How the data is built.

No inferred or synthetic values — every field traces back to a source record. The pipeline runs the same four steps on every registry, every cycle: Ingest, Normalize, Validate, Version.