Visual Arts Collection Query Agent

A conversational interface for exploring the collection

Ask Questions Naturally

The query agent provides a natural language interface to explore the visual arts collection. Instead of navigating menus or learning a specialized query language, you can simply ask questions in plain English. Ask about paintings by location, explore techniques and materials, discover connections between works, or browse by medium. The agent understands art terminology and can provide context—when you ask about a technique, it draws from both the collection's own definitions and broader art historical knowledge.

Try asking:

show me artworks from Korea what is cold-pressed paper find works that share tags with Seoul show me sculptures show me notes for Circular Courtyard

Connected Results

Results are presented conversationally with clickable links back to the artworks, tags, and concepts in the collection. When you ask about painting notes or related works, the agent not only summarizes the content but also provides direct links so you can explore further. The system handles everything from simple lookups to more sophisticated queries that traverse relationships in the data, such as finding works connected through shared thematic tags or discovering what criticism exists for a particular piece.

How It Works

Behind the scenes, the agent is powered by Claude, Anthropic's AI assistant, which translates your questions into precise database queries against the collection's structured knowledge graph. This semantic web approach means the collection data is richly interconnected—artworks link to techniques, materials, locations, and interpretive notes—and the agent can navigate these connections to surface relevant information. The combination of natural language understanding with structured data enables a conversational exploration experience that would be difficult to achieve with traditional search interfaces.

Query List

Future Capabilities

At this time the agent can only handle text-based analyses since the LLM is not vision capable. In the future, the intention is for the agent to use vision-capable LLMs. At that time the following queries should be supported:

Powered by Claude ยท visualartsdna.org