Context

Currently the way to initialize MongoDB Atlas Vector Search is via passing an initialized collection 

*https://github.com/langchain-ai/langchain-mongodb/blob/main/libs/langchain-mongodb/langchain_mongodb/vectorstores.py#L207*

vector_store = MongoDBAtlasVectorSearch(
 collection=MONGODB_COLLECTION,
 embedding=embeddings,
 index_name=ATLAS_VECTOR_SEARCH_INDEX_NAME,
 relevance_score_fn="cosine",
)

This requires the user to do a bunch of pre-work in initializing the collection, which involves steps like 

# initialize MongoDB python client
client = MongoClient(MONGODB_ATLAS_CLUSTER_URI)

DB_NAME = "langchain_test_db"
COLLECTION_NAME = "langchain_test_vectorstores"
ATLAS_VECTOR_SEARCH_INDEX_NAME = "langchain-test-index-vectorstores"

MONGODB_COLLECTION = client[DB_NAME][COLLECTION_NAME]

If we can modify the MongoDBAtlasVectorSearch to also allow for passing the required params (collection_name, db_name, cluster_uri) we can initialize the client ourselves. 

vector_store = MongoDBAtlasVectorSearch(
 collection=MONGODB_COLLECTION, :: OPTIONAL 
 collection_name= XXX
 db_name= XXX
 connection_string = XXX
embedding=embeddings, 
index_name=ATLAS_VECTOR_SEARCH_INDEX_NAME, 
relevance_score_fn="cosine", )

We want to maintain backward compatibility, and don't want to create a breaking change.

To ideate on the options: 

• The earlier 'collection' was a required field, which will can be made optional? The new parameters can be optional as well
• Make another init method that can provide the desired param (and not have the "collection" param)?

Definition of done

What must be done to consider the task complete?

Pitfalls

What should the implementer watch out for? What are the risks?