Trace lineage without turning lineage into ownership

ENTITY records causal relationships among governed objects and can deterministically distribute explicitly modeled value across contribution weights. It does not infer legal title, royalties or payment obligations merely because one object contributed to another.

1. Register source and derived objects

source = fabric.register_object(owner_id, "DATASET", "Source Dataset", content_sha256=SOURCE_SHA)
derived = fabric.register_object(owner_id, "MODEL", "Derived Model", content_sha256=MODEL_SHA)

2. Add a provenance edge

edge = fabric.add_provenance(
    actor_entity_id=owner_id,
    parent_object_id=source["object_id"],
    child_object_id=derived["object_id"],
    relation="TRAINING_SOURCE",
    contribution_bps=6500,
    evidence={"training_run":"run-2026-001"},
)

The graph must remain acyclic, and incoming contribution weights cannot exceed 10,000 basis points. The record explicitly states provenance_is_not_ownership=true.

3. Add additional contributors when justified

fabric.add_provenance(
    owner_id,
    second_source_id,
    derived["object_id"],
    "TRAINING_SOURCE",
    contribution_bps=3500,
    evidence={"training_run":"run-2026-001"},
)

4. Compute a deterministic distribution

distribution = fabric.contribution_distribution(
    derived["object_id"],
    amount_units=100000,
)

This gives a deterministic mathematical allocation from the recorded contribution graph. It is not, by itself, a royalty agreement. An actual payment obligation still needs explicit economic terms, a valid obligation and settlement evidence.

5. Keep value state explicit

ENTITY uses the value states POTENTIAL, OFFER, CONTRACTED, ACCRUED, SETTLED and REALIZED. A system should not jump directly from “this contributed” to “this is realized cash.”