Résumé
The Common European Agricultural Data Space (CEADS) aims to enable trusted, interoperable data sharing across the agri-food sector through a federated ecosystem of platforms and services. However, achieving meaningful interoperability requires not only governance and technical alignment but also shared, machineinterpretable semantics that support discovery, integration, and reuse of heterogeneous data assets. This paper presents three preliminary complementary semantic services implemented within the AgrospAI Data Sharing Initiative, built on the Pontus-X compute-to-data infrastructure. First, the AgroAnnotator performs ontologybased annotation of agricultural documents and datasets relying on AgroPortal Annotator within a trusted execution environment, producing reusable semantic outputs. Second, a publication-time connector integrates AgroPortal-hosted controlled vocabularies into the asset (dataset, algorithm, or service) metadata, enabling consistent thematic classification aligned with DCAT specification, while preserving compatibility with Pontus-X metadata models. Third, a wrapper exposes Pontus-X native catalogue metadata records as GeoDCAT-APcompliant JSON-LD. SHACL validation confirms successful end-to-end execution and standards compliance. Collectively, these three early contributions demonstrate a practical pathway for embedding ontology-driven semantics within a federated data space, providing a foundation for future CEADS interoperability and AI-ready data ecosystems.