Scientific Transparency

A key requirement of the challenge is scientific transparency. Quagga Agent addresses this at multiple levels:

SPARQL Query Exposure

Every answer includes the exact SPARQL query used, with all prefixes, formatting, and whitespace preserved. The validate_sparql_tool extracts the query from the subagent response and presents it alongside the natural-language answer.

Also since we use OpenWebUI, we can export the conversation using a web link for other users of the Quagga Agent or for wider use using a PDF or JSON file export.

Tool Call Visibility

The OpenWebUI interface displays all tool calls and their results, making the agent's reasoning process fully inspectable. Users can see:

• Which KG was selected and why (relevance scores)
• What was the query which led to the answer and metadata like: how many SPARQL queries were attempted?
• The final validated answer and its results (presented as markdown)

Federated Query Potential

The CKG uses NFDI_0001006 (external identifier) to link entities to Wikidata, ROR, GND, and other authority records—over 11 million such links. This shared authority data enables querying other knowledge graphs (Wikidata/GND) to verify the underlying data or use SPARQL SERVICE federated queries across the CKG and external endpoints, which the agent can generate when the user's question requires cross-KG data.

Cross-KG exploration in practice is illustrated in the Bach case study and the Detmolder Hoftheater study.