Reference
Helm values
The chart values most installs touch, grouped by purpose, with their defaults.
This page lists the telark chart values
(oci://ghcr.io/telark/charts/telark) most installs touch. The complete list is
charts/telark/VALUES.md,
generated from the chart; the
chart README
explains each one.
Set values with --set key=value or -f values.yaml. Helm does not
remember --set flags: pass them again on every helm upgrade, or keep
them in a values file.
| Key | Default | Notes |
|---|
app.auth.bootstrap.admin | "" | The one bootstrap admin's email, enrolled and recovered with break-glass --enroll, and the only account that can turn on Google sign-in or self-registration. The render fails while it is empty. Once that account's first passkey pins it, a change only lets break-glass open a replacement request (Recovery requests). |
| Key | Default | Notes |
|---|
app.name | telark | Prefix of every resource name. The API group stays telark.io. |
app.namespace | telark | Must match the release namespace (-n). |
app.mode | standard | minimal, standard or performance: replicas, resources, rate limits and disruption budgets of every Telark service. Subcharts keep their own sizing. |
crds.enabled | true | Installs the telark-crds subchart. Set false to manage CRDs out of band. |
app.selfMonitoring.enabled | false | true discovers Telark's own namespace and shows its components as applications. See Self-monitoring. |
app.image.registry | ghcr.io/telark | Registry hosting the service images. |
app.image.pullPolicy | Always | Pull policy of every service container. |
app.image.pullSecrets, global.imagePullSecrets | [] | Only for a private registry or mirror. |
app.image.kubectl | registry.k8s.io/kubectl:v1.37.1 | Pinned by digest. Runs the uninstall hooks. Mirror it when air-gapped. |
| Key | Default | Notes |
|---|
app.auth.passkey.id | "" | WebAuthn relying-party ID. Empty follows the request host. Required with ingress.enabled or gateway.enabled. |
app.auth.passkey.origin | "" | Allowed origins, comma-separated. Empty follows the request Origin. Required with ingress.enabled or gateway.enabled. It also names the session cookie: __Host-telark_session when every entry starts with https://, else telark_session, passed to the services as SESSION_COOKIE_HOST_PREFIX. |
app.auth.passkey.name | Dashboard App | Name shown in the browser's passkey prompt. |
app.auth.oidc.existingSecret | "" | A Secret you manage (key googleJwkJson) for the pinned Google keys, for cluster-less renders. |
services.auth.env.ENROLL_INVITE_TTL_SEC | 3600 | Lifetime, in seconds, of an enrollment link created from Members. Break-glass links always live 10 minutes. See Enrollment links. |
The Google client ID, the Google switch and self-registration are
runtime settings on TelarkConfig, changed in Settings by the
bootstrap admin, not chart values.
| Key | Default | Notes |
|---|
ingress.enabled | false | Ingress to the dashboard. |
ingress.className, ingress.host, ingress.tls, ingress.annotations | "", "", [], {} | Controller class, hostname and TLS. |
gateway.enabled | false | Gateway API HTTPRoute instead of an Ingress. |
gateway.parentRefs, gateway.hostnames | [], [] | Gateway listeners to attach to, and hostnames. |
Without either, reach the dashboard with
kubectl port-forward -n telark svc/telark-ui-service 3000:8080.
| Key | Default | Notes |
|---|
kyverno.enabled | true | Installs the Kyverno subchart, which enforces protection plans. false uses a Kyverno you already run: 1.18 or later, with the policies.kyverno.io CRDs. |
kyverno.features.forceFailurePolicyIgnore.enabled | true | While the bundled Kyverno is down, admission lets requests through, enforce plans included. |
kyverno.admissionController.replicas | 2 | With a disruption budget of minAvailable: 1. |
| Key | Default | Notes |
|---|
ollama.enabled | true | Installs Ollama, the local model runtime for Insights. |
app.ollama.autoPull | true | Lets the analyzer download a missing model; false for air-gapped installs (no runtime egress). |
app.ollama.runtimeUrl | "" | An Ollama-API endpoint you run. Empty uses the subchart. |
ollama.resources | requests 250m and 1536Mi, limit 2 CPU | One size for every mode. See Insights hardware profiles. |
ollama.persistentVolume.size | 6Gi | Model volume, kept on uninstall: the default model plus any one catalog model, since a model switch deletes the other catalog models. 20Gi to try 8B models. Never shrinks: see Upgrading. |
services.analyzer.env.ANALYZER_MODE | fast | deep for the tool-calling mode. |
| Key | Default | Notes |
|---|
app.persistence.enabled | true | Provisions the exporter's snapshot and report volumes, kept on uninstall. |
app.persistence.storageClass | "" | Class of both exporter volumes. "" is the cluster default on install and keeps the claims' current class on upgrade; a different class copies the data to new claims. See Exporter storage. "-" disables provisioning. |
services.exporter.replicas | 1 | More than one shares the volumes and needs a ReadWriteMany app.persistence.storageClass. minimal always runs one. |
app.persistence.snapshotsSize | 512Mi | Snapshot volume (performance: 2Gi). Only grows, where the class allows expansion. |
app.persistence.reportsSize | 512Mi | Report volume (performance: 2Gi). Only grows. |
redis.master.persistence.size | 2Gi | Redis volume. Fixed once installed: see Upgrading. |
nats.persistence.size | 1Gi | NATS JetStream volume. Fixed once installed: see Upgrading. |
| Key | Default | Notes |
|---|
app.security.crdGuard.enabled, app.security.crdGuard.enforce | true, true | Only Telark's own service accounts may write telark.io resources and the OIDC trust Secret. enforce: false only audits. |
app.security.crdGuard.extraAllowedUsers | [] | Break-glass usernames also allowed to write them. |
app.security.breakGlass.waitHours | 24 | Hours, at least 1, with no sign-in by the bootstrap account after which a break-glass recovery or replacement request completes. See Recovery requests. |
app.security.networkPolicy.enabled | true | Ingress NetworkPolicies for Telark pods and NATS. Needs a CNI that enforces them. |
app.security.serviceToken.value, app.security.serviceToken.existingSecret | "", "" | Shared service token. Empty generates one on install and reads it back on upgrade; set existingSecret for cluster-less renders. |
nats.existingSecrets.publisher, nats.existingSecrets.consumer | "", "" | Your own NATS user Secrets, for cluster-less renders. |
redis.auth.existingSecret | <release>-redis-secret | The Redis password Secret (key redis-password). Generated on install and read back on upgrade; set your own for cluster-less renders. Redis always requires the password. |
| Key | Default | Notes |
|---|
app.webhooks.targets | [] | Endpoints the notifier sends signed events to: name, url, secret, events, optional previousSecret and allowPrivate. Rendered into the Secret telark-webhooks. See Webhooks. |
app.webhooks.existingSecret | "" | A Secret you manage (key targets.json) holding the same list as JSON, to keep targets out of values files and for cluster-less renders. |
| Key | Default | Notes |
|---|
app.serviceDefaults.autoscaling.enabled | true | CPU autoscaling of every service except the exporter (minimal: off). |
app.serviceDefaults.autoscaling.minReplicas, maxReplicas | 1, 3 | performance: up to 5. |
app.serviceDefaults.autoscaling.targetCPUUtilizationPercentage | 80 | |
vpa.enabled | false | Installs the VPA operator and one VPA per service. |
metrics-server.enabled | true | Needed for usage-based Insights rules and autoscaling. Set false if the cluster already runs one. |
monitoring.serviceMonitor.enabled | false | Prometheus-Operator ServiceMonitor. No Telark service serves /metrics, so keep it off: it would scrape nothing. |
monitoring.serviceMonitor.labels | {} | Must match your Prometheus serviceMonitorSelector. |
Each service has a services.<svc> block (exporter, discovery,
analyzer, notifier, auth, ui) with enabled, replicas,
nodeSelector, tolerations, affinity and env. Tunables go in
services.<svc>.env; see Environment variables.
Resources come from the shared app.shared.resources block (requests
100m and 128Mi, limits 1000m and 512Mi), resized by app.mode.
Its ephemeral storage (request 50Mi, limit 2Gi) is the same in every
mode.
| Subchart | Version |
|---|
| Kyverno | 3.9.2 |
| Redis (Bitnami) | 23.0.10 |
| NATS (Bitnami) | 9.0.28 |
| Ollama | 1.50.0 |
| metrics-server | 3.12.2 |
| VPA (Fairwinds) | 5.1.0 |
The chart requires Kubernetes 1.30 or later.