Vigya Sharma

This change stores `node` and `score` into a single `ScoreNode` object instead of maintaining separate arrays. I think it simplifies operations like sorting the neighbor nodes when they're added out...

Test coverage for `OldGenDecisionPolicy` has lowered, creating an issue to track pending tests.

> Extending discussion on PR #402 There can be several reasons for high old heap usage, cache and queues cover only a subset of them. Currently however, these are the...

[WIP] Multi-Vector support for HNSW search

30

Adds support for multi-valued vectors to Lucene. In addition to max-similarity aggregations like parent-block joins, this change supports ColBERT style distance functions that compute interaction across all query and document...

Add support for KNN ParentJoin Benchmarks

1

With Lucene supporting [parent-block joins](https://github.com/apache/lucene/pull/12434) for KNN vectors, and ongoing work around [Multi-Vector support](https://github.com/apache/lucene/issues/12313), it would be good to have common, shared benchmarks that can test recall, latency etc. for...

Does HNSW search sometimes match NO_MORE_DOCS?

1

We have a check to separately handle `NO_MORE_DOCS` while parsing the results from an HNSW approximate search. We should not be returning `NO_MORE_DOCS` in the first place. Creating an issue...