Querygraph is built around a simple but demanding product promise: an
agent should be able to answer a serious question over enterprise data
without turning the enterprise into an unbounded prompt. The system must
know what the data is, where it came from, who is asking, which action
is allowed, which model was used, what the answer depends on, and how to
replay the run.
That promise becomes the spine of the platform:
Semantic Croissant describes the data as agents will encounter
it.
CDIF publishes the same data as interoperable FAIR metadata.
DIDs identify agents, datasets, bundles, issuers, and
attestations.
ODRL expresses governed data rights as machine-actionable
policies.
TypeSec turns those identities and policies into typed
capabilities.
Grust gives the navigator a graph of meaning, policy, lineage, and
agents.
OSI gives business concepts stable names and relationships.
Sail executes the lakehouse and keeps the audit data queryable.
OpenLineage records runs, inputs, outputs, and derivations.
QG Lakehouse ties all of it together in Rust.
Diagram 4
The rest of the book walks that spine component by component. Each
chapter answers three questions: why the component exists, how
Querygraph uses it, and what becomes possible when it is combined with
the others.
