Wine label images are stored in Google Cloud Storage. The offline-first design ensures full usability without internet connectivity through local caching and automatic synchronization mechanisms once online. Comprehensive unit and integration testing validated the reliability of core components, AI services, and synchronization workflows. The results confirm that the Smart Cellar system achieves its goal of providing a robust, intelligent, and user-friendly solution that reduces manual effort while supporting seamless management of digital wine cellars in both online and offline modes.