Profiles and interoperability

Profiles

The core protocol uses generic entities, locators, selectors, activities, and attributes. A profile adds domain vocabulary without changing the derivation model.

Profiles can define:

• selector types for document regions, database rows, source-code ranges, or media time ranges
• activity-type conventions
• namespaced attributes
• entity and agent conventions

Profile identifiers are strings. Namespaced attributes keep profile data separate from the stable core schema.

For example, the OpenParser extraction profile defines document-region selectors, OCR confidence methods, and openparser:* attributes. Those conventions ship from @openparser/lineage/openparser; the core package remains useful to producers that know nothing about documents or OCR. The profile pages cover verbatim extraction, evidence and confidence, and human review.

Complementary standards

These systems preserve related information at different levels. They can be connected without making one format pretend to replace the others.

lineage@1

Which exact inputs and operation produced this value?

• W3C PROV

  How can provenance be exchanged using shared semantics?

  The semantic foundation for entities, activities, agents, and PROV-JSON export.

• OpenLineage

  Which job transformed which dataset?

  Jobs, runs, datasets, and column-level dependencies.

• OpenTelemetry

  What happened while the operation was running?

  Runtime traces, spans, timing, and operational diagnostics.

• in-toto / C2PA

  Can this artifact’s origin and integrity be verified?

  Signed attestations over artifacts and media.

W3C PROV

lineage@1 adopts the semantics of the W3C PROV Data Model:

• a lineage entity maps to prov:Entity
• an activity maps to prov:Activity
• an agent maps to prov:Agent
• derivations map to wasDerivedFrom, wasGeneratedBy, and used
• associations, attributions, specializations, alternates, and collection membership retain their PROV meaning

toProvJson() exports a validated graph to PROV-JSON.

OpenLineage

OpenLineage describes jobs, runs, datasets, and column dependencies. Activities can carry OpenLineage run and job identifiers, while value entities point to dataset fields.

Derivation inputs use OpenLineage's direct and indirect distinction and can record transformation vocabulary such as identity, aggregation, join, filter, window, or conditional.

OpenTelemetry

OpenTelemetry records runtime execution and timing. Store trace and span identifiers on activities to connect operational telemetry with durable derivation history.

Signed provenance

in-toto and C2PA provide signed attestations for software artifacts and media. Entity digests and lineage document identifiers provide stable subjects for those attestations.

These systems answer complementary questions. lineage@1 focuses on the data derivation itself: which exact inputs and operation produced this value?