fix(memory): centralize entity cleanup and skip malformed LLM relation dicts (#4515)

tests/memory/test_memory_utils.py

@@ -0,0 +1,57 @@
+import pytest
+from mem0.memory.utils import remove_spaces_from_entities, sanitize_relationship_for_cypher
+
+
+class TestRemoveSpacesFromEntities:
+    """
+    Covers behavior used by Neo4j, Memgraph (sanitize_relationship=True),
+    Kuzu, and Neptune (sanitize_relationship=False). All backends delegate here.
+    """
+
+    @pytest.mark.parametrize(
+        "sanitize",
+        [True, False],
+        ids=["cypher_sanitized", "plain"],
+    )
+    def test_filters_empty_and_incomplete_dicts(self, sanitize):
+        mixed = [
+            {},
+            {"source": "a"},
+            {"source": "a", "relationship": "r"},
+            {"source": "x", "relationship": "rel", "destination": "y"},
+        ]
+        out = remove_spaces_from_entities(mixed, sanitize_relationship=sanitize)
+        assert len(out) == 1
+        assert out[0]["source"] == "x"
+        assert out[0]["destination"] == "y"
+
+    @pytest.mark.parametrize("sanitize", [True, False])
+    def test_all_empty_returns_empty(self, sanitize):
+        assert remove_spaces_from_entities([{}, {}, {}], sanitize_relationship=sanitize) == []
+
+    def test_skips_non_dict_entries(self):
+        assert remove_spaces_from_entities([None, "not-a-dict", 123, {"source": "a", "relationship": "r", "destination": "b"}]) == [
+            {"source": "a", "relationship": "r", "destination": "b"}
+        ]
+
+    def test_sanitize_true_relationship_uses_sanitizer(self):
+        """Neo4j / Memgraph path: special characters mapped via sanitize_relationship_for_cypher."""
+        entities = [{"source": "A", "relationship": "x/y", "destination": "B"}]
+        out = remove_spaces_from_entities(entities, sanitize_relationship=True)
+        assert out[0]["relationship"] == sanitize_relationship_for_cypher("x/y".lower().replace(" ", "_"))
+
+    def test_sanitize_false_relationship_plain_only(self):
+        """Kuzu / Neptune path: only lowercase and spaces to underscores."""
+        entities = [{"source": "A", "relationship": "Works At", "destination": "B Co"}]
+        out = remove_spaces_from_entities(entities, sanitize_relationship=False)
+        assert out[0]["relationship"] == "works_at"
+        assert out[0]["source"] == "a"
+        assert out[0]["destination"] == "b_co"
+
+    def test_sanitize_true_vs_false_slash_in_relationship(self):
+        """Slash is rewritten when sanitizing (Cypher path); kept as-is for plain path."""
+        base = {"source": "s", "relationship": "a/b", "destination": "d"}
+        t = remove_spaces_from_entities([dict(base)], sanitize_relationship=True)[0]["relationship"]
+        f = remove_spaces_from_entities([dict(base)], sanitize_relationship=False)[0]["relationship"]
+        assert t == sanitize_relationship_for_cypher("a/b")
+        assert f == "a/b"