Find answers from the community

Updated 3 months ago

```

Plain Text

index = GPTSimpleKeywordTableIndex(nodes=nodes,
                            storage_context=storage_context,
                            service_context=service_context,
                            show_progress=True
                            )

retriever = index.as_retriever()
nodes = retriever.retrieve("formation")

Hello,

I need your help for my use case.

How to retrieve all matching nodes; not only 2 or similarity_top_k. I want to retrieve all nodes matching my query.

7 comments

LLogan M

You are using a keyword index

So it already returns all nodes with matching keywords 👀

iisaac47

Really? because when I check in the text from my nodes I found much more nodes than the want returns.

And what if I use GPTVectorStoreIndex?

Here is my use case. I have a data base with job description. I want to retrieve all nodes releated to the query via semantic or keyword search.

LLogan M

How it works here is it takes your query, extracts keywords, and retrieves nodes that have the same keywords from the table. Pretty basic. You can see when you query I think which keywords it extracted too

If you want to combine keyword/semantic, maybe try BM25 + vector store index (you could skip the re-ranking here)
https://gpt-index.readthedocs.io/en/stable/examples/retrievers/bm25_retriever.html#advanced-hybrid-retriever-re-ranking

iisaac47

Thanks alot for our answer. So we do not need in this case to add similarity top k param right ?

LLogan M

yea correct. Right now, your keywrod index will just return every matching node

iisaac47

And how to check extracted keywords from the query?

LLogan M

it should be printing them to the console, but if not, run this first

Plain Text

import logging
import sys

logging.basicConfig(stream=sys.stdout, level=logging.INFO)
logging.getLogger().addHandler(logging.StreamHandler(stream=sys.stdout))

Add a reply