User Experience

Conversational Interface

Quagga Agent provides a conversational interface via OpenWebUI, a widely-used open-source chat UI. The agent appears as a model named "Graphia SPARQL Agent" in the model selector. Users interact entirely in natural language—no knowledge of SPARQL, the CTO ontology, or the CKG endpoint is required.

Streaming and Progress Indicators

The agent streams responses in real-time using a custom SSE protocol with multiple chunk types:

Chunk Type	Purpose
progress	Status updates ("Selecting knowledge graph(s)", "Retrieving metadata", "Generating SPARQL query")
tool_call / tool_result	Tool execution transparency
llm_text	Final answer text
reasoning_content	Renders as "Thinking..." in OpenWebUI for progress steps

The QuaggaProgressPipe OpenWebUI filter (filters/progress_pipe.py) translates progress chunks into native OpenWebUI status indicators, giving users real-time feedback during the 10–60 second query generation process.

Multi-Turn Conversations

The agent maintains conversation state via PostgreSQL-backed checkpointing (AsyncPostgresSaver). Follow-up questions inherit context from previous turns—the subagent receives formatted parent context including prior questions, answers, and SPARQL queries. This enables natural research workflows like:

User: Are there any Russian operas in the graph? Which ones? Agent: [lists operas] User: Romeo and Julia by Tchaikovsky? What's that? 1812 Ouvertuere - from which work is it? Agent: [refines query based on previous context]

A full multi-turn example is available in this Quagga conversation; see also Russian operas and ballets for a critical assessment of similar queries.