참고소스 수정본

tests/test_generic_ontology_core.py

@@ -0,0 +1,208 @@
+from sqlalchemy.orm import Session
+
+from crawler_platform.app.config.loader import ProjectConfig, SourceConfig, load_project_config
+from crawler_platform.app.core.database import models
+from crawler_platform.app.core.database.repository import KnowledgeRepository
+from crawler_platform.app.core.database.session import make_engine
+from crawler_platform.app.core.extractor.base import ExtractedClaim, ExtractedEntity, ExtractionBundle, ExtractionPageContext
+from crawler_platform.app.core.extractor.validation import attach_page_context
+from crawler_platform.app.core.ontology.gap_detector import KnowledgeGapDetector
+from crawler_platform.app.core.ontology.registry import OntologyRegistry
+from crawler_platform.app.core.research.graph_query import SemanticGraphQuery
+
+
+def make_repo():
+    config = load_project_config("configs/perfume_subscription.yaml")
+    engine = make_engine("sqlite:///:memory:")
+    models.Base.metadata.create_all(engine)
+    session = Session(engine)
+    repo = KnowledgeRepository(session)
+    project = repo.upsert_project(config)
+    source = repo.get_source(project.id, "official_brand_site")
+    return config, session, repo, project, source
+
+
+def save_validated_note_claim(repo, project, source, config):
+    page = repo.upsert_page(
+        project.id,
+        source.id,
+        "https://example.com/product/cotton",
+        "Cotton Example",
+        200,
+        "Cotton Example\nTop notes: Bergamot",
+    )
+    bundle = ExtractionBundle(
+        entities=[
+            ExtractedEntity("Perfume", "Cotton Example", evidence_text="Cotton Example", confidence=0.95),
+            ExtractedEntity("Note", "Bergamot", evidence_text="Top notes: Bergamot", confidence=0.95),
+        ],
+        claims=[
+            ExtractedClaim(
+                "Cotton Example",
+                "Perfume",
+                "hasTopNote",
+                "Bergamot",
+                "Note",
+                evidence_text="Top notes: Bergamot",
+                confidence=0.96,
+                confidence_reason="direct evidence",
+            )
+        ],
+        extractor_name="llm_json_extractor",
+        provider="lm_studio",
+        raw_output={"extraction_mode": "primary"},
+    )
+    context = ExtractionPageContext(
+        url=page.url,
+        final_url=page.url,
+        title=page.title,
+        page_type="ProductPage",
+        clean_text=page.cleaned_text_summary,
+        source_zones=[
+            {
+                "zone_type": "product_description",
+                "selector": "main",
+                "text": "Top notes: Bergamot",
+                "claim_allowed": True,
+            }
+        ],
+    )
+    return repo.save_extraction_bundle(project.id, source, page, attach_page_context(bundle, context), config)
+
+
+def test_project_upsert_seeds_schema_registry():
+    _config, session, _repo, project, _source = make_repo()
+
+    registry = OntologyRegistry(session).registry_payload(project.id)
+
+    entity_names = {row["name"] for row in registry["entity_types"]}
+    relation_names = {row["name"] for row in registry["relation_types"]}
+    assert {"Entity", "Product", "Perfume"}.issubset(entity_names)
+    assert {"relatedTo", "hasTopNote", "hasBrand"}.issubset(relation_names)
+    session.close()
+
+
+def test_validated_claim_creates_separate_ontology_triple():
+    config, session, repo, project, source = make_repo()
+
+    claims = save_validated_note_claim(repo, project, source, config)
+
+    triples = session.query(models.OntologyTriple).all()
+    assert len(claims) == 1
+    assert len(triples) == 1
+    assert triples[0].predicate == "hasTopNote"
+    assert triples[0].status == "merged"
+    assert claims[0].metadata_json["ontology_triple_id"] == triples[0].id
+    session.close()
+
+
+def test_schema_proposal_and_knowledge_gap_are_governance_records():
+    _config, session, _repo, project, _source = make_repo()
+    registry = OntologyRegistry(session)
+
+    proposal = registry.propose_schema_change(
+        project.id,
+        proposal_type="relation_type",
+        name="influencedBy",
+        reason="New cross-domain relation observed in evidence.",
+        evidence="A research paper influenced a product strategy.",
+        confidence=0.73,
+    )
+    gaps = KnowledgeGapDetector(session).list_open(project.id)
+
+    assert proposal.status == "pending_review"
+    assert any(gap["gap_type"] == "schema_governance" and gap["target_name"] == "influencedBy" for gap in gaps)
+    session.close()
+
+
+def test_generic_graph_query_uses_ontology_triples():
+    config, session, repo, project, source = make_repo()
+    save_validated_note_claim(repo, project, source, config)
+
+    rows = SemanticGraphQuery(session).relation_summary(project.id)
+
+    assert rows[0]["predicate"] == "hasTopNote"
+    assert rows[0]["support_count"] == 1
+    session.close()
+
+
+def test_config_driven_academic_relation_without_perfume_adapter():
+    config = ProjectConfig(
+        project_name="academic_demo",
+        domain="academic",
+        target_entities=["ResearchPaper", "Person"],
+        fields=["title", "author"],
+        sources=[SourceConfig(name="research_site", type="paper_index", trust_level=0.9)],
+        ontology={
+            "entity_types": ["ResearchPaper", "Person"],
+            "predicates": ["authoredBy"],
+            "relation_types": {
+                "authoredBy": {
+                    "allowed_subject_types": ["ResearchPaper"],
+                    "allowed_object_types": ["Person"],
+                    "allowed_page_types": ["ArticlePage"],
+                    "allowed_source_zones": ["article_body"],
+                    "confidence_rules": {"min_confidence": 0.82},
+                }
+            },
+        },
+    )
+    engine = make_engine("sqlite:///:memory:")
+    models.Base.metadata.create_all(engine)
+    session = Session(engine)
+    repo = KnowledgeRepository(session)
+    project = repo.upsert_project(config)
+    source = repo.get_source(project.id, "research_site")
+    page = repo.upsert_page(
+        project.id,
+        source.id,
+        "https://example.org/papers/semantic-systems",
+        "Semantic Systems",
+        200,
+        "Semantic Systems was authored by Ada Kim.",
+    )
+    bundle = ExtractionBundle(
+        entities=[
+            ExtractedEntity("ResearchPaper", "Semantic Systems", evidence_text="Semantic Systems", confidence=0.95),
+            ExtractedEntity("Person", "Ada Kim", evidence_text="authored by Ada Kim", confidence=0.95),
+        ],
+        claims=[
+            ExtractedClaim(
+                "Semantic Systems",
+                "ResearchPaper",
+                "authoredBy",
+                "Ada Kim",
+                "Person",
+                evidence_text="Semantic Systems was authored by Ada Kim.",
+                confidence=0.97,
+                confidence_reason="directly stated",
+            )
+        ],
+        extractor_name="llm_json_extractor",
+        provider="lm_studio",
+        raw_output={"extraction_mode": "primary"},
+    )
+    context = ExtractionPageContext(
+        url=page.url,
+        final_url=page.url,
+        title=page.title,
+        page_type="ArticlePage",
+        clean_text=page.cleaned_text_summary,
+        source_zones=[
+            {
+                "zone_type": "article_body",
+                "selector": "article",
+                "text": "Semantic Systems was authored by Ada Kim.",
+                "claim_allowed": True,
+            }
+        ],
+    )
+
+    claims = repo.save_extraction_bundle(project.id, source, page, attach_page_context(bundle, context), config)
+
+    triples = session.query(models.OntologyTriple).all()
+    assert len(claims) == 1
+    assert triples[0].predicate == "authoredBy"
+    assert triples[0].status == "merged"
+    assert session.query(models.OntologyRelationType).filter_by(name="authoredBy").one().domain == "academic"
+    session.close()