Merge pull request #228 from reconurge/fix/node-upsert-clobber-exclude-unset

fix(core): stop node upsert from nulling unset FlowsintType fields

flowsint-core/src/flowsint_core/core/graph/serializer.py

@@ -7,7 +7,7 @@ into Neo4j-compatible primitive types, following the Single Responsibility Princ
 
 from typing import Any, Callable, Dict, List, Optional, Type, Union
 
-from pydantic import ValidationError
+from pydantic import BaseModel, ValidationError
 
 from flowsint_core.utils import flatten, unflatten
 from flowsint_types import FlowsintType
@@ -160,8 +160,20 @@ class GraphSerializer:
 
         Serializes the model to JSON-compatible types and flattens nested
         structures into dot-notation keys for Neo4j property storage.
+
+        nodeProperties uses exclude_unset: repository._build_node_query does
+        `SET n += $props`, so a full dump of a partial FlowsintType would
+        null out fields a prior enricher already set on the same node.
         """
-        neo4j_dict = node.model_dump(mode="json", serialize_as_any=True)
+        neo4j_dict = node.model_dump(
+            mode="json", serialize_as_any=True, exclude={"nodeProperties"}
+        )
+        node_properties = node.nodeProperties
+        neo4j_dict["nodeProperties"] = (
+            node_properties.model_dump(mode="json", exclude_unset=True)
+            if isinstance(node_properties, BaseModel)
+            else node_properties
+        )
         neo4j_dict_flatten = flatten(neo4j_dict, remove_empty=False)
         neo4j_dict_flatten.pop(
             "nodeProperties.nodeLabel", None

flowsint-core/tests/core/graph/test_graph_serializer.py

@@ -8,7 +8,7 @@ from flowsint_core.core.graph import (
     GraphSerializer,
     NodeMetadata,
 )
-from flowsint_types import Domain, Ip
+from flowsint_types import Domain, Ip, Website
 
 
 def test_serializer():
@@ -112,6 +112,47 @@ def test_serialize_from_flowsint_type():
     assert "T" in neo4j_dict["nodeMetadata.created_at"]
 
 
+class TestNodePropertiesExcludeUnset:
+    def test_thin_website_only_sends_fields_it_set(self):
+        website = Website(url="https://example.com/")
+        node = GraphNode(
+            id="id",
+            nodeLabel="https://example.com/",
+            nodeType="website",
+            nodeProperties=website,
+            nodeMetadata=NodeMetadata(),
+        )
+        neo4j_dict = GraphSerializer.graph_node_to_neo4j_dict(node)
+
+        assert neo4j_dict["nodeProperties.url"] == "https://example.com/"
+        for unset_field in (
+            "active",
+            "status_code",
+            "headers",
+            "title",
+            "description",
+            "content",
+            "technologies",
+            "redirects",
+            "domain",
+        ):
+            assert f"nodeProperties.{unset_field}" not in neo4j_dict
+
+    def test_richer_website_keeps_explicitly_set_fields(self):
+        website = Website(url="https://example.com/", active=False, status_code=404)
+        node = GraphNode(
+            id="id",
+            nodeLabel="https://example.com/",
+            nodeType="website",
+            nodeProperties=website,
+            nodeMetadata=NodeMetadata(),
+        )
+        neo4j_dict = GraphSerializer.graph_node_to_neo4j_dict(node)
+
+        assert neo4j_dict["nodeProperties.active"] is False
+        assert neo4j_dict["nodeProperties.status_code"] == 404
+
+
 class TestCleanEmptyValues:
     def test_removes_empty_strings(self):
         data = {"key1": "value", "key2": "", "key3": "another"}