Exploring the Document Summary Index in LlamaIndex

1. Yes you can create DocumentSummaryIndex from your existing index but there is no direct way. you'll have to do something like:

1. Docstore allows you for easy access to the nodes created. You can easily iterate over the nodes by accessing the nodes. Also docstore is created if you are creating index locally or you are specifying the docstore in case of vector store.

Vector Store: Its basically a combination of your nodes and embeddings.

You can find more detail here: https://docs.llamaindex.ai/en/stable/module_guides/storing/#concept

i see! if i were to understand it correctly, essentially we have to rescontruct the documents from the nodes in vectorStore and create a new DocumentSummaryIndex with the documents?


it seems that persisting nodes, and indexes in a docStore and indexStore respectively in mongoDB is the 'right way' to do it as its more extensible to support the different indexes.

Can i confirm that persisting docStore and indexStore in mongoDB is possible? i ultimately want to use both vectorIndex and summary Index.

if i were to understand it correctly, essentially we have to rescontruct the documents from the nodes in vectorStore and create a new DocumentSummaryIndex with the documents? Yes, if you dont want to ingest document again and want to use the same nodes present in vector store.

Yes you can persist docstore, indexstore in mongoDB: https://docs.llamaindex.ai/en/stable/examples/docstore/MongoDocstoreDemo/

interesting! i will try this approach! 🧑‍🚀

thank alot for the help!

oyes, i have another qn.

suppose i have a document which represents a pdf page. I further chunk it into 5 nodes.
will i expect to see 6 entities in my docStore? 1 document + 5 nodes

No docstore will only show the chunked nodes: that is 5 nodes

thanks alot whitefang