October 2026 Summaries

Cube Core 1.8 introduces stricter, safer defaults and performance improvements, with development mode now explicitly controlled by CUBEJS_DEV_MODE, query values redacted from production logs, and stronger validation for data models, requests, time zones, and access policies. The release includes breaking changes such as removal of the TypeScript-based cube.ts configuration file, compilation errors for empty member-level access policies and duplicate joins to the same cube, and an Angular 20 requirement for the Angular client library. SQL API enhancements include data-source pushdown for compatible UNION queries, window and aggregate functions, COPY FROM STDIN support, and faster streaming results, which are reported as up to 1.76 times faster for 10-million-row queries. Data modeling gains memo() for caching asynchronous results across compilation stages, while expanded source support adds query attribution, driver improvements, cancellation support, and additional database compatibility. Cube Store receives default-enabled performance toggles, encryption and operational controls, resource limits, and lower memory use, while multi-tenant deployments can share compiled model state to reduce memory consumption. The release also deprecates the legacy Cube CLI and Hive driver, updates its Docker runtime to Node.js 24.21.0, and adds Python 3.14 support for native Python and Jinja data models.

Cube has introduced Apache Ossie integration, allowing users to export Cube data models or individual views into vendor-neutral Ossie YAML files and import Ossie models back into Cube as generated cubes and views. Available through Cube CLI version 1.8.3 or later and the Platform API, the feature uses Cube’s open-source converter maintained in the Apache Ossie repository, supporting local, cloud, automated, and offline conversion workflows. Apache Ossie, formerly Open Semantic Interchange, is an Apache-backed standard designed to reduce manual rebuilding of semantic models when organizations move between BI tools by representing datasets, relationships, metrics, AI context, and extensions in a common format. Cube maps most model elements directly, while retaining Cube-specific features such as pre-aggregations, segments, policies, and formats through custom extensions to support round trips. The converter identifies potential join fan-out problems that could cause incorrect metric totals in other tools, safely rewrites applicable metrics such as counts, and issues warnings or errors for unsafe or unrepresentable cases. Single-view exports are stricter and expose only the selected view’s members and dependencies, while conversion reports classify outcomes as errors, warnings, or informational changes. Conversion runs in an isolated Cube environment using only model files, does not access warehouse credentials or data, uses no AI, and currently supports YAML models rather than JavaScript or Jinja files.