This paper presents SAPSAI (sap-sigh) – System for Acquisition, Preprocessing, and Storage of Agricultural Images, a modular and scalable system for the acquisition, preprocessing, and structured storage of agricultural images aimed at enhancing early crop disease diagnosis. The proposed solution integrates a web application that allows non-expert users to directly capture leaf images and a microservice powered by a Vision Transformer (ViT) model responsible for validating and classifying leaves based on their morphology. We trained the ViT model from scratch on a reorganized dataset grouped by leaf type. We achieved an average accuracy of 73.64% on the test set. The results show high performance for well-represented classes, highlighting challenges in distinguishing morphologically similar categories. The system enables the collaborative construction of high-quality datasets, supporting the subsequent training of disease classification models. This approach promotes the adoption of intelligent technologies in precision agriculture, contributing to improved efficiency and sustainability in crop production.