lance-format/lance

Allow issue vector search in batch mode

Open

#876 opened on May 17, 2023

View on GitHub
 (3 comments) (0 reactions) (0 assignees)Rust (695 forks)github user discovery
A-indexA-pythonenhancementgood first issuehelp wantedrust

Repository metrics

Stars
 (6,582 stars)
PR merge metrics
 (Avg merge 6d 1h) (219 merged PRs in 30d)

Description

Problem Statement

Instead of query one vector a time, allows user to issue multiple queries in the same request, exposing more opportunity for internal parallisms.

enable vector searching for a batch of queries. Currently Lance only allows 1D queries, but allowing 2D queries (parallelized) would be useful. Use case #1 is: at query time to run multiple simultaneous queries 'around' the root query to diversify search results, and #2 is useful for ETL operations (like the above deduplication) Agree it would be amazing that lance support this api directly to minimize read requests to object store. Recent work to cache the index already helps

Contributor guide

Research direction
Study the existing single vector search implementation and design an interface for batch queries. Consider parallelization and minimizing reads to object store.
Tech stack
rust
Domain
backenddatabase
Issue type
Feature
DifficultyEstimated implementation difficulty for a new contributor, from 1 for very small changes to 5 for expert-level work.
3
Estimated timeA rough time range for an experienced contributor to investigate, implement, test, and prepare a pull request.
3-5 days
Activity statusHow available the issue appears right now: fresh, active, stale, blocked, or waiting on maintainer input.
Active
ClarityHow clearly the issue explains the expected change, acceptance criteria, and next step.
Clear
Prerequisites
RustGit
Newbie friendlinessA 1-100 score estimating how approachable this issue is for first-time contributors.
40