Skip to content

Conversation

@dishaprakash
Copy link
Collaborator

@dishaprakash dishaprakash commented Oct 7, 2025

This PR fixes the 2nd Major issue in #234, related to Configuration and Querying Issues

Issue breakdown:

Inconsistent k Parameter: For hybrid searches, the dense search LIMIT was tied to the final k parameter, while the sparse search used its own secondary_top_k.
Late Initialization of HybridSearchConfig: The HybridSearchConfig was initialized after the k parameter was calculated, which could lead to incorrect behavior if the config was passed dynamically via kwargs.

Overview of the changes:

  1. The __query_collection method now uses a separate dense_limit for the dense search query.
    If HybridSearchConfig is active, dense_limit is set to primary_top_k. For dense-only searches, dense_limit defaults to the final k value, preserving the existing behavior.

  2. The hybrid_search_config is now initialized at the beginning of __query_collection, ensuring that any kwargs overrides are handled before k or dense_limit are calculated.

@dishaprakash dishaprakash marked this pull request as ready for review October 9, 2025 12:51
@averikitsch averikitsch merged commit 396282f into langchain-ai:main Oct 10, 2025
2 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

2 participants