SEARCH clause with WHERE in Community Edition — GA status and supported predicates for multi-tenant filtering?

Neo4j Graph Platform
1

  1. Is SEARCH ... IN (VECTOR INDEX ... WHERE ...) GA on neo4j:2026.04.0-community, or still preview?

  2. Is list membership supported in the inner WHERE? I need this for role-based access:

    SEARCH section IN (
  VECTOR INDEX sectionEmbedding
  FOR $query_vec
  WHERE section.company_id = $company_id
    AND $role_id IN section.allowed_roles
  LIMIT $top_k
) SCORE AS score

  3. Do existing vector indexes created via db.index.vector.createNodeIndex work with the new SEARCH clause, or do they need to be rebuilt?

2

On 3, as far as I know the indexes created through db.index.vector.createNodeIndex will work with the SEARCH clause. However, they will not have any additional properties to filter on as those are added upon index creation in the Cypher CREATE VECTOR INDEX command.

So to have anything to filter on in the WHERE clause, they would need to be rebuilt/recreated with the relevant filtering properties specified

CREATE VECTOR INDEX my_index FOR (n:Label) ON (n.vectorProperty) WITH [n.filteringProperty1, n.filteringProperty2]
3

Is SEARCH ... IN (VECTOR INDEX ... WHERE ...) GA on neo4j:2026.04.0-community, or still preview?
can anyone confirm this?

4

Hi @swamy.prateek11

Sorry for not getting back to you sooner :disappointed_face:

However, yes the the feature in 2026.04.0 and is no longer in preview. Regarding your second question, in 2026.04.0 we still don't support list membership but the feature has been added internally and we think it will be available in the upcoming 2026.06 release.

Indexes created with the db.index.createNodeIndex procedure will work fine with the search clause without rebuilding, using the CREATE-VECTOR-INDEX is the preferred way going forward and if you need the search-with-filter feature you have to rebuild the index using the new syntax, something like

CREATE VECTOR INDEX sectionEmbedding
FOR (m:YourLabel) ON m.yourEmbedding
WITH [m.company_id, m.role_d]

where you have to replace YourLabel and yourEmbedding with what you are using.

We have also done some improvements to the indexes so by rebuilding the index you could also get you better performance.

Hope that answers your questions!

Best regards,

Pontus