Issues

LSM_VECTOR: a DOT_PRODUCT index accepts non-unit vectors, which silently breaks the metric's score range

bug

#5750

· lvca opened

on Aug 2, 2026

LSM_VECTOR: a query with an allow-list narrower than k costs a full index scan, every time

bug

#5748

· lvca opened

on Aug 2, 2026

Database.close() is O(size) for vector indexes and analytical views (34.5 s at 1M vectors), O(1) for everything else

#5747

· tae898 opened

on Aug 2, 2026

openCypher multi-hop patterns return zero rows when planned through GAVFusedChain; SQL MATCH correct on the same data

#5746

· tae898 opened

on Aug 2, 2026

Graph Analytical View: traversal is accelerated but far-endpoint property access is not, and the vectorized column accessor has no callers

#5745

· tae898 opened

on Aug 1, 2026

An index rebuilt from its own definition still loses its full-text and geospatial configuration at four sites

bug

#5742

· lvca opened

on Aug 1, 2026

Follow-ups from #5680: an append racing a vertex delete leaves a ghost edge, and deleteVertex materialises the whole edge list

bug

Bug

#5725

· lvca opened

on Aug 1, 2026

·

26.8.1

HASH: the key-type check still sits in the bypassable factory handler, unlike the page-size check

#5724

· lvca opened

on Aug 1, 2026

·

26.8.1

Renaming a type resets every index's page size to the default

bug

#5723

· lvca opened

on Aug 1, 2026

·

26.8.1

HTTP query endpoint: no hard ceiling on the rows a single response materializes

enhancement

#5719

· lvca opened

on Aug 1, 2026

·

26.8.1

[Cypher] the two count push-downs disagree on their preconditions and their entry points: a dropped SKIP/LIMIT, and three shapes that scan when they need not

Bug

#5715

· lvca opened

on Aug 1, 2026

·

26.8.1

HASH index: the sorted slot directory makes inserts scale with page size, and the 64KB default costs ~2x

performance

Feature

#5712

· lvca opened

on Aug 1, 2026

·

26.9.1

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Uh oh!

All issues

LSM_VECTOR: a DOT_PRODUCT index accepts non-unit vectors, which silently breaks the metric's score range

LSM_VECTOR: a query with an allow-list narrower than k costs a full index scan, every time

Database.close() is O(size) for vector indexes and analytical views (34.5 s at 1M vectors), O(1) for everything else

openCypher multi-hop patterns return zero rows when planned through GAVFusedChain; SQL MATCH correct on the same data

Graph Analytical View: traversal is accelerated but far-endpoint property access is not, and the vectorized column accessor has no callers

An index rebuilt from its own definition still loses its full-text and geospatial configuration at four sites

Follow-ups from #5680: an append racing a vertex delete leaves a ghost edge, and deleteVertex materialises the whole edge list

HASH: the key-type check still sits in the bypassable factory handler, unlike the page-size check

Renaming a type resets every index's page size to the default

HTTP query endpoint: no hard ceiling on the rows a single response materializes

[Cypher] the two count push-downs disagree on their preconditions and their entry points: a dropped SKIP/LIMIT, and three shapes that scan when they need not

HASH index: the sorted slot directory makes inserts scale with page size, and the 64KB default costs ~2x

Uh oh!

Uh oh!

Search results