Research data management infrastructures for agricultural sciences must handle heterogeneous datasets spanning geospatial layers, tabular records, and model outputs while satisfying the FAIR principles and supporting interactive spatial exploration. Existing research repositories provide robust curation and persistent identifier workflows but lack geospatial visualisation capabilities to address direct interaction of researches and research data on first met. This work presents the extension of GeoNode, an open-source geospatial content management system, into a FAIR-compliant research data repository deployed at the Leibniz Centre for Agricultural Landscape Research (ZALF) to operate the BonaRes Repository. The platform introduces a multi-step publication workflow with DataCite DOI registration, an extended metadata model, non-spatial dataset support. To limit long-term maintenance costs, ingestion, automation, and bibliographic presentation are implemented as independent API-driven components. The full stack is deployed on Kubernetes providing reproducible and horizontally scalable operation. The platform currently hosts about one thousand DOI-registered datasets and is integrated into FairAgro, the German national research data infrastructure for agricultural sciences. All components are released as free and open-source software, providing a transferable reference architecture for institutions seeking to build domain-specific research data infrastructures on open geospatial technology.