Agent Mesh
Go a2aa2a-protocolagent-meshagentic-aiinferencemcpmeshmesh-networknetwork-securitynetworking-aip2p-network
image:
repository: agentmesh-node
tag: local
pullPolicy: IfNotPresent
replicaCount: 1
# Required: control plane URL the node enrolls with,
# e.g. http://agentmesh-control-plane:8080. A plaintext http:// URL makes the
# chart pass --insecure-control-plane: whoever answers it becomes the node's
# trust root, so use https:// for anything beyond the cluster network.
controlPlaneUrl: ""
# Audience of the projected ServiceAccount token; must be one of the control
# plane's allowedAudiences.
audience: agentmesh-audience
# Bearer token protecting the node's local REST API. Leave empty to generate a
# random one on first install (kept stable across upgrades) in the Secret
# <release>-api-token; read it with
# kubectl get secret <release>-api-token -o jsonpath='{.data.api-token}' | base64 -d
# Set explicitly (e.g. via --set) to pin a known value. The generated form
# relies on Helm's lookup, which sees no cluster under `helm template`,
# `--dry-run` or a GitOps diff: those render a fresh value each time. Pin the
# token, or pre-create the Secret, when the manifests are rendered outside
# `helm install`/`upgrade`.
apiToken: ""
# Pod and container security contexts. The image is distroless nonroot;
# runAsNonRoot needs the numeric uid.
podSecurityContext:
runAsNonRoot: true
runAsUser: 65532
runAsGroup: 65532
seccompProfile:
type: RuntimeDefault
securityContext:
allowPrivilegeEscalation: false
capabilities:
drop: ["ALL"]
# Address the node API binds inside the pod. Loopback keeps the API
# pod-private; set 0.0.0.0:8080 to expose it on the pod IP.
bindAddr: "127.0.0.1:8080"
# Plain HTTP port for /metrics, /healthz and /readyz, separate from bindAddr
# so neither the scraper nor the kubelet needs the API token. Unauthenticated,
# so it is only ever a containerPort. Set to 0 to disable it and the probes.
metricsPort: 9090
# Scraping of the metrics port. Each is a CRD the cluster may not have, so
# both are off until you pick the collector you run.
monitoring:
interval: 30s
# Google Managed Prometheus (monitoring.googleapis.com/v1 PodMonitoring),
# present by default on GKE.
podMonitoring:
enabled: false
# prometheus-operator (monitoring.coreos.com/v1 PodMonitor).
podMonitor:
enabled: false
# Extra labels, e.g. the release selector a Prometheus instance matches on.
labels: {}
# Extra agentmesh-node args, e.g. ["--discovery-interval=200ms"].
extraArgs: []
# Rendered verbatim as the node's agentmesh-node.yaml: static services the node
# advertises, and local attenuation.
config:
version: v1alpha1
# Operator-declared labels the control plane attests, subject to the role's
# allowed_labels grant (e.g. region: us-east-1).
labels: {}
attenuation:
policies: []
checks: []
rules: []
services: []
# Optional service container next to the node; they share localhost, so
# config.services target_url entries point at 127.0.0.1:<port>.
# Rendered only when image is non-empty.
service:
name: service
image: ""
command: []
env: []
ports: []
resources: {}
# Set create: false and name to reuse an existing ServiceAccount; annotations
# cover IAM bindings (e.g. GKE Workload Identity).
serviceAccount:
create: true
name: ""
annotations: {}
resources: {}
nodeSelector: {}
tolerations: []
affinity: {}