add value-shape field classification and auto-apply

Field roles are now inferred from observed values instead of names,
with src/dst disambiguation, value-pattern detector hints, an Ollama
advisor scoped to ambiguous fields only, confidence-gated auto-apply
with manual-field protection, and drift flagging for review

tests/test_history.py

@@ -0,0 +1,71 @@
+import tempfile
+import time
+import unittest
+from pathlib import Path
+
+from fgai.history import FieldDiscoveryStore
+
+
+class FieldDiscoveryConfidenceHistoryTests(unittest.TestCase):
+    def test_record_and_read_applied_confidence(self):
+        with tempfile.TemporaryDirectory() as tmp:
+            store = FieldDiscoveryStore(str(Path(tmp) / "history.sqlite3"))
+            store.record_applied_confidence(
+                "custom_vendor",
+                ["x7_field_a", "x7_field_b"],
+                field_confidence={"x7_field_a": 85, "x7_field_b": 40},
+                field_shapes={"x7_field_a": "ip", "x7_field_b": "categorical_low_cardinality"},
+            )
+
+            applied = store.applied_confidence("custom_vendor")
+
+            self.assertEqual(applied["x7_field_a"]["applied_confidence"], 85)
+            self.assertEqual(applied["x7_field_a"]["applied_shape"], "ip")
+            self.assertEqual(applied["x7_field_b"]["applied_confidence"], 40)
+
+    def test_record_ignores_fields_without_confidence_entry(self):
+        with tempfile.TemporaryDirectory() as tmp:
+            store = FieldDiscoveryStore(str(Path(tmp) / "history.sqlite3"))
+            store.record_applied_confidence(
+                "custom_vendor",
+                ["x7_field_a", "timestamp"],
+                field_confidence={"x7_field_a": 85},
+                field_shapes={"x7_field_a": "ip"},
+            )
+
+            applied = store.applied_confidence("custom_vendor")
+
+            self.assertIn("x7_field_a", applied)
+            self.assertNotIn("timestamp", applied)
+
+    def test_record_keeps_first_snapshot_and_ignores_later_calls(self):
+        with tempfile.TemporaryDirectory() as tmp:
+            store = FieldDiscoveryStore(str(Path(tmp) / "history.sqlite3"))
+            store.record_applied_confidence("custom_vendor", ["x7_field_a"], {"x7_field_a": 65}, {"x7_field_a": "ip"})
+            store.record_applied_confidence("custom_vendor", ["x7_field_a"], {"x7_field_a": 90}, {"x7_field_a": "ip"})
+
+            applied = store.applied_confidence("custom_vendor")
+
+            self.assertEqual(applied["x7_field_a"]["applied_confidence"], 65)
+            self.assertEqual(len(applied), 1)
+
+    def test_confidence_history_maintenance_prunes_old_rows(self):
+        with tempfile.TemporaryDirectory() as tmp:
+            db_path = Path(tmp) / "history.sqlite3"
+            store = FieldDiscoveryStore(str(db_path))
+            store.record_applied_confidence("custom_vendor", ["x7_field_a"], {"x7_field_a": 85}, {"x7_field_a": "ip"})
+
+            import sqlite3
+            old_time = int(time.time()) - 200 * 86400
+            with sqlite3.connect(db_path) as connection:
+                connection.execute("update field_confidence_history set applied_at = ?", (old_time,))
+
+            result = store.confidence_history_maintenance(retention_days=90)
+
+            self.assertEqual(result["removed"], 1)
+            self.assertEqual(result["remaining"], 0)
+            self.assertEqual(store.applied_confidence("custom_vendor"), {})
+
+
+if __name__ == "__main__":
+    unittest.main()