Quagga Agent — From Notes to Nodes

HERMES Data Challenge 2026 · Team MusiGraphia

Challenge	HERMES Data Challenge 2026 — "From Notes to Nodes"
Team	MusiGraphia
Members	Harshdeep Singh, Polina Proutskova, Olga Velichkina, Matteo Romanello
System	Quagga Agent — LLM-powered exploration of the Culture Knowledge Graph (CKG)

Overview

Quagga Agent is a conversational AI that lets musicologists query the Culture Knowledge Graph — and other music-related knowledge graphs — in plain language, without writing SPARQL. Given a research question, the agent identifies the relevant graph, discovers its schema, generates and executes SPARQL queries, validates the results, and returns a structured answer alongside the query used to produce it.

The agent was developed within the EU-funded GRAPHIA project (grant 101188018) and adapted for this challenge to cover CKG datasets including RISM, musiconn.performance, APSearch, and the Detmolder Hoftheater.

Contents

• Technical note — in-depth technical documentation of the agent: architecture, SPARQL pipeline, musicological adequacy, UX, and known limitations
• Explorations — case studies on Bach sources, German venues, Russian operas, Irish traditional music, court theatre repertoire, and more. Team members bring complementary backgrounds spanning AI, musicology, and ethnomusicology; the explorations reflect this diversity, with research questions ranging from technical system probes to substantive musicological inquiries
• Team — MusiGraphia team members

Try the agent

The live system is at quagga-agent.graphia-ssh.eu. Every answer includes the SPARQL query used; conversations can be shared via link or exported as PDF/JSON.