[
https://issues.apache.org/jira/browse/SOLR-18074?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18076032#comment-18076032
]
ASF subversion and git services commented on SOLR-18074:
--------------------------------------------------------
Commit 2c1465d10c354e70afbf2463f18770dcbf5e7437 in solr's branch
refs/heads/branch_9x from David Smiley
[ https://gitbox.apache.org/repos/asf?p=solr.git;h=2c1465d10c3 ]
SOLR-18074: DenseVectorField multi-valued fields / block-join (#4191)
Support for multi-valued dense vector fields (through nested vectors and
diversifying children query)
Co-authored-by: copilot-swe-agent[bot]
<[email protected]>
Co-authored-by: Alessandro Benedetti <[email protected]>
> MultiValued dense vectors support (through inner nested vectors)
> ----------------------------------------------------------------
>
> Key: SOLR-18074
> URL: https://issues.apache.org/jira/browse/SOLR-18074
> Project: Solr
> Issue Type: New Feature
> Components: vector-search
> Affects Versions: 10.0
> Reporter: Alessandro Benedetti
> Assignee: Alessandro Benedetti
> Priority: Major
> Labels: pull-request-available
> Time Spent: 0.5h
> Remaining Estimate: 0h
>
> Scope of this issue is to give an additional way to users to handle documents
> with multiple vectors.
> This is already possible through block join nested vectors and it requires to
> model the information with nested documents (see
> [https://solr.apache.org/guide/solr/latest/indexing-guide/indexing-nested-documents.html])
> This issue adds the possibility of handling multi valued vectors directly as
> multi-valued fields (useful when the nested document has no other field than
> the vector).
> This contribution ease the indexing process hiding the complexity of knowing
> that behind the scene, there are nested documents (different context but
> similar approach to partial updates - that hide fetch, update and re-index).
> The contribution introduces support both at indexing and query time:
> At indexing time the user can index multi valued vectors as they would do for
> any other multi valued fields.
> At query time, the child transformer allow to return in the results the
> original document with the best child vector, or all child vectors.
> A draft pull request will be available soon, then documentation and
> discussion will follow.
> N.B. this is not meant to be perfect, it's a sponsored contribution by a
> client of ours that I believe can be useful to a wider audience
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
