Knowledge & the ontology
Why Chiron models an ontology, not an index — and what that buys an agent.
Chiron models an *ontology*, not an index — and the difference is the whole product. This page explains what that means and what it buys an agent.
An index tells an agent which documents mention a thing. An ontology tells it what the thing *is*, what it is made of, what it depends on, and what breaks when it changes. Getting there requires three commitments most retrieval systems never make.
Three commitments
- Entities are typed. A module, an issue, and a table are different kinds of thing, with different questions attached. A search index treats all three as text.
- Relations are typed and directional. "This code writes to that table" is a different fact from "this code reads it." Direction is not decoration — it is what makes blast-radius answerable.
- Meaning is shared across sources. The same concept named in a ticket, a commit message, and a column resolves to *one* entity, not three strings that happen to look alike.
An index makes an agent better at finding text. An ontology makes it capable of reasoning about a system it was never trained on.
One ontology, three planes, linked
Code, work, and data are modeled as separate planes and then connected into a single ontology. The cross-plane edges are the point: anyone can index a repository, but the useful question — *which code did this ticket change*, or *which service breaks if this column is dropped* — is only answerable when the planes share one ontology instead of sitting in three tools.
Deterministic first, models second
Structure is derived by parsers and is reproducible; the same commit yields the same graph. Only then does a model run, over the structure rather than raw files, and its output is labeled as interpretation. Every fact carries its origin — computed or inferred — and every returned claim keeps a resolvable ref. The consumer can verify instead of trust.
How you touch it
You never query the graph directly. You go through one contract — discover, search, fetch, activity — that fans your request across every source and returns one ranked, cited answer.
Next: The Ontology Layer overview.

