So we only support graph nodes upload for now. The reason is, even if you are able to upload output nodes, you won't be able to upload code nodes and output nodes have to be tied to their respective code nodes.
Moreover, we need to specify the condition on which a particular output node is connected to its corresponding code node. In a bulk upload(which is mostly a CSV), such nested relations could quickly become a mess. We thus stick to bulk uploads for simpler and more straightforward use cases.
Hope this answer helps.