You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Kubernetes resource lists load the complete requested kind and selected namespaces through compact shared caches, automatic small pages and live watches. Session-wide read budgets and bounded idle retention reduce unnecessary API traffic and memory retention.
Large All Namespaces reads stop with a request to choose specific namespaces, without presenting a partial list as complete.
Consistent refresh icons re-read lists and spin until completion. Loading shows elapsed time and received inventory payload bytes when available. Cancel replaces the list with a persistent cancelled-request placeholder; Retry or a new connection starts it again.
The cluster graph loads only after an explicit click for each connection. Selection, search and opening details preserve the camera.
Connection navigation, resource-kind labels, table tooltips, scoped command-palette navigation and log selection have been refined. Re-rendering selected-resource ports no longer repeats the same information request.
Local inventory and memory measurements, raw samples, methodology and their limits are documented in docs/performance.md; they are not universal production-performance guarantees.
Resource lists avoid repeated JSON decoding, reducing CPU work while loading large Kubernetes inventories.