Odoma
About the organisation
Odoma is a Swiss-based SME that combines AI and UX design to serve the cultural and creative sectors. It works with museums, libraries, archives, and research groups, building bespoke AI workflows that parse images, recognise handwritten and multilingual texts, and structure data from historical documents, while designing user-centred interfaces that turn complex collections into searchable knowledge resources.
What challenges can the GRAPHIA project help solve?
As a commercial partner in the GRAPHIA project, Odoma contributes its AI expertise to the creation of a comprehensive Knowledge Graph for SSH enriched with AI and LLM applications, working alongside research infrastructures and industry partners to make SSH data more machine-readable, interoperable, and useful to both academic and commercial users.
What is the proposed use case?
GraphChat is a conversational interface that lets users query knowledge graphs in plain language, without SPARQL expertise. It translates a question into a SPARQL query, executes it against the user’s own knowledge graph endpoint, passes the result bindings to an LLM, and returns a natural-language answer together with the query, the bindings, and a timestamp.
Every response is fully traceable: the source endpoint, the generated query, and the underlying triples are visible, so claims can be reproduced, cited, and audited. When data is missing, GraphChat reports the gap rather than inventing a plausible answer, eliminating hallucinations by design. Data never leaves the institution — queries are executed against the existing endpoint, preserving governance and security.
Results can be rendered as tables, charts, or lists, and the agent can be integrated into other systems through an API or MCP. GraphChat is model-agnostic, working with major commercial and open-source LLMs.
GraphChat currently serves two complementary audiences — domain experts and managers who want trustworthy answers in plain language, and data specialists who can inspect and refine the generated queries — with use cases across museums, libraries, archives, and auction houses.
What type of SSH data or content are involved?
GraphChat currently supports the exploration of five knowledge graphs from the cultural heritage sector (museums, libraries and archives), including the EHRI KG developed by GRAPHIA partner EHRI (European Holocaust Research Infrastructure).
What technical aspect would be used?
GraphChat is powered by LLMs served and deployed in GRAPHIA as part of the LLM4SSH service offer. GraphChat is being developed by Odoma as a fork of GRAPHIA’s Quagga agent, an agentic LLM that enables conversational exploration of any semantic knowledge graph. Moreover, GraphChat builds upon EXYGEN (EXplore Your Graphs ENgine), an API developed by Odoma in the context of GRAPHIA for the automatic generation of KG metadata at scale.