NemoGuardrails

Overview

The NemoGuardrails resource provides guardrail services for AI safety and content moderation. It integrates with NIM endpoints and supports configuration via ConfigMap or PVC. API Group: apps.nvidia.com
API Version: v1alpha1
Kind: NemoGuardrail

Spec Fields

image

object

required

Container image configuration.

Show properties

repository

string

required

Container image repository

tag

string

required

Image tag

pullPolicy

string

Image pull policy

pullSecrets

array

Image pull secret names

configStore

object

required

Guardrail configuration source. Either ConfigMap or PVC must be specified, not both.

Show properties

configMap

object

ConfigMap containing guardrail configuration

Show properties

name

string

required

ConfigMap name (minimum length: 1)

pvc

object

PVC containing guardrail configuration

Show properties

name

string

PVC name

create

boolean

Whether to create the PVC

storageClass

string

Storage class name

volumeAccessMode

string

Volume access mode

size

string

Storage size

subPath

string

Subdirectory in PVC (defaults to “guardrails-config-store”)

nimEndpoint

object

NIM endpoint configuration for LLM inference.

Show properties

baseURL

string

required

Base URL for NIM service or NIM proxy (pattern: ^https?:\/\/[^\s]+\/v1\/?$, format: uri)

apiKeySecret

string

Secret containing API key for NIM proxy access

apiKeyKey

string

default:"NIM_ENDPOINT_API_KEY"

Key in secret containing API key

databaseConfig

object

Optional PostgreSQL database configuration for persistent storage.

Show properties

host

string

required

Database hostname (minimum length: 1)

port

integer

default:"5432"

Database port (1-65535)

databaseName

string

required

Database name (minimum length: 1)

credentials

object

required

Database credentials

Show properties

user

string

required

Database username

secretName

string

required

Secret containing password

passwordKey

string

default:"password"

Key in secret for password

otel

object

OpenTelemetry configuration.

Show properties

enabled

boolean

Enable OpenTelemetry tracing

exporterOtlpEndpoint

string

OTLP collector endpoint URL

disableLogging

boolean

Disable Python logging auto-instrumentation

exporterConfig

object

Exporter configuration

Show properties

tracesExporter

string

default:"otlp"

Traces exporter (enum: otlp, console, none)

metricsExporter

string

default:"otlp"

Metrics exporter (enum: otlp, console, none)

logsExporter

string

default:"otlp"

Logs exporter (enum: otlp, console, none)

excludedUrls

array

default:"[\"health\"]"

URLs to exclude from tracing

logLevel

string

default:"INFO"

Log level (enum: INFO, DEBUG)

expose

object

Service exposure configuration.

Show properties

service

object

Service configuration

Show properties

type

string

Service type (e.g., ClusterIP, LoadBalancer)

port

integer

default:"8000"

Service port (1-65535)

annotations

object

Service annotations

router

object

Ingress/Gateway router configuration

replicas

integer

default:"1"

Number of replicas (minimum: 1). Cannot be set when autoscaling is enabled.

scale

object

Autoscaling configuration.

Show properties

enabled

boolean

Enable horizontal pod autoscaling

hpa

object

HPA specification

Show properties

minReplicas

integer

default:"1"

Minimum replicas

maxReplicas

integer

required

Maximum replicas

metrics

array

Metrics for autoscaling

metrics

object

Metrics collection configuration.

Show properties

enabled

boolean

Enable metrics collection

serviceMonitor

object

Prometheus ServiceMonitor configuration

command

array

Override container command

args

array

Container arguments

env

array

Additional environment variables

resources

object

Resource requirements (CPU, memory)

nodeSelector

object

Node selector labels

tolerations

array

Pod tolerations

affinity

object

Pod affinity rules

labels

object

Custom labels

annotations

object

Custom annotations

userID

integer

User ID for container security context (default: 1000)

groupID

integer

Group ID for container security context (default: 2000)

runtimeClass

string

Runtime class name

Status Fields

conditions

array

Current state conditions

availableReplicas

integer

Number of available replicas

state

string

Current state (Pending, NotReady, Ready, Failed)

Example

apiVersion: apps.nvidia.com/v1alpha1
kind: NemoGuardrail
metadata:
  name: nemoguardrails-sample
  namespace: nemo
spec:
  configStore:
    pvc:
      name: "pvc-guardrail-config"
      create: true
      storageClass: ""
      volumeAccessMode: ReadWriteOnce
      size: "1Gi"
  nimEndpoint:
    baseURL: "http://meta-llama-3-1-8b-instruct.nemo.svc.cluster.local:8000/v1"
  databaseConfig:
    host: guardrail-pg-postgresql.nemo.svc.cluster.local
    port: 5432
    databaseName: guardraildb
    credentials:
      user: guardrailuser
      secretName: guardrail-pg-existing-secret
      passwordKey: password
  expose:
    service:
      type: ClusterIP
      port: 8000
  image:
    repository: nvcr.io/nvidia/nemo-microservices/guardrails
    tag: "25.10"
    pullPolicy: IfNotPresent
    pullSecrets:
      - ngc-secret
  metrics:
    serviceMonitor: {}
  replicas: 1
  resources:
    limits:
      cpu: "1"
      ephemeral-storage: 10Gi

NIM Resources

NeMo Resources

Overview

Spec Fields

Status Fields

Example

Build docs developers (and LLMs) love

NIM Resources

NeMo Resources

​Overview

​Spec Fields

​Status Fields

​Example

Build docs developers (and LLMs) love

Overview

Spec Fields

Status Fields

Example