Skip to content

Design details for ranking strategy #346

@arindam-g-ghosh-capgemini-com

Description

Once the vector search is done for a user query, for each and every result row, a single llm is done to get raking etc. details. I wanted to understand the design considerations for the same instead of doing a batch ranking. Is it because of putting quality + precision over performance + token usage ?

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions