Release Notes v0.6.0

What's Changed

feat: K8s node host agents — Telegraf (metrics) and Fluent Bit (logs) installed per node via privileged Jobs, with per-node metric selection by @ish-hcc in #327, #332
feat: Extend Insight, Alert, and Log to Kubernetes clusters and nodes by @ish-hcc in #329
fix: Resolve exec-plugin kubeconfigs (EKS/GKE/NKS) via a cb-tumblebug token broker, schedule Jobs onto the real node (instance-id → node name), and install the log agent as a Fluent Bit host binary so it works on EKS Amazon Linux nodes by @ish-hcc in #334, #340, #342
feat: Persist K8s node agent task state with a per-node lock so install/uninstall is consistent and shown across browser tabs by @ish-hcc in #348
fix: K8s cluster-list cache with a background warmer so the NS/Infra overview never pays the cold cb-spider discovery cost by @ish-hcc in #332, #345

feat: Add distributed tracing (Grafana Tempo) and rename MCI/VM to Infra/Node throughout by @ish-hcc in #302
feat: Windows OTel Java auto-instrumentation and Beyla-based tracing by @basinibi in #294, #274, #275, #276
feat: Add paging to the Tracing trace list by @ish-hcc in #308

feat: New React-based observability frontend by @ish-hcc in #278
feat: Namespace-scoped iframe UI for embedding consoles — cross-origin parent-namespace detection via postMessage, land on the Monitoring view, and remember the last selected namespace/section by @ish-hcc in #326, #335, #336, #337, #338, #339
feat: Make the Insight menu usable — anomaly detection, prediction, and server error analysis by @ish-hcc in #307, #300
fix: Display timestamps in the browser's local timezone and show point markers for sparse metric samples by @ish-hcc in #319, #314

feat: Editable trigger policies — thresholds and settings after creation, multi-channel notifications by @ish-hcc in #318, #320
feat: Discord and Teams notification channels by @tryoo0607 in #298
feat: Kakao AlimTalk notifications via NCP-registered templates by @tryoo0607 in #312, #313, #324

feat: GPU monitoring with DCGM Exporter by @ish-hcc in #299
feat: OpenAI-compatible LLM provider and base-URL support for Insight by @50thk in #323

feat: Caffeine-based monitoring metric + CSP cache with periodic, portal-range-aligned warming by @ish-hcc in #270, #271, #281, #287, #273
feat: Downsampling DB support across multiple InfluxDB instances for long-range queries by @ish-hcc in #264, #273

refactor: Route all infrastructure calls through cb-tumblebug only (no direct cb-spider) and align with the v0.12.x Infra/Node model by @basinibi @ish-hcc in #297, #291, #302
feat: Update cb-tumblebug / cb-spider across the v0.12.x line by @ish-hcc in #265, #288, #292, #295

fix: Harden VM/K8s agent install and status — node-id registration, cross-tab task guard, status caching, Azure AKS node-name mapping, and a 63-char Job-name limit; drop GROUP BY time() when the projection has no aggregate (empty graphs); bypass the apt sandbox so the K8s log agent installs on hardened nodes by @ish-hcc in #346, #347
fix: Surface backend-unreachable errors in the overview with a Retry instead of a silent empty state by @ish-hcc in #333
fix: Precise CSP "unsupported" labelling, persistent infra selector, and no agent buttons on non-running VMs by @ish-hcc in #341, #343
fix: Auto-create InfluxDB databases and reset stuck agent task status on manager startup by @ish-hcc in #303, #305
fix: Auto-derive VM-reachable host for Fluent Bit (Loki) and Beyla (Tempo) agents by @ish-hcc in #310

Full Changelog: v0.5.0...v0.6.0