Feature request / question: Retrieval-only function with top_k support

[bmelloul]

Hi,

When running
from graphrag.api.query import local_search
answer, context = await local_search(...)
The retrieved context is in context["sources"]

I have two questions about the retrieval step:

1. Is there a way to specify a parameter like top_k that returns the top k chunks from the retrieval?
2. Is there a function that performs only retrieval, without also performing generation (unlike local_search which does both retrieval and generation)? Essentially, I want to fetch the relevant chunks first, then handle generation separately.

Thanks in advance!

[natoverse]

The ContextBuilder classes for each search type can be used to perform the retrieval process. For example, BasicSearchContext will just execute the top k search against the vector store to get the relevant text chunks.