Release 0.5.20 — Self-Service Integrations, Smarter Health Checks, and Reliability Fixes

Released: 2026-06-23 Chart: oci://ghcr.io/cnoe-io/charts/ai-platform-engineering:0.5.20 Previous release: 0.5.19

Highlights

0.5.20 lets team members connect their own Slack and Webex channels without involving a platform admin. Platform Health gets a meaningful upgrade — instead of a single "everything is fine" indicator, you now see the status of each dependency individually with direct links to fix what's broken. Webex works out of the box with no extra setup. Several reliability fixes land too, including one that stopped random logouts for teams running multiple replicas.

What's New

Connect your own channels — no admin required

Team members can now grant their own Slack or Webex channel access directly from the Integrations tab. Previously this required a platform admin to do it on your behalf. Admins can also onboard multiple connectors at once using the new bulk onboarding flow — select connectors, assign an agent and team, and submit them all in one step. (#1985)

Your tools stay connected when you re-authorize

When you re-consent to a connected OAuth provider (GitHub, Jira, Confluence, etc.), any tools that were already using your credentials automatically pick up the new authorization. You no longer need to manually re-link tools after refreshing your OAuth connection. (#1985)

Platform Health shows you exactly what's wrong

The Platform Health page now probes each dependency — MongoDB, Keycloak, RAG, Webex, and others — individually. If something is down, you see which service, not just a generic "unhealthy" status. Each failing item links directly to the recommended fix. (#1992)

Fix Slack routing issues directly from the UI

When Slack channel routing breaks, the admin panel now shows exactly which channels have a problem and lets you repair them with a single click — or fix all of them at once. (#1992)

Webex works out of the box

Webex space discovery and admin operations are now enabled by default. No manual configuration step is needed before Webex features start working. (#1992)

New MCP tools connect faster

When you add a new MCP server, it now uses Streamable HTTP by default — the current recommended transport that AgentGateway is optimized for. Existing servers are unaffected. (#1992)

Bug Fixes

• Random logouts in multi-replica deployments: If you run more than one UI replica, users were sometimes logged out when their requests landed on a different pod. Sessions are now shared across replicas, so this no longer happens. (#1987)
• Platform Health showed "Down" incorrectly on Kubernetes: The health check was crashing due to how Kubernetes formats some environment variables, causing the entire health widget to show as unavailable. This is fixed. (#1983)
• Fresh docker-compose installs failed at the migrations step: A first-time install using the default profiles would reach the setup wizard but fail during access-control migrations. The docker-compose configuration is corrected so the full setup flow works end-to-end. (#1907)
• Shared conversation links exposed unshared chats: Under some conditions, the shared chat link could surface conversations that were never explicitly shared. This is now scoped correctly.
• Confluence tools failed authentication: Confluence MCP tools were not passing credentials through the authorization layer correctly, causing tool calls to fail. This is fixed.

Breaking Changes

No breaking changes. Drop-in upgrade from 0.5.19.

---

Upgrade Guide: 0.5.19 → 0.5.20

Helm Values Changes

One default changed: Webex admin API is now enabled by default.

Key	Old default	New default
caipe-ui.env.WEBEX_ADMIN_API_ENABLED	"false"	"true"

If you do not have a Webex admin bot token and want to suppress the resulting health warnings, pin it off in your values.yaml:

caipe-ui:
  env:
    WEBEX_ADMIN_API_ENABLED: "false"

Otherwise no values.yaml changes are needed.

Upgrade Runbook

helm upgrade ai-platform-engineering \
  oci://ghcr.io/cnoe-io/charts/ai-platform-engineering \
  --version 0.5.20 \
  -f your-values.yaml

Verify:

kubectl get pods -n <namespace>

Full values diff (0.5.19 → 0.5.20)

-    WEBEX_ADMIN_API_ENABLED: "false"
+    WEBEX_ADMIN_API_ENABLED: "true"