Remove schema:subjectOf edges from KG extraction (#695)

The subjectOf triples were redundant with the subgraph provenance model
introduced in e8407b34. Entity-to-source lineage can be traced via
tg:contains -> subgraph -> prov:wasDerivedFrom -> chunk, making the
direct subjectOf edges unnecessary metadata polluting the knowledge graph.

Removed from all three extractors (agent, definitions, relationships),
cleaned up the SUBJECT_OF constant and vocabulary label, and updated
tests accordingly.

tests/integration/test_agent_kg_extraction_integration.py

@@ -14,7 +14,7 @@ from unittest.mock import AsyncMock, MagicMock, patch
 from trustgraph.extract.kg.agent.extract import Processor as AgentKgExtractor
 from trustgraph.schema import Chunk, Triple, Triples, Metadata, Term, Error, IRI, LITERAL
 from trustgraph.schema import EntityContext, EntityContexts, AgentRequest, AgentResponse
-from trustgraph.rdf import TRUSTGRAPH_ENTITIES, DEFINITION, RDF_LABEL, SUBJECT_OF
+from trustgraph.rdf import TRUSTGRAPH_ENTITIES, DEFINITION, RDF_LABEL
 from trustgraph.template.prompt_manager import PromptManager
 
 
@@ -174,10 +174,6 @@ class TestAgentKgExtractionIntegration:
         label_triples = [t for t in sent_triples.triples if t.p.iri == RDF_LABEL]
         assert len(label_triples) >= 2  # Should have labels for entities
 
-        # Check subject-of relationships
-        subject_of_triples = [t for t in sent_triples.triples if t.p.iri == SUBJECT_OF]
-        assert len(subject_of_triples) >= 2  # Entities should be linked to document
-
         # Verify entity contexts were emitted
         entity_contexts_publisher = mock_flow_context("entity-contexts")
         entity_contexts_publisher.send.assert_called_once()

tests/integration/test_kg_extract_store_integration.py

@@ -17,7 +17,7 @@ from trustgraph.extract.kg.relationships.extract import Processor as Relationshi
 from trustgraph.storage.knowledge.store import Processor as KnowledgeStoreProcessor
 from trustgraph.schema import Chunk, Triple, Triples, Metadata, Term, Error, IRI, LITERAL
 from trustgraph.schema import EntityContext, EntityContexts, GraphEmbeddings, EntityEmbeddings
-from trustgraph.rdf import TRUSTGRAPH_ENTITIES, DEFINITION, RDF_LABEL, SUBJECT_OF
+from trustgraph.rdf import TRUSTGRAPH_ENTITIES, DEFINITION, RDF_LABEL
 
 
 @pytest.mark.integration

tests/unit/test_knowledge_graph/test_agent_extraction.py

@@ -13,7 +13,7 @@ from unittest.mock import AsyncMock, MagicMock, patch
 from trustgraph.extract.kg.agent.extract import Processor as AgentKgExtractor
 from trustgraph.schema import Chunk, Triple, Triples, Metadata, Term, Error, IRI, LITERAL
 from trustgraph.schema import EntityContext, EntityContexts
-from trustgraph.rdf import TRUSTGRAPH_ENTITIES, DEFINITION, RDF_LABEL, SUBJECT_OF
+from trustgraph.rdf import TRUSTGRAPH_ENTITIES, DEFINITION, RDF_LABEL
 from trustgraph.template.prompt_manager import PromptManager
 
 
@@ -183,12 +183,6 @@ This is not JSON at all
         assert def_triple.s.iri == f"{TRUSTGRAPH_ENTITIES}Machine%20Learning"
         assert def_triple.o.value == "A subset of AI that enables learning from data."
 
-        # Check subject-of triple
-        subject_of_triple = next((t for t in triples if t.p.iri == SUBJECT_OF), None)
-        assert subject_of_triple is not None
-        assert subject_of_triple.s.iri == f"{TRUSTGRAPH_ENTITIES}Machine%20Learning"
-        assert subject_of_triple.o.iri == "doc123"
-
         # Check entity context
         assert len(entity_contexts) == 1
         assert entity_contexts[0].entity.iri == f"{TRUSTGRAPH_ENTITIES}Machine%20Learning"
@@ -228,10 +222,6 @@ This is not JSON at all
         assert rel_triple.o.iri == object_uri
         assert rel_triple.o.type == IRI
 
-        # Check subject-of relationships
-        subject_of_triples = [t for t in triples if t.p.iri == SUBJECT_OF and t.o.iri == "doc123"]
-        assert len(subject_of_triples) >= 2  # At least subject and predicate should have subject-of relations
-
     def test_process_extraction_data_literal_object(self, agent_extractor, sample_metadata):
         """Test processing of relationships with literal objects"""
         data = [
@@ -274,10 +264,6 @@ This is not JSON at all
 
         triples, entity_contexts, _ = agent_extractor.process_extraction_data(data, metadata)
 
-        # Should not create subject-of relationships when no metadata ID
-        subject_of_triples = [t for t in triples if t.p.iri == SUBJECT_OF]
-        assert len(subject_of_triples) == 0
-
         # Should still create entity contexts
         assert len(entity_contexts) == 1
 

tests/unit/test_knowledge_graph/test_agent_extraction_edge_cases.py

@@ -13,7 +13,7 @@ from unittest.mock import AsyncMock, MagicMock
 from trustgraph.extract.kg.agent.extract import Processor as AgentKgExtractor
 from trustgraph.schema import Chunk, Triple, Triples, Metadata, Term, IRI, LITERAL
 from trustgraph.schema import EntityContext, EntityContexts
-from trustgraph.rdf import TRUSTGRAPH_ENTITIES, DEFINITION, RDF_LABEL, SUBJECT_OF
+from trustgraph.rdf import TRUSTGRAPH_ENTITIES, DEFINITION, RDF_LABEL
 
 
 @pytest.mark.unit
@@ -187,10 +187,6 @@ class TestAgentKgExtractionEdgeCases:
         data = [{"type": "definition", "entity": "Test", "definition": "Test def"}]
         triples, contexts, _ = agent_extractor.process_extraction_data(data, metadata)
 
-        # Should not create subject-of triples when ID is empty string
-        subject_of_triples = [t for t in triples if t.p.iri == SUBJECT_OF]
-        assert len(subject_of_triples) == 0
-
     def test_process_extraction_data_special_entity_names(self, agent_extractor):
         """Test processing with special characters in entity names"""
         metadata = Metadata(id="doc123")
@@ -338,7 +334,7 @@ class TestAgentKgExtractionEdgeCases:
 
         # Should process all relationships
         # Note: The current implementation has some logic issues that these tests document
-        assert len([t for t in triples if t.p.iri != RDF_LABEL and t.p.iri != SUBJECT_OF]) >= 7
+        assert len([t for t in triples if t.p.iri != RDF_LABEL]) >= 7
 
     @pytest.mark.asyncio
     async def test_emit_empty_collections(self, agent_extractor):