Product changelog

v25.3.6

Fixed an issue affecting COMPOSITE data source type that depleted connection pool when constructing querying database metadata concurrently from several connections and the same time.

Fixed an issue where a column defined as call_func, when invoking standard SQL functions (introduced in v25.3.2), was not correctly processed when the parameters provided were not part of the query’s projected columns. The query engine now automatically expands the projection to include the required fields before passing the query to the translator, ensuring proper execution on the remote data source without affecting the original projection requested in the command.

Fixed an issue affecting all Database datasources where the schema component of the fully qualified name (NAMEINSOURCE) was erroneously removed by the translator before being passed to the remote data source, potentially causing query errors in certain scenarios.

Fixed an issue where essential classes for Spatial operations were incorrectly removed during native image generation. Although the necessary dependencies were present, the classes were marked as unreachable and excluded from the final binary.
Added support for using SQL functions within the DDL call_func directive.

Fixed an issue that created the soft transaction row locks DB even when the feature was disabled at the application level.
Added support to mark data sources as contributors to the global health check.
Introduced the new SYSMETRICS.DataSources table, which provides data source–level health status.

Introduced agentic endpoints that enable capability propagation to any upstream agent following the expected message contract.
Fixed an issue in the Cassandra data source where Kubling SQL expressions like field = field + 1 were not correctly translated to Cassandra’s += operator for COUNTER fields.
Added support for defining properties directly in the application configuration, useful when an external properties file is not an option.
Added caching support for the Redis data source, based on user feedback.
Added support for Redis counter field types.
Added the ability to transparently define table names as part of keys in Redis, to avoid mixing up entries when using LIKE.
Added support for batch updates to the Redis data source (e.g., when performing a SELECT ... INTO).
Added importer.tableNamePattern to the Cassandra data type.
Added support for DataStax configuration file definition in the Cassandra data type.

Fixed an issue in the Cassandra Data Source that prevented SELECT queries from completing correctly when processing millions of tuples.
Fixed the cosineDistance function, which failed when applied to high-dimensional vectors (1000+).

Fixed an issue where expiration was being ignored in the Redis data source during UPDATE operations on the hash data structure.

Introduced Redis as a new Data Source type.
Introduced Cassandra as a new Data Source type.
Introduced Composite as a new Data Source type.
Introduced Routing as a new Data Source type.
Enabled the definition of additional columns in Database Data Sources.
Added query functions for basic vector operations.
Documented the TIMESTAMPADD and TIMESTAMPDIFF query functions.
Introduced an HTTP endpoint to display query plans, similar to the EXPLAIN command in other database engines.

Added support for SQL query functions, defined in modules. Configuration mechanism has changed, please check it here
Added Portable Local data source.
Enhanced clarity and detail in error messages for critical failures.

It is now possible to use a parent field in a synthetic table with the format synthetic_type 'parent', synthetic_parent_field 'another_field' in filters.
Descriptor Module Bundle now supports scheduled scripts, similar to other modules.
Resolved Out-Of-Memory (OOM) problems encountered when running certain queries on the KUBERNETES data source.
Introduced a new environment variable flag, KBL_SKIP_ENV_VAR_TEMP_FILE_ENCRYPTION, to prevent the local environment variable buffer file (used for restarting the instance) from being encrypted. This resolves issues where containers could not be stopped and restarted due to an invalid random key.
Enhanced the built-in DDL for the Kubernetes data source to cover reported use cases.

Introduced a new system schema called SYSMETRICS, which includes a table named METRICS for retrieving the engine’s metrics values.
Removed the metric kubling.engine.sql.plan.enqueued_time.max and replaced it with kubling.engine.sql.plan.enqueued.max.

Implemented the distinction between passthrough and regular access in Database data sources. Therefore, you can now extend schemas also in these type of data sources. Learn more.
Added support for parameter lists in SQL functions.
Updated function module loading (SQL and Template) to support multiple modules instead of a single one. Changes to naming resolution were introduced; read more here and here.
It is no longer necessary to pass random values for generated fields in INSERT and UPDATE operations. The Query Engine now identifies and handles these fields automatically before executing the final command.
Fixed an issue that prevented the use of non-generated fields as filters with subqueries.
Updated embedded web console to reflect the above-mentioned changes.
Official images were renamed to kubling/kubling-ce and kubling/kubling-cli.

Implemented Cache. See here.
Added metrics to monitor Cache. See here.
Implemented first-class Document functions. See here.
Engine Stream Buffer can now use local disk. See the buffer configuration option in the main application configuration.
Updated embedded web console to reflect the above-mentioned changes.

Added blankNamespaceStrategy option to the Kubernetes Data Source Configuration.
Added a new aggregator schema button in the console, under Virtual Databases.
Fixed the JavaScript filesystem cache, which in some circumstances resulted in a false cache miss when the resource had been visited before, especially during import.
Added metadata__uid in all deployment related TABLE, in the built-in Kubernetes DDL