Find answers from the community

Updated 2 years ago

Use case

My use case is I want to go through each document node and extract information that might be relevant to an arbitrary question/prompt. Then put anything relevant into a json object and return that, and keep concatenating to that

6 comments

LLogan M

That sounds like a list index.

A list index will check every node in your index

And vector index will only check the top k most similar (according to embeddings)

nnbulkz

thanks, that what I figured. That's why I was surprised to see the embeddings word in the ListIndex documentation

nnbulkz

you said, it will generate embeddings at query time if you set that. Can you give me an example of that? is that extra api call charges? under the hood what data is traveling back and forth

LLogan M

Under the hood it's the same as a vector index, your embed model will get hit with all the text in the index.

So either openai or some other model you set up will calculate the embeddings

nnbulkz

ok so it effectively creates a GPTSimpleVectorIndex out of the GPTListIndex?

LLogan M

Yea pretty much. Bit of an odd feature tbh lol but it's there

Add a reply