Knowledge graph
Concept hierarchies, taxonomies, and relationship networks.
Taxonomy
Strict tree: every node has exactly one parent. Good for file systems, category browsing, biological classification.
doc { cols: 3, rows: 3 }
icon :anim @B1 tabler/paw "Animal" sizeScale=1.3
icon :mam @A2 tabler/paw "Mammal"
icon :bir @C2 tabler/feather "Bird"
icon :cat @A3 tabler/cat "Cat"
icon :dog @B3 tabler/dog "Dog"
icon :eag @C3 tabler/feather "Eagle"
anim --> mam
anim --> bir
mam --> cat
mam --> dog
bir --> eagFaceted tags
Items categorised along multiple independent axes. A "cross-cutting" model — nothing is in a strict hierarchy anymore.
doc { cols: 4, rows: 3 }
icon :item @B2 tabler/box "Item"
icon :col @A1 tabler/palette "Color"
icon :siz @C1 tabler/ruler "Size"
icon :use @D1 tabler/tag "Use"
icon :mat @A3 tabler/stack "Material"
item <-> col
item <-> siz
item <-> use
item <-> matConcept relations
Nouns connected by named edges — the skeleton of a small ontology. RDF / Wikidata graphs look like this with more vocabulary.
doc { cols: 4, rows: 3 }
icon :author @A2 tabler/user "Author"
icon :book @B2 tabler/book "Book"
icon :topic @D1 tabler/tag "Topic"
icon :pub @D2 tabler/building "Publisher"
icon :year @D3 tabler/calendar "Year"
author --> book "wrote"
book --> topic "about"
book --> pub "by"
book --> year "in"Knowledge graph for an agent
What an LLM "sees" when you hand it a knowledge graph for retrieval. Each edge is a typed relation the agent can traverse.
doc { cols: 4, rows: 3 }
icon :q @A2 tabler/search "Query"
icon :ent @B2 tabler/box "Entity"
icon :rel1 @C1 tabler/link "relatedTo"
icon :rel2 @C3 tabler/link "partOf"
icon :res @D2 tabler/bulb "Answer"
q --> ent
ent --> rel1
ent --> rel2
rel1 --> res
rel2 --> res