fix: sanitize inverted affiliation dates in git integration

Reject inverted or incomplete project-registry stint date ranges before
they are written to member organizations and segment affiliations.

Signed-off-by: Yeganathan S <63534555+skwowet@users.noreply.github.com>

Author: skwowet

services/apps/git_integration/src/crowdgit/models/affiliation_info.py

@@ -6,7 +6,7 @@
 
 import orjson
 from loguru import logger
-from pydantic import BaseModel, Field, TypeAdapter, ValidationError
+from pydantic import BaseModel, Field, TypeAdapter, ValidationError, model_validator
 
 
 class AffiliationContributor(BaseModel):
@@ -41,6 +41,20 @@ class AffiliationOrganizationStint(BaseModel):
 
     model_config = {"populate_by_name": True}
 
+    @model_validator(mode="after")
+    def sanitize_date_range(self) -> AffiliationOrganizationStint:
+        if self.date_end is not None and self.date_start is None:
+            self.date_start = None
+            self.date_end = None
+        elif (
+            self.date_start is not None
+            and self.date_end is not None
+            and self.date_end < self.date_start
+        ):
+            self.date_start = None
+            self.date_end = None
+        return self
+
 
 class AffiliationContributorEntry(BaseModel):
     contributor: AffiliationContributor

services/apps/git_integration/src/crowdgit/services/affiliation/affiliation_service.py

@@ -329,6 +329,7 @@ def get_extraction_prompt(self, content_to_analyze: str) -> str:
 
         Time period (only when the file states it):
         - "dateStart" and "dateEnd" as ISO dates (YYYY-MM-DD).
+        - dateStart must be on or before dateEnd; if unclear, use null for both.
         - Use null for any bound the file does not state (open-ended or undated).
         - When a contributor has multiple affiliations over time, emit a separate
           entry for each period. Do not merge, deduplicate, or keep only the latest.

services/apps/git_integration/src/test/test_affiliation_date_sanitization.py

@@ -0,0 +1,104 @@
+from datetime import date
+
+import orjson
+
+from crowdgit.models.affiliation_info import (
+    AffiliationContributor,
+    AffiliationContributorEntry,
+    AffiliationOrganizationStint,
+    RepoAffiliationRegistry,
+)
+
+
+def test_stint_nulls_inverted_date_range():
+    stint = AffiliationOrganizationStint(
+        domain="example.com",
+        date_start=date(2025, 8, 11),
+        date_end=date(2022, 9, 20),
+    )
+
+    assert stint.date_start is None
+    assert stint.date_end is None
+
+
+def test_stint_nulls_end_without_start():
+    stint = AffiliationOrganizationStint(
+        domain="example.com",
+        date_end=date(2022, 9, 20),
+    )
+
+    assert stint.date_start is None
+    assert stint.date_end is None
+
+
+def test_stint_keeps_valid_date_range():
+    stint = AffiliationOrganizationStint(
+        domain="example.com",
+        date_start=date(2019, 6, 24),
+        date_end=date(2022, 9, 20),
+    )
+
+    assert stint.date_start == date(2019, 6, 24)
+    assert stint.date_end == date(2022, 9, 20)
+
+
+def test_snapshot_load_sanitizes_inverted_dates():
+    snapshot = [
+        {
+            "contributor": {"email": "user@example.com"},
+            "organizations": [
+                {
+                    "domain": "example.com",
+                    "dateStart": "2025-08-11",
+                    "dateEnd": "2022-09-20",
+                }
+            ],
+        }
+    ]
+
+    registry = RepoAffiliationRegistry.from_db(
+        {
+            "repoId": "00000000-0000-0000-0000-000000000001",
+            "status": "success",
+            "snapshot": orjson.dumps(snapshot).decode(),
+        }
+    )
+
+    assert registry.snapshot is not None
+    stint = registry.snapshot[0].organizations[0]
+    assert stint.date_start is None
+    assert stint.date_end is None
+
+
+def test_snapshot_load_keeps_valid_dates():
+    snapshot = [
+        AffiliationContributorEntry(
+            contributor=AffiliationContributor(email="user@example.com"),
+            organizations=[
+                AffiliationOrganizationStint(
+                    domain="example.com",
+                    date_start=date(2019, 6, 24),
+                    date_end=date(2022, 9, 20),
+                )
+            ],
+        )
+    ]
+
+    registry = RepoAffiliationRegistry(
+        repo_id="00000000-0000-0000-0000-000000000001",
+        status="success",
+        snapshot=snapshot,
+    )
+
+    serialized = registry.snapshot_for_db()
+    loaded = RepoAffiliationRegistry.from_db(
+        {
+            "repoId": "00000000-0000-0000-0000-000000000001",
+            "status": "success",
+            "snapshot": serialized,
+        }
+    )
+
+    stint = loaded.snapshot[0].organizations[0]
+    assert stint.date_start == date(2019, 6, 24)
+    assert stint.date_end == date(2022, 9, 20)