Changelog

2.1.3 (2026-10-06)

• /opt/metricshub/lib/connectors is now on the agent volume (subPath connectors): custom connectors copied there survive restarts and upgrades. The bundled connectors are copied from the image at every start, so they follow image upgrades; add your own files, do not edit the bundled ones.

2.1.2 (2026-10-05)

• README, repository index and chart description explain what the stack is for: an AI assistant in Slack that answers from live MetricsHub data, Prometheus history, the MetricsHub documentation and web search.

2.1.1 (2026-10-05)

• Architecture diagram (architecture.svg) replaces the text diagram in the README, also on the site: embedded OTel Collector pushing to Prometheus, outbound Slack and LLM connections, stage Jobs, Secrets and network rules.

2.1.0 (2026-10-03)

• M8B bot image 3.1.0 (docker.metricshub.com/metricshub/m8b-slack:3.1.0); appVersion 3.1.0.
• Knowledge-base errors name the Helm setting to use (m8b.knowledgeBase.bootstrap.mode=always) instead of kit commands.
• Documentation: GETTING-STARTED pins the chart version and checks both private images; README, KNOWLEDGE-BASE and NETWORK-POLICIES aligned with it (step numbers, Job retries, full list of allowed network flows).
• Documentation links to the MetricsHub Agent docs (Enterprise on Docker, resource settings) and to the M8B Slack bot docs (Slack app creation, configuration reference, AI backends).

2.0.0 (2026-10-02)

First release published from github.com/metricshub/helm-charts (Helm repository and GHCR OCI). Breaking for installations made with the 1.5.0 kit: resource and PVC names change.

• Agent image switched to MetricsHub Enterprise (docker.metricshub.com/metricshub-enterprise:3.9.07). The agent pushes OTLP over loopback to its embedded OpenTelemetry Collector, which exports to Prometheus and exposes 13133 (health) and 24375 (Prometheus exporter). New metricshub.otel.configText and metricshub.otel.ingressCidrs.
• Prometheus promotes OTLP resource attributes to labels (replaces the agent's append_resource_attributes).
• m8b.allowSelfSignedMcp defaults to true: the agent serves MCP with a self-signed certificate.
• Names follow Helm conventions (fullname, nameOverride, fullnameOverride); names.* removed. Selectors and labels use app.kubernetes.io/{name,instance,component} plus helm.sh/chart, managed-by, version. Local PV paths become <basePath>/<namespace>/<fullname>/<component>.
• Doctor and KB index Jobs are suffixed with the Helm revision, so re-running check/index works.
• metricshub.egress rules overlapping network.podCidrs/serviceCidrs are rejected unless excepted or network.allowClusterCIDROverlap: true.
• Removed kit-only values: namespace, releaseName, names, secrets.mode, secrets.bootstrapUser, storage.keep, storage.local.prepareOnThisHost, metricshub.bootstrap, m8b.ai.requireApiKey, m8b.doctorOnStartup, operations.
• README documents private registry access and the Enterprise/OTel architecture; examples/portable.yaml shows monitored networks and cluster CIDRs.

1.5.0

• Standard Kubernetes NetworkPolicies by default (network.provider: kubernetes); Cilium is opt-in.
• Generic CSI storage defaults; examples contain placeholders only.
• NOTES.txt prints the next step for the current stage; README documents the full staged sequence with plain Helm.