Keep raw file references when refreshing imported conversations

src/backend/services/ai_chat_session_ingestion_service.py

@@ -119,6 +119,7 @@ class SessionDecision:
     record: SessionRecord
     document_concept_id: str
     repair_file_copy_concept_id: str | None = None
+    existing_file_copy_concept_id: str | None = None
 
 
 @dataclass
@@ -781,6 +782,7 @@ def run(
                     "source_session_id": decision.record.source_session_id,
                     "document_concept_id": decision.document_concept_id,
                     "action": "unchanged",
+                    "file_copy_concept_id": decision.existing_file_copy_concept_id,
                     "reason": decision.reason,
                     "success": True,
                     "storage_object_written": False,
@@ -1073,6 +1075,9 @@ def _classify_record(self, record: SessionRecord) -> SessionDecision:
                 reason="content_hash_match",
                 record=record,
                 document_concept_id=document_concept_id,
+                existing_file_copy_concept_id=self._read_attribute(
+                    existing, "current_file_copy_concept_id"
+                ),
             )
 
         return SessionDecision(

src/backend/services/external_conversation_import_service.py

@@ -30,7 +30,7 @@
 
 
 EXTERNAL_CONVERSATION_PACKAGE_SCHEMA_VERSION = "external_conversation_package.v1"
-EXTERNAL_CONVERSATION_PARSER_VERSION = "2026-09-16.1"
+EXTERNAL_CONVERSATION_PARSER_VERSION = "2026-09-16.2"
 EXTERNAL_CONVERSATION_ORIGIN_KIND = "external_conversation_import"
 EXTERNAL_CONVERSATION_LINEAGE_SCHEMA_VERSION = "conversation_lineage.v1"
 MAX_EXTERNAL_CONVERSATION_SOURCE_BYTES = 32 * 1024 * 1024

tests/backend/test_ai_chat_session_ingestion_service.py

@@ -203,7 +203,6 @@ def discover(self):
 def test_run_unchanged_record_skips_mutation(monkeypatch, tmp_path):
     from src.backend.services.ai_chat_session_ingestion_service import (
         AIChatSessionIngestionService,
-        SessionDecision,
     )
 
     record = _record(tmp_path=tmp_path)
@@ -220,15 +219,20 @@ def discover(self):
     )
 
     monkeypatch.setattr(service, "ensure_ontology_types", lambda: [])
+    _install_source_profile_authority(service)
     monkeypatch.setattr(
         service,
-        "_classify_record",
-        lambda _rec: SessionDecision(
-            action="unchanged",
-            reason="content_hash_match",
-            record=_rec,
-            document_concept_id=_rec.document_concept_id,
-        ),
+        "_get_concept",
+        lambda _concept_id: {
+            "concept_id": record.document_concept_id,
+            "attributes": {
+                "source_content_sha256": record.content_sha256,
+                "current_file_copy_concept_id": "#V#retained_file_copy",
+            },
+            "relationships": {
+                service._document_has_file_predicate_id(): ["#V#retained_file_copy"]
+            },
+        },
     )
 
     result = service.run(dry_run=False)
@@ -239,6 +243,7 @@ def discover(self):
     assert result.counters.skipped == 1
     assert result.counters.intended_mutations == 0
     assert result.counters.executed_mutations == 0
+    assert result.records[0]["file_copy_concept_id"] == "#V#retained_file_copy"
 
 
 def test_dry_run_uses_ontology_validation_not_bootstrap(monkeypatch, tmp_path):
@@ -272,6 +277,7 @@ def _validate():
 
     monkeypatch.setattr(service, "ensure_ontology_types", _bootstrap)
     monkeypatch.setattr(service, "validate_ontology_types", _validate)
+    monkeypatch.setattr(service, "_get_concept", lambda _concept_id: None)
 
     result = service.run(dry_run=True)