Skip to main content

Release 0.5.25 — Platform Health Overhaul, Sharing Fixes, and Webex Routing Simplification

Sri Aradhyula
Sri Aradhyula

Released: 2026-06-26 Chart: oci://ghcr.io/cnoe-io/charts/ai-platform-engineering:0.5.25 Previous release: 0.5.24

Highlights​

0.5.25 is a dense release. Platform Health is rebuilt around a capabilities model — each probe is now profile-aware and reports integration status independently. Chat sharing is restored after a regression. Webex space onboarding and direct routing are simplified. Slack channel health auditing is batched for speed. Several UI quality fixes land around sharing, knowledge base ownership, and session auth caching.

What's New​

Platform Health rebuilt as profile-aware capabilities​

Platform Health probes are now structured as named capabilities tied to the deployment profile. Each capability reports its status independently, and integrations that are not part of the active profile are shown as "not applicable" rather than failing. This makes the health view meaningful for partial deployments and removes false positives. (#2005)

Webex space onboarding and direct routing simplified​

The Webex space onboarding flow is streamlined and direct message routing is simplified. Fewer configuration steps are needed to get a Webex space routing to an agent, and the logic handling direct (non-space) message delivery is cleaner. (#2032)

Bug Fixes​

  • Chat share visibility broken: User-level and team-level conversation sharing was not displaying correctly after recent changes. Share visibility is restored. (#2031)
  • MCP tool schema and health scans unstable: Tool schema parsing was not handling edge cases in some MCP server responses, causing health scans to report false errors. Schema parsing and scan logic are now more tolerant. (#2033)
  • MCP seedConfig source ignored: The seedConfig source for MCP server configuration was not being used — the UI was reading from a different source. This is corrected so seeded configuration is applied as expected. (#2042)
  • Admin user team lists did not scroll for long lists: When a user belonged to many teams, the team list in the admin user view overflowed the panel without scrolling. Long lists now collapse. (#2034)
  • Keycloak OOM under Slack admin load: Keycloak was running out of memory when the Slack admin panel performed large user listing operations. Memory headroom is added to the Keycloak chart defaults. (#2038)
  • Session auth re-checked on every API call: The session authorization check was running a full validation on every API request rather than caching the result within a session. This caused unnecessary latency and load. Auth results are now cached for the session lifetime. (#2039)
  • Slack channel health audit too slow: The configured-channel health audit was scanning channels one at a time. It now batches lookups, significantly reducing the time to surface channel health status for large Slack deployments. (#2043)
  • Sharing and transfer UX inconsistencies: Several edge cases in the conversation sharing and knowledge base transfer UI were fixed, including ownership consolidation for knowledge bases shared across multiple users. (#2041)

Breaking Changes​

No breaking changes. Drop-in upgrade from 0.5.24.

Upgrade Guide: 0.5.24 → 0.5.25​

Helm Values Changes​

No required values changes. The Keycloak memory default is increased — if you have pinned Keycloak memory limits below the new default in your values.yaml, review and adjust if needed.

Upgrade Runbook​

helm upgrade ai-platform-engineering \
  oci://ghcr.io/cnoe-io/charts/ai-platform-engineering \
  --version 0.5.25 \
  -f your-values.yaml

Verify:

kubectl get pods -n <namespace>