search_kwargs not being used in vectorstore as_retriever

Checked other resources

[X] I added a very descriptive title to this issue.
[X] I searched the LangChain documentation with the integrated search.
[X] I used the GitHub search to find a similar question and didn't find it.
[X] I am sure that this is a bug in LangChain rather than my code.
[X] The bug is not resolved by updating to the latest stable version of LangChain (or the specific integration package).

Example Code

In the following code:

vector_store = AzureSearch(...)
retriever = vector_store.as_retriever(
            search_type = "similarity",
            search_kwargs = {
                "k": 8,
                "search_type": 'hybrid',
                "filters": "(x eq 'foo') and (y eq 'bar')"  # Azure AI Search filter
            }
        )

When _get_relevant_documents is called, the provided search_kwargs are not used -- the defaults (k=4, similarity, no filter) are used instead. Although these are stored in the retriever in _lc_kwargs, this doesn't seem to be referenced anywhere.

This seemed to work before -- there may have been an issue introduced somewhere in between these versions:

langchain==0.1.16 -> 0.1.17 langchain-community==0.0.32 -> 0.0.36 langchain-core==0.1.42 -> 0.1.50 langchain-openai==0.1.3 ->0.1.6

I am using the retriever above as part of a custom doc retriever -- the work-around is for me to set search_kwargs directly in the retriever returned by as_retriever right before I call get_relevant_documents, rather than depending on the args I gave as_retriever. (I believe I got an error from pydantic if I try to set these earlier.)

Error Message and Stack Trace (if applicable)

N/A

Description

I'm trying to use search_kwargs to set the ACS query options. The expected behavior is that they should be honored. What is currently happening is that defaults are used instead.

System Info

langchain==0.1.17 langchain-community==0.0.36 langchain-core==0.1.50 langchain-openai==0.1.6

Windows Python 3.10.11

langchain-ai / langchain