Digital Humanities projects often use different tools, standards, and data models. Even when they focus on the same historical period or similar sources, their data usually remains separated. This makes cross-project research difficult.

At the Centre for Medieval Studies at KUL, several projects work on related medieval materials but use different formats, including TEI, wiki and structured databases. To make these resources searchable in one place, we are developing Scrinium, a multisearch application for all research data produced within the Centre.

Scrinium is based on the Typesense search engine and supports full-text search and filtering. The main challenge, however, is not the search engine itself, but the integration of different data models. In order to connect heterogeneous project data, we use Wikibase as a central semantic layer. It serves as a shared core that links entities and concepts across projects.

In this paper, I present the architecture of Scrinium, the strategy for integrating different data models, and the practical problems we encountered. I also discuss how this infrastructure can support the development of a RAG-based system for more precise semantic search.

Scrinium shows how a multisearch environment can be built for diverse DH projects while respecting their individual data structures.