feat(api): Allow the `node_execution_id` field to be `NULL`.

api/migrations/versions/2025_06_06_1424-4474872b0ee6_workflow_draft_varaibles_add_node_execution_id.py

@@ -34,7 +34,7 @@ def upgrade():
         )
 
     with op.batch_alter_table('workflow_draft_variables', schema=None) as batch_op:
-        batch_op.add_column(sa.Column('node_execution_id', models.types.StringUUID(), nullable=False))
+        batch_op.add_column(sa.Column('node_execution_id', models.types.StringUUID(), nullable=True))
     # ### end Alembic commands ###
 
 

api/models/workflow.py

@@ -919,7 +919,14 @@ class WorkflowDraftVariable(Base):
 
     # The `node_execution_id` field identifies the workflow node execution that created this variable.
     # It corresponds to the `id` field in the `WorkflowNodeExecutionModel` model.
-    node_execution_id: Mapped[str] = mapped_column(StringUUID, nullable=False)
+    #
+    # This field is not `None` for system variables and node variables, and is  `None`
+    # for conversation variables.
+    node_execution_id: Mapped[str | None] = mapped_column(
+        StringUUID,
+        nullable=True,
+        default=None,
+    )
 
     def get_selector(self) -> list[str]:
         selector = json.loads(self.selector)