Skip to content

Release 6.7

May 2026

Kelvin API

What's New

Asset Insights & Asset Status

  • Added Asset Status Indicator support.
  • Added asset status indicator configuration to asset types.
  • Added calculation and inclusion of asset status indicators in asset endpoints.
  • Migrated Asset Insights API to ClickHouse.
  • Added a job queue for Asset Insights time series backfills.
  • Added stale job cleanup loop for Asset Insights workers.
  • Added ClickHouse paginator provider with:
    • Keyset cursor pagination
    • Null positioning
    • Post actions
    • Processed query exposure
  • Added ClickHouse query utilities and expression builders.
  • Added the latest values view improvements for time series.

Apps / Workloads

  • Added app image management support.
  • Added a new workload flag (enabled).
  • Added data tags support to the app and workload manifests.
  • Added workload/app runtime schema synchronization.

Data Tags

  • Added the created_since filter to the datatags list endpoint.

Control Change Policies

  • Added Control Change Policy APIs.
  • Added cluster sync policies endpoint.
  • Added control change policy resource existence validation.
  • Added resource data type validation for CC policies.
  • Added control change expiration delta support.
  • Added SSL certificates and edge-k8s client configuration for Control Change Manager.
  • Added synchronization of control change policies to edge clusters.

Core / Infrastructure

  • Added ClickHouse data qualities importer.
  • Added GitHub workflows and AI skill tooling.
  • Added codex/copilot/Claude skills and repo tooling documentation.

What's Changed

Asset Insights

  • Improved Asset Insights performance:
    • Early filtering
    • Reduced memory usage
    • Separated enrichment from filtering
    • Removed unnecessary ordering
  • Refactored time series backfill strategy:
    • Adaptive chunking
    • Weekly chunking
    • Recursive splitting on OvercommitTracker
    • Partition iteration
  • Refactored extra field handling into post-actions.
  • Missing related resources now return null instead of errors.

ClickHouse & Importers

  • Upgraded ClickHouse Go library.
  • Improved importer performance with streamed imports.
  • Refactored importer migrations to use the ClickHouse migrator.
  • Import schedules are now loaded on demand.
  • Importers now rebuild through event consumers and job queues.
  • Improved race-condition handling across importers.

Apps / Workloads

  • Unified app status calculation.
  • Filtered empty workload states.
  • Build status fields are now preserved across state transitions.
  • Workload updates now correctly refresh row_updated.
  • Dynamic IO apps no longer allow map_to.

Data Tags

  • Improved worker performance with indexes.
  • Removed timestamp download buffer logic.
  • Seeded default model configuration values.
  • Improved model session handling and worker reliability.

API / Platform

  • Refactored template engine to internal template functions.
  • Refactored health checks and service architecture.
  • Improved cloud spec descriptions and payloads.
  • Improved policy-related error messages.
  • Improved build worker logging.
  • Refactored the status indicator worker with dirty-set architecture.

CI / Dependencies

  • Updated vulnerable dependencies and Python packages.
  • Updated external container images and Alpine base version.
  • Added deployment skip options for alpha/beta.
  • Updated GitHub Actions versions and Slack notifications.
  • Improved git hooks compatibility and workflow tooling.

What's Deleted

  • Removed has and !has operators from the asset status indicator configuration.
  • Removed Asset Insights RPC health check from API Instance (replaced with HTTP health check).
  • Removed staged filter from workload queries.
  • Removed median aggregate state from Asset Insights.
  • Removed service account source constraint from Create Data Tag endpoint.
  • Removed previous ClickHouse Nullable(JSON) workaround.

What's Deprecated

  • Older importer scheduling model deprecated in favor of on-demand imports.
  • Legacy backfill strategies are deprecated in favor of adaptive chunking architecture.
  • Older synchronous importer rebuild patterns replaced with queue-driven rebuild workflows.
  • Older asset enrichment query patterns are deprecated in favor of the post-action enrichment flow.

Whats New in our Documentation ?

This release is a big one. The Kelvin UI sidebar has been completely restructured into four sections — Operations, Engines, Connectors, and Administration — and the documentation has been updated throughout to match. If things look different when you navigate, that's why, and every page across Operations Tools, Developer Tools, and Platform Administration now reflects the new layout.

On top of that, the Kelvin Academy launches for the first time. It's the new home for onboarding and training, replacing the old Use Cases and Quick Start guides. Courses cover both Operations and Developer tracks, and it will keep growing across future releases.

In the Overview section a new Engines concept page has also been added to help understand this major concept change. Many other pages also have been updated to cover asset scheduling, data tags, asset types, manual controls, and more.

What's New

What's Updated

What's Deleted

  • None

Others

  • The documentation site has moved to the Zensical engine. Zensical ships a brand new search engine built from scratch — client-side, privacy-friendly, no third-party services, and it works offline too. Fast results, better relevance.
  • All screenshots have been updated to reflect the new Kelvin UI look and feel.
  • The Changelog has been reorganized and moved into the Overview section.
  • The overview menu layout has been improved into clear sections; Components, Infrastructure, Intelligence and For Developers.