Skip to content

agentrax

Directory actions

More options

Directory actions

More options

Latest commit

 

History

History

agentrax

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
crds
crds
 
 
templates
templates
 
 
Chart.yaml
Chart.yaml
 
 
README.md
README.md
 
 
values.yaml
values.yaml
 
 

README.md

Agentrax Helm Chart

Official Helm chart for installing and managing Agentrax — a declarative, multi-tenant Kubernetes operator for autonomous AI agents and LLM inference workloads.

Features Supported by this Chart

  • Multi-Tenant Quota Enforcement: Dynamic admission and scaling boundaries preventing GPU, agent instance, and replica over-commitment.
  • Canary Progressive Traffic Splitting: Automated Gateway API canary routing with Prometheus error rate and latency threshold evaluation.
  • In-Cluster MCP Discovery Registry: Embedded Model Context Protocol (MCP) tool discovery server on port :9090.
  • Zero-Trust Multi-Cloud Identity: Native secretless workload identity for Microsoft Azure (AKS Workload Identity) and Amazon Web Services (AWS EKS IRSA).
  • Hardened Pod Security: Runs non-root (UID 65532), read-only root filesystem, dropped capabilities (ALL), and RuntimeDefault seccomp profile.

Prerequisites

  • Kubernetes: v1.28+
  • Helm: v3.14+
  • cert-manager: v1.13+ (required for admission webhook TLS certificates)
  • Gateway API CRDs: v1.0+ (required if using canary traffic splitting)

Adding the Helm Repository

helm repo add agentrax https://gitcommitankit.github.io/agentrax
helm repo update

Installing the Chart

1. Default Installation

Install Agentrax into the agentrax-system namespace:

helm install agentrax agentrax/agentrax \
  --namespace agentrax-system \
  --create-namespace

2. Installation with Prometheus Canary Evaluation

Wire Agentrax to an in-cluster Prometheus instance to enable canary threshold evaluations:

helm install agentrax agentrax/agentrax \
  --namespace agentrax-system \
  --create-namespace \
  --set prometheus.url="http://kube-prometheus-stack-prometheus.monitoring.svc:9090"

3. Installation for Azure AKS (Workload Identity)

helm install agentrax agentrax/agentrax \
  --namespace agentrax-system \
  --create-namespace \
  --set workloadIdentity.provider=azure \
  --set workloadIdentity.enabled=true \
  --set workloadIdentity.azureClientId="<MANAGED_IDENTITY_CLIENT_ID>" \
  --set workloadIdentity.azureTenantId="<AZURE_TENANT_ID>"

4. Installation for AWS EKS (IRSA)

helm install agentrax agentrax/agentrax \
  --namespace agentrax-system \
  --create-namespace \
  --set workloadIdentity.provider=aws \
  --set workloadIdentity.enabled=true \
  --set workloadIdentity.awsRoleArn="arn:aws:iam::<ACCOUNT_ID>:role/AgentraxOperatorRole"

Upgrading the Chart

helm repo update
helm upgrade agentrax agentrax/agentrax \
  --namespace agentrax-system \
  --reuse-values

Uninstalling the Chart

helm uninstall agentrax --namespace agentrax-system

Note: Custom Resource Definitions (CRDs) are retained by default to prevent accidental data loss. To remove them manually:

kubectl delete crd agentdeployments.agentrax.io tenantquotas.agentrax.io

Configuration Reference (values.yaml)

Parameter Description Default
replicaCount Number of controller-manager pod replicas 1
image.repository Image repository for the controller-manager ghcr.io/gitcommitankit/agentrax
image.tag Image tag (defaults to Chart.appVersion) ""
image.pullPolicy Image pull policy IfNotPresent
serviceAccount.create Create a dedicated ServiceAccount true
serviceAccount.name Custom ServiceAccount name (auto-generated if empty) ""
serviceAccount.annotations Custom annotations for ServiceAccount {}
podSecurityContext.runAsNonRoot Run controller container as non-root user true
podSecurityContext.seccompProfile.type Container runtime default seccomp profile RuntimeDefault
securityContext.allowPrivilegeEscalation Disallow privilege escalation false
securityContext.readOnlyRootFilesystem Mount root filesystem as read-only true
securityContext.capabilities.drop Linux capabilities to drop ["ALL"]
resources.limits.cpu CPU limit for controller manager 500m
resources.limits.memory Memory limit for controller manager 128Mi
resources.requests.cpu CPU request for controller manager 10m
resources.requests.memory Memory request for controller manager 64Mi
manager.leaderElect Enable leader election (required if replicaCount > 1) false
manager.metricsBindAddress Metrics endpoint bind address (:8443, :8080, or 0 to disable) "0"
manager.metricsSecure Serve metrics securely via HTTPS true
manager.healthProbeBindPort Port for /healthz and /readyz probes 8081
manager.gpuResourceName Resource name used for GPU quota accounting "nvidia.com/gpu"
prometheus.url Base URL of Prometheus for canary metric evaluation ""
gateway.name Name of Gateway API Gateway for canary routing "agentrax-gateway"
gateway.namespace Namespace of Gateway API Gateway "agentrax-system"
registry.service.port In-cluster port for the MCP discovery service 9090
registry.bindPort Container bind port for MCP discovery HTTP server 9090
registry.ttl Inactivity TTL window before purging stale agents "90s"
mcp.healthInterval Interval between background MCP health checks "30s"
otlp.endpoint gRPC endpoint for OTLP trace exporter (empty to disable tracing) ""
otlp.insecure Use plaintext gRPC for local development tracing false
workloadIdentity.enabled Enable keyless cloud IAM pod credentials false
workloadIdentity.provider Cloud provider (azure or aws) "azure"
workloadIdentity.azureClientId Azure User-Assigned Managed Identity Client ID ""
workloadIdentity.azureTenantId Azure Active Directory Tenant ID ""
workloadIdentity.awsRoleArn AWS IAM Role ARN for IRSA ""