[Question]: Will NodeRelationship affect index building?

[fisipro2]

Question Validation

• I have searched both the documentation and discord for an answer.

Question

Hi

I'm building the nodes (TextNode) maually, I also setup the child/parent relationships for the node.
I want to know:

1. Will the relationships be used when building up the index ( embeedings ), like if I don't setup the relationships, will there be a big difference in query ?

2. Do I have to setup the SOURCE relationship as well? Will there be a difference if set/not set the SOURCE realtionship for the node?

Actually I'm using pinecode to hold the vector, I want to update just some nodes each time, I found the index.delete_nodes has been implemented yet, the only way to delele the node is to call delete_ref_doc, which will delete all the nodes belongs to the same doc_id, that' not what I want, I want to just delete specific nodes, not all of it, so my idea is that if I can assign the SOURCE relationship a unique random id/tag for each node, by doing this, the SOURCE realtionship will has nothing to do with the relationship, it will be just used for holding custom data). Later I can delete each node individually by calling the delete_ref_doc and passing the random id I generated, not sure if this will affect the query result ? )

Thanks for your help.

[dosubot[bot]]

🤖

Hi,

Thanks for your questions. Let's address them one by one:

1. The relationships between nodes (including parent-child relationships) are indeed used when building the index. The LlamaIndex handles these relationships in a bottom-up fashion during the index building process. This is done by the GPTTreeIndexBuilder in the _build_index_from_nodes method. Each node in the resulting tree-structured index is a summary of its children nodes. If you don't set up these relationships, the structure of your index and the results of your queries could be significantly different.

2. The SOURCE relationship is used to link nodes to their original source documents, which can be used to retrieve additional information when querying the index. If you don't set up the SOURCE relationship, you might not be able to retrieve as much information from your queries.

As for your idea of assigning a unique random id/tag to each node's SOURCE relationship to delete specific nodes, unfortunately, the current version of LlamaIndex does not support the deletion of specific nodes individually. The _delete_node method in the KnowledgeGraphIndex class raises a NotImplementedError with the message "Delete is not supported for KG index yet." This suggests that the functionality to delete specific nodes is not yet implemented.

However, each node does have a unique ID (node_id) that is generated when the node is created. This ID could potentially be used to identify and delete specific nodes if such functionality were to be implemented in the future.

I hope this answers your questions. If you have any more questions or need further clarification, feel free to ask.

Best,
Dosu

Sources

• llama_index/indices/knowledge_graph/base.py
• llama_index/schema.py
• llama_index/indices/tree/base.py

---

This response is meant to be useful and save you time. It is not meant to be a precise solution, but rather a starting point for your own research.

Help me be more useful! Please leave a 👍 if this is helpful and 👎 if it is irrelevant.

If you want to continue the conversation, start your reply with @dosu-bot.