Neo4j Aura June Database Release

The Neo4j Aura June release has started to roll out, commencing on AuraDB Free and progressing to higher tiers later.

Cypher 25 Features

CALL { ... } IN CONCURRENT TRANSACTIONS
now supports
DISJOINT BY
, a deadlock prevention mechanism that eliminates lock contention in parallel write workloads natively. High concurrency ingestion, particularly when merging nodes with unique constraints or creating relationships, has always been vulnerable to deadlock, as parallel threads compete for the same locks simultaneously.
DISJOINT BY
resolves contention at scheduling time, before any transaction is attempted. See the Cypher Manual → Batch scheduling and deadlock prevention with DISJOINT BY.
GRAPH TYPE
moves from public preview to GA in this release. You can now define, enforce, and validate your graph schema at the database level in production workloads with the stability guarantees that come with a fully supported feature. See the Cypher Manual → Graph types.
Extend
SHOW CURRENT GRAPH TYPE
to be able to return lists of virtual nodes and relationships representing the graph type (similar in format to the output of db.schema.visualization()) instead of the current string representation. To get the graph version of the command, use
SHOW CURRENT GRAPH TYPE AS GRAPH
. See the Cypher Manual → SHOW CURRENT GRAPH TYPE AS GRAPH.
For customers using Change Data Capture (CDC), we have added a
txCommitTime
column to the
db.cdc.current()
procedure. This update allows clients to conveniently identify the commit time of their most recent transaction alongside the transaction identifier.
Vector
SEARCH
queries now support
IN
for filtering, allowing queries like:

MATCH (movie:Movie)
  SEARCH movie IN (
    VECTOR INDEX moviePlots
    FOR $queryVector
    WHERE movie.genre IN ["Horror", "SciFi"]
    LIMIT $topK
  )
RETURN movie.title AS title, movie.rating AS rating

See the Cypher Manual → Predicates for vector search with filters.

Other relevant updates

Preview feature

: an improved vector search that uses our new Hi-Fidelity Quantized Vector Search method. HFQ performs an initial expanded search on the quantized vectors, then re-ranks the results using the full precision vectors. This should make queries faster, reduce memory requirements, and increase recall accuracy when compared to the previous default quantized vector index. HFQ adds the following options to

CREATE VECTOR INDEX

, vector.quantization.type and vector.default_search_expansion_factor.

To try the preview of HFQ:

CREATE VECTOR INDEX moviePlots IF NOT EXISTS
FOR (m:Movie)
ON m.embedding
OPTIONS { indexConfig: {
  `vector.quantization.type`: 'binary',
  `vector.default_search_expansion_factor`: 2.0,
  `vector.dimensions`: 1536,
  `vector.similarity_function`: 'cosine'
}}

HFQ requires an index re-build and is expected to be GA in the next release.

Aura Graph Analytics (AGA)

Native projections are now available in AGA via both Cypher and the GDS Python client.
Project-level Machine Learning model publishing is now supported, automatically granting model access to all members of a project.

Security

Upgraded netty to version
4.2.15.Final
in order to resolve several CVEs.

Bug fixes

Fixed a bug that allowed cached pages to exceed the configured limit.
Fixed a bug where outbound network chunks for secondaries were not flushed properly. This could lead to a buffer overflow in the outbound.
Fixed a bug in the pipelined runtime where an
OR EXISTS
subquery containing ordering could silently drop a result row:

MATCH (c:Customer)
WHERE c.vipStatus = true OR EXISTS { MATCH (c)-[:PLACED]->(o:Order) RETURN o ORDER BY o.total DESC }
RETURN c.name

Fixed a record format bug where undirected scans of multiple relationship types omitted sibling relationships, resulting in a ~50% undercount in the output.
Fixed an issue that let a User Defined Function (UDF) be defined as part of a Property-Based Access Control (PBAC) privilege. This isn't supported and wouldn't have worked as a user might have expected.

For full details of all updates and fixes in this release, as well as links to relevant documentation please visit: Release Notes: Neo4j Aura Database – June 2026.