DIAL Extension chart configuration

The dial/dial-extension chart deploys a container as a Kubernetes Deployment and Service, with optional routing, storage, scaling, and observability resources. It is used for adapters, applications, Chat, themes, and Admin frontend/manager aliases. This reference describes chart 3.1.1 for platform engineers preparing workload values.

The release's README parameter table and values.yaml list every nested field. The tables below group those fields by the resource they control. When used under another chart, put values below the alias, such as openai, frontend, or manager.

Image and runtime defaults

Value	Default	Purpose
image.registry, image.repository, image.tag	docker.io, empty, latest	Set a repository and pin the image tag or digest for the chosen service.
image.digest, .pullPolicy, .pullSecrets	Empty, IfNotPresent, []	Digest and registry pull behavior.
replicaCount	1	Replicas before optional HPA control.
containerPorts.http, containerPorts.metrics	5000, 9464	Container ports; override for services using different ports.
resourcesPreset	nano	Common-library CPU/memory preset; none disables the preset.
resources	{}	Explicit requests/limits; when nonempty, takes precedence over resourcesPreset.
service.type, service.ports.http	ClusterIP, 80	Service exposed inside the cluster.
livenessProbe.enabled, readinessProbe.enabled, startupProbe.enabled	false, false, false	Enable the appropriate health checks. Default probe paths are /health.
containerSecurityContext.readOnlyRootFilesystem	true	Container filesystem policy. Add writable mounts where the image needs them.

Shared workload settings

These setting families also appear in the Core chart and Admin Backend workload, with chart-specific defaults. The Extension chart is the source of values for its dependency aliases.

Value family	Default / behavior
global.imageRegistry, .imagePullSecrets, .storageClass	Shared registry, registry Secrets, and storage class overrides.
global.compatibility.openshift.adaptSecurityContext	disabled; OpenShift security-context adaptation.
nameOverride, fullnameOverride, namespaceOverride	Empty; resource naming and namespace overrides.
commonLabels, commonAnnotations, podLabels, podAnnotations	{}; shared and pod-specific metadata.
extraDeploy	[]; additional manifests.
diagnosticMode.enabled	false; workload diagnostics mode.
command, args	[]; container startup overrides.
env, secrets, extraEnvVars, extraEnvVarsCM, extraEnvVarsSecret	Plain env, generated Secret, native env entries, and existing configuration references. See Configuration sources.
extraContainerPorts	[]; additional container ports.
livenessProbe, readinessProbe, startupProbe	Native probe fields and timing; enabled gates each generated probe.
customLivenessProbe, customReadinessProbe, customStartupProbe	{}; custom probe definitions.
podSecurityContext	Enabled; group 1001 with filesystem group policy Always, plus configurable sysctls/supplemental groups.
containerSecurityContext	Enabled; UID/GID 1001, non-root, no privilege escalation, drop all capabilities, RuntimeDefault seccomp. The Extension default root filesystem is read-only.
automountServiceAccountToken	false; pod token mounting.
serviceAccount.create, .name, .annotations, .automountServiceAccountToken	true, empty, {}, false; create or select a service account.
hostAliases, hostNetwork	[], false; pod host resolution/networking.
nodeSelector, affinity, tolerations	Empty; Kubernetes scheduling constraints.
topologySpreadConstraints, priorityClassName, schedulerName	Empty; topology, priority, and scheduler choices.
updateStrategy	RollingUpdate.
terminationGracePeriodSeconds, lifecycleHooks	Empty / {}; shutdown timing and hooks.
pdb.create, .minAvailable, .maxUnavailable	Disabled, empty availability bounds.
autoscaling.hpa	Disabled; default replica bounds 1–3. Configure CPU/memory targets, custom rules, annotations, and behavior as needed.
extraVolumes, extraVolumeMounts, sidecars, initContainers	[]; pod composition and mounts.

Service and external routing

Value family	Purpose
service	Type, HTTP port/node port, ClusterIP, load-balancer IP/source ranges, traffic policy, annotations, extra ports, and session affinity.
ingress.enabled	false by default.
ingress.ingressClassName, .hosts, .path, .pathType	Controller class, hostnames, path (/), and path type (Prefix).
ingress.tls, .annotations, .serviceName, .extraPaths, .extraRules	TLS and additional routing controls.
httpRoute.enabled	false by default; generates a Gateway API HTTPRoute.
httpRoute.parentRefs, .hostnames, .rules, .annotations	Attach to existing Gateways and configure HTTP routing.

Creating an Ingress or HTTPRoute does not install a controller, Gateway, or TLS issuer. Select routes and backend ports appropriate for the application. Prefer one intended exposure path for each workload.

Storage

Value	Default	Purpose
persistence.enabled	false	Attach persistent storage.
persistence.mountPath, .size, .accessModes	/data, 8Gi, [ReadWriteOnce]	Volume mount and claim defaults.
persistence.existingClaim, .storageClass, .subPath	Empty	Existing claim, storage class, and mount subpath.
persistence.annotations, .dataSource, .selector	{}	Claim metadata, source, and selection.
temporary.mountPath	/tmp	Temporary volume mount.
temporary.subPath, .sizeLimit, .medium	Empty	Temporary-volume controls.

The container must use the selected mount path; chart persistence does not alter application configuration. Review claim access modes before scaling a workload that mounts a volume.

Network policy

networkPolicy.enabled defaults to true in Extension chart 3.1.1. Both allowExternal and allowExternalEgress default to true. Configure the policy for the intended peers, and use a cluster network implementation that enforces NetworkPolicy.

extraIngress and extraEgress add rules. ingressPodMatchLabels, ingressNSMatchLabels, and ingressNSPodMatchLabels select allowed peers. networkPolicy.metrics configures metrics ingress separately, including its allowExternal and namespace/pod selectors.

Metrics

Value family	Default / behavior
metrics.enabled	false; creates the dedicated metrics Service.
metrics.service	ClusterIP, HTTP port 9464, plus the same exposure/session settings as the main Service.
metrics.serviceMonitor.enabled	false; requires the metrics Service and Prometheus Operator CRDs.
metrics.serviceMonitor.port, .path	http-metrics, /metrics.
metrics.serviceMonitor.namespace, .annotations, .labels, .selector	Resource placement and discovery metadata.
metrics.serviceMonitor.interval, .scrapeTimeout, .jobLabel, .honorLabels	Scrape timing and labels; honorLabels defaults to false.
metrics.serviceMonitor.metricRelabelings, .relabelings	Scrape relabeling rules.
metrics.prometheusRule.enabled	false; creates a PrometheusRule when enabled.
metrics.prometheusRule.namespace, .annotations, .labels, .rules	Rule placement, metadata, and expressions.

The application must actually expose metrics on that port/path. For Deployment Manager, for example, use /actuator/prometheus rather than the chart's /metrics default. Exporter configuration belongs to the deployed service.

Example: DIAL Model Provider workload

image:
  repository: epam/ai-dial-model-provider
  tag: 0.1.0-rc.0
containerPorts:
  http: 5000
env:
  DIAL_URL: http://dial-core.dial.svc.cluster.local
resources:
  requests:
    cpu: 100m
    memory: 256Mi
  limits:
    cpu: "1"
    memory: 1Gi

This example renders one DIAL Model Provider workload, which serves all its adapters. Core still needs model configuration pointing to its Service, and provider credentials must be configured as described in DIAL Model Provider configuration. For an application image, use Custom Apps deployment.

Further reading

• Platform chart — use these values below an assembly alias
• Admin chart — frontend and manager aliases
• Deploy Custom Apps to a cluster — run a Custom App as a Kubernetes service beside Core
• Install the Quick Apps feature — deploy the Quick Apps backend and register it in Core
• DIAL Model Provider — the service that runs the OpenAI, Bedrock, and Vertex AI adapters