Conference Agenda

This presentation showcases a next-generation, AI-powered enhancement layer for DSpace repositories, focused on two core areas: intelligent search and improved accessibility. Moving beyond traditional keyword-based discovery, the solution introduces AI-driven retrieval based on Retrieval-Augmented Generation (RAG), combining structured metadata, full-text content, and vector search to deliver accurate, context-aware answers grounded strictly in repository holdings. This approach improves relevance, supports multilingual and cross-language discovery, and scales to very large collections without compromising trustworthiness.

In parallel, the presentation demonstrates AI-powered PDF processing designed to improve accessibility and reuse of repository content. Advanced OCR and document conversion pipelines transform complex or scanned PDFs into fully searchable, screen-reader-friendly, WCAG-aligned formats, significantly broadening access for users with disabilities and enabling downstream AI processing.

Through short live demonstrations, architectural overviews, and real-world use cases, this presentation illustrates how DSpace repositories can evolve from passive storage systems into active, inclusive research discovery platforms – supporting FAIR principles, lowering barriers for new users, and responding pragmatically to the opportunities of emerging AI technologies.

This presentation aims to share the strategies and solutions adopted at 4Science to provide a reliable, high-performance hosting service for DSpace repositories. In the cloud era, repository platforms must support the adoption of cloud-native paradigms to deliver reliable, performant, and cost-effective services.

In 2025, 4Science transitioned its hosting infrastructure from a traditional VM-based architecture to a modern, containerized deployment powered by Kubernetes and AWS cloud-native services. This transition has required changes and fine-tuning to the DSpace application codebase, a review of the development life cycle, and the adoption of new operational tools.

We will explain the reasons behind the changes and the benefits obtained. Based on the operational data, we have identified several areas of improvements across the different application layers, from the frontend to the backend. Lack of support for horizontal scalability has been addressed to provide HA and consistent performance under heavy load. Many improvements have already been contributed to the DSpace codebase directly or via the ongoing merger of DSpace-CRIS; others will be discussed with the community and offered for inclusion in future versions.

This approach keeps DSpace service sustainable, stopping institutions from over-provisioning of costly resources and therefore reducing the environmental impact.