← Back to Changelog
Header image for changelog: August 2025 | Product releases
31st August 2025

August 2025 | Product releases

This release brings GPUs front and center with powerful new workload support across in PaaS for A100s, H100s, B200s and more, alongside bring-your-own registry for container builds, expanded logging and addon options, and smarter scheduling.

We’ve introduced the Northflank AI Co-Pilot to help teams ship faster, streamlined build and template enhancements, and delivered dozens of fixes and improvements across infrastructure, workflows, and developer experience, all to keep scaling on Northflank smoother and more powerful than ever.

🏗 Infrastructure & workloads

Added

  • GPU workloads on PaaS: Run AI inference and training on A100s, H100s, B200s and other GPU types at competitive pricing. Includes stack templates, workload scheduling, and plan selection improvements.
  • Push Northflank container builds directly to your own private registry.
  • Northflank hosted Loki can now store logs in GCP storage as well as S3.
  • Northflank Organisation BYOC clusters now support wildcard domains.
  • Added support for snapshot thresholds on addons.
  • Support for custom vnet subnets and Cilium overlay mode on Azure clusters.
  • Global Secrets available at the team level; can be inherited and combined in templates and pulled into projects.
  • Organisation-level API tokens can now be created from existing RBAC roles.
  • PostgreSQL addons now support the h3-pg and pg_partman extensions.

Enhancements

  • GPU scheduling logic improved for higher reliability.
  • Horizontal pod autoscaling improved for custom addon types.
  • Metrics added for Bring Your Own Kubernetes (BYOK).
  • Addon disk size limit raised to 1.5TB.
  • More robust handling of PostgreSQL permissions on startup.
  • Improved MySQL HA metrics representation for decimal values.
  • Ceph cluster config options now populate correctly on page load.
  • AWS cluster validation ensures nodes have enough pod capacity before creation.
  • Crontab validation now supports dynamic template arguments.
  • Pod termination dates recalculated correctly after termination events.
  • GCP provisioning error messages now display all errors if multiple occur.
  • Billing system updated: GPU usage requires pre-purchased credits, grace periods are more dynamic.

Fixed

  • Fixed filtering of GPU plans in workload creation.
  • GPU stack templates now correctly filter regions/clusters based on GPU support.
  • Fixed addon storage overrides when using ref values in templates.
  • Fixed volume creation issues in clusters/regions without multi-RW support.
  • Fixed SSO user login issues during first sign-in.
  • Fixed API tokens being revoked when org roles were modified.
  • Fixed race condition that could crash service dashboard on restart.
  • Fixed PostgreSQL edge case where permissions were reset on startup.

🧩 Templates & workflow

Added

  • Added support for Buildkit build secrets using secret mounts in Northflank builds.
  • Added support for custom addon Helm values in YAML.
  • New workflow loop node in templates, useful for generating multiple nodes dynamically (e.g. with secrets).

Enhancements

  • Initial stack template configuration UX refined for smoother setup.
  • Template improvements: easier initial creation, draft name/description validation less restrictive, and commit hashes now shown correctly.

Fixed

  • Template list no longer flashes on load and search works reliably.
  • Template count in projects now shows correct values.

👩‍💻 Developer experience

Added

  • New command menu for faster navigation across resources.
  • Added command exec support in CLI and js-client for custom addon types (BYOA - Bring your own Addon).
  • Introduced AI Copilot Assistant — ask questions about Northflank primitives and platform usage.
  • Team dashboard, BYOC cluster info, and subdomain list redesigned for clarity.

Enhancements

  • Build logs improved: cache promotion logic accounts for cache misses, log viewer displays all builds, and BuildKit progress is now shown step-by-step.
  • Cluster/node views updated: node pool list now shows architecture, deletion modals clearer, new buttons for cordoning/draining nodes.
  • Networking forms: domain add form has better descriptions; subdomain verification warns when Cloudflare proxy may interfere.
  • Billing page restructured for new credit system and incremental billing.
  • Faster fetching of repos, branches, and PRs from VCS providers.
  • Observe pages now scale better with larger pod counts.
  • Improved validation performance for DockerHub images.
  • General UI polish: sliding tabs highlight active tab, buttons and selectors behave consistently across screens, responsive improvements across many components.

Fixed

  • White-labelling now shows the correct URL in GitHub deployments.
  • Combined Service fields no longer highlight without input.
  • Resource headers no longer shift when hovering.
  • Fixed display glitches in node affinity tags and volume performance metrics.
  • Password managers (e.g. Bitwarden) now autofill update password prompts correctly.
  • Fixed crashes on node pool form and build options page for legacy services.
  • Fixed log line sharing to use correct timestamp.
  • Fixed subdomain path service selector so project names can be searched.
  • Fixed project list hitting internal rate limits when listing many projects.
  • Fixed display of team/org resource quotas (now has its own page).
  • Audit logs now display deleted users correctly.

See you next month!

Share this article with your network
X