Hi everyone,

sstorm

Hi everyone,
I’m trying to implement something similar to this example: https://docs.llamaindex.ai/en/stable/examples/query_engine/multi_doc_auto_retrieval/multi_doc_auto_retrieval/

Where I would retrieve an IndexNode from one index then, because its an IndexNode, recursively retrieve the referenced document in the IndexNode using a retriever stored in the obj.

However, LlamaIndex doesn’t persist the obj for later use, as its saved as null. Is there a way to store the obj or is there a way around this?

7 comments

LLogan M

What is the obj pointing to? If its serialiable, it will get saved, otherwise you have to provide it again at load time

sstorm

The obj is a retriever pointing to a vector database

LLogan M

yea thats not going to be serializable 😅

You can do this when re-connecting to your vector store

Plain Text

from llama_index.core.schema import IndexNode

vector_obj = IndexNode(
    index_id="vector", obj=vector_retriever, text="Vector Retriever"
)

index = VectorStoreIndex.from_vector_store(vector_store, objects=[vector_obj], ...)

sstorm

I’m pretty much doing it the same way as the example i sent, the summary vector store holds IndexNodes where the ‘obj’ is a retriever with a filter to another vector store.

I’m assuming this wouldn’t work when reconnecting?

LLogan M

When reconnecting you have to pass in the objects again, like the above

LLogan M

since the index node objects aren't serializable

sstorm

Gotcha, thank you!

Add a reply

Find answers from the community

Hi everyone,