Using the Grounded Assistant
How AsteriaStar's grounded assistant answers from the knowledge graph alone — explaining entities, comparing concepts, tracing evidence paths, and building learning paths — and how it stays honest: it returns only real facts and citations, never generates prose, and says 'not enough graph evidence' rather than guess.
Related entities
- Assistant capability
Object Explanation
A grounded explanation of any entity — what it is, how it connects to the rest of the graph, and where the knowledge comes from — assembled entirely from the entity's own description, its relations, and its cited sources.
- Assistant capability
Concept Comparison
Compares two concepts by the real common ground between them — the entities they both connect to in the graph. Mars and Venus, for instance, share atmospheric escape, climate evolution, and their place in the Solar System. A comparison built from relations, not rhetoric.
- Assistant capability
Evidence Chains
Traces the chain of real relations connecting a claim to its supporting entities — for example, from Edwin Hubble through the expansion of the universe to dark energy. Every link in the chain is a genuine edge in the graph, so the reasoning can be followed and checked.
- Assistant capability
The No-Hallucination Layer
The design principle beneath the assistant: it surfaces only facts already in the knowledge graph, each with its provenance and a traceable chain of relations. A future language model would phrase these grounded facts, never add to them — so an answer can always be checked against the graph.