NVIDIA / NVIDIA/cuvs

[FEA] Allow search params to vary within each batch of queries

Open
#2,002 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

feature request
Dominant language
Cuda
Stars
854
Forks
236
Avg merge
3d 3h
Merged PRs (30d)
62

Description

Currently, our search APIs assume all queries in a batch should share the same search parameters.

This makes it challenging for databases that might already have an optimizer or heuristics for selecting parameters based on other properties of the vector space or metadata distributions and want to specify different search parameters for each query. In short, this forces these users to assume they will always be doing an online query (batch size of 1) even if they have several 10s or hundreds of queries that can be batched together.

Users have requested the ability to specify separate search params for each query vector, and for cuVS to handle this case efficiently no matter what params are being varied.

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start by tracing the batch search API and the current handling of shared search parameters. Determine how per-query parameters should be represented and processed efficiently across varied batches. Done means batched queries can use separate search parameters without falling back to one-query-at-a-time execution.

Written by the indexing model from the issue text.

Assessment

Domain
api, search
Issue type
Feature
Difficulty
5/5
Estimated time
Over a week
Activity status
Quiet
Clarity
Needs clarification
Newbie friendliness
35/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.