Fixed
Full Changelog: https://github.com/giantswarm/muster/compare/v5.2.0...v5.2.1
AI Agents
Fixed
- Test framework: expectations are now evaluated by a single implementation for both step kinds, so an assertion means the same thing whether it is written on an MCP tool step or a
test_*step. The two paths each carried their own copy of the checks and had drifted three times —json_path(#1036),not_containsandwait_for_state(#1078) were each implemented on one path and silently ignored on the other, so a scenario declaring one on the wrong kind of step passed regardless of the response. Each path now only adapts its own response shape and shares the checks. Two behaviours are unified in the process:error_containsis evaluated whenever it is declared (the test-tool path previously skipped it unlesssuccess: false), and asuccess: falsepayload now fails an MCP step that expected success (previously checked only fortest_*steps). All 186 scenarios are unaffected by both. A new test asserts that every field ofTestExpectationis enforced on both paths, and fails the build when an expectation kind is added without being wired up — the class of bug behind #1038. - Test framework:
status_codein a step’sexpectedblock is now rejected at load time, with a message pointing atjson_path. It was accepted by the schema and printed by the reporter but never evaluated by either path, so a scenario asserting an HTTP status asserted nothing. No scenario in the repository used it.
- Test framework: expectations are now evaluated by a single implementation for both step kinds, so an assertion means the same thing whether it is written on an MCP tool step or a
Changed
components.muster.versionRangedefault changed from">=4.1.0 <5.0.0"to"5.x", so the meta-package resolves the current muster line (v5.1.0). The>=4.1.0floor is subsumed by the major: it existed because the muster values here setobservability.metrics.prometheus.prometheusRule, and every5.xaccepts that key. No installation needs to act. musterv5.0.0(muster#1078) is breaking only for muster instances running without their own OAuth: with no bearer there was no token family, so the transport injected thedefault-userplaceholder and every concurrent MCP client collapsed onto one session — one client’s authenticated backend became visible and callable in every other client’s session, on the first client’s token. Such sessions are now keyed by MCP transport session, which means a session ends with the connection and a reconnecting client re-runscore_auth_login. Every installation of this umbrella runsmuster.muster.oauth.server.enabled: true(the default here, overridden in no*-configsinstallation patch across the 14 deployments), so a real token-family session ID still wins and the crossing is a no-op for the fleet; stdio is unaffected either way. musterv5.1.0(muster#1079) additionally fails the render on an explicitly emptiedrbac.mcpServerEditor.subjects/rbac.workflowEditor.subjects; both are bound to the Giant Swarm admin groups by this chart (since #232) and overridden nowhere, so the guard is satisfied. The crossing also picks up muster#1077 (the chart values-schema fix), which the never-publishedv4.1.1chart left unreachable.
Fixed
- The
kagent-pgCNPG Cluster now setsinheritedMetadata.labelswithobservability.giantswarm.io/tenant: giantswarm, which the operator copies onto the PodMonitor it generates. Without it the PodMonitor carried only CNPG’s stock labels and matched neither of alloy-metrics’ discovery selectors, so it was created and silently never scraped (zerocnpg_*series in Mimir). The previouspodMonitorRelabelingsattempt is removed: it rewrote scraped series, not the PodMonitor object’s labels, so it could never make the monitor discoverable.
- The
Changed
- Point kagent wrapper-chart references at the renamed repo in #235 by @QuentinBisson
Full Changelog: https://github.com/giantswarm/agent-platform/compare/v2.11.0...v2.11.1
Added
Full Changelog: https://github.com/giantswarm/agent-platform/compare/v2.10.0...v2.11.0
Added
Full Changelog: https://github.com/giantswarm/muster/compare/v5.0.0...v5.1.0
Fixed
- Security (aggregator): a muster without inbound OAuth no longer shares one client’s downstream logins with every other client. Without a bearer token there is no token family to key a session by, so the transport layer injects the
default-userplaceholder — and that constant was used as the session key, collapsing every concurrent MCP connection onto a single session. Session-scoped state is keyed by that ID: the OAuth token store, the capability store that backs session-scoped tool visibility (ADR 006), and the session connection pool. So once any one client rancore_auth_loginagainst an OAuth-protected backend, that backend’s tools appeared inlist_toolsfor — and were callable by — every other client of the same muster, using the first client’s token. Requests that carry no authenticated identity are now keyed by their MCP transport session instead (one key per client connection, per ADR 006 §4.1); stdio reports a constant session ID, so the inherently single-user stdio transport keeps one stable session. Consequence for unauthenticated HTTP: a session ends with the connection, so a reconnecting client re-runscore_auth_loginfor its downstream backends. Enable muster’s own OAuth to get durable per-user sessions that survive reconnects. Covered by thesession-multi-user-progressive-authandsession-multi-user-tool-isolationscenarios, whose isolation assertions previously passed vacuously (see below). - Test framework:
expected.not_containsis now evaluated fortest_*steps. The scenario loader accepted the field and the regular tool path implemented it, but the test-tool path only logged a warning and passed, so every absence assertion written on atest_*step was vacuous — including the session-isolation assertions insession-multi-user-progressive-authandsession-multi-user-tool-isolation, which is why the leak above went unnoticed. It now fails the step when the response contains an unexpected string, matching the non-test-tool path.
- Security (aggregator): a muster without inbound OAuth no longer shares one client’s downstream logins with every other client. Without a bearer token there is no token family to key a session by, so the transport layer injects the
Fixed
- Helm:
values.schema.jsonno longer rejects valid chart values.podDisruptionBudget.minAvailableaccepts a percentage string ("50%") as well as an integer,podDisruptionBudget.maxUnavailableis now a declared value (the PDB template already honoured it, but the schema refused it as an unknown key), and the free-formgatewayAPI.httpRoute.{labels,annotations}/gatewayAPI.backendTrafficPolicy.{labels,annotations}maps accept arbitrary keys. Setting any of these previously failed athelm install/helm templatetime withvalues don't meet the specifications of the schema(s).
- Helm:
Added
Full Changelog: https://github.com/giantswarm/agent-platform/compare/v2.9.0...v2.10.0