deploy/k8s/ucs/manifest.yaml

# Copyright (c) 2025, NVIDIA CORPORATION.  All rights reserved.
#
# NVIDIA CORPORATION and its licensors retain all intellectual property
# and proprietary rights in and to this software, related documentation
# and any modifications thereto.  Any use, reproduction, disclosure or
# distribution of this software and related documentation without an express
# license agreement from NVIDIA CORPORATION is strictly prohibited.

type: msapplication
specVersion: 2.5.0
name: ucf.svc.ace-controller
chartName: ace-controller
description: ACE Controller
version: 1.0.2
displayName: "ACE Controller Microservice"
category:
  functional: "Conversational AI"
  industry: "General"
tags: []
keywords: []
nSpectId: NSPECT-XGIZ-EB0C

publish: false

egress-endpoints:
  - name: "redis"
    description: Redis message broker
    protocol: TCP
    scheme: asyncio
    mandatory: False
    data-flow: in-out
  - name: "riva-speech"
    description: Riva Speech Skills API
    scheme: grpc
    protocol: TCP
    mandatory: False
    data-flow: in-out
  - name: "animgraph-http"
    description: Animation Graph HTTP API
    scheme: http
    protocol: TCP
    mandatory: False
    data-flow: out
  - name: "animgraph-grpc"
    description: Animation Graph GRPC API
    scheme: grpc
    protocol: TCP
    mandatory: False
    data-flow: out
  - name: "a2f-grpc"
    description: Audio2Face service GRPC API
    scheme: grpc
    protocol: TCP
    mandatory: False
    data-flow: out

ingress-endpoints:
  - name: http-api
    description: ACE Controller REST API
    scheme: http
    data-flow: in-out

secrets:
  - name: nvidia-api-key-secret
    description: Secret for NVIDIA API key
    mandatory: False
    mountPath: /secrets
    fileName: nvidia_api_key.txt
  - name: openai-key-secret
    description: Secret for passing OpenAI key
    mandatory: False
    mountPath: /secrets
    fileName: openai_api_key.txt
  - name: elevenlabs-api-key-secret
    description: Secret for ElevenLabs API key
    mandatory: False
    mountPath: /secrets
    fileName: elevenlabs_api_key.txt
  - name: custom-env-secrets
    description: Secret for passing custom env variables and API keys
    mandatory: False
    mountPath: /secrets
    fileName: custom.env


externalFiles:
  - name: config.yaml
    description: config
    mandatory: true
    isDirectory: false

params:
  OTEL_SDK_DISABLED: 'false'
  #> description: when enabled, tracing data will be exported
  #> type: string
  OTEL_SERVICE_NAME: ace-controller
  #> description: protocol for exporting OTel data
  #> type: string
  OTEL_EXPORTER_OTLP_ENDPOINT: ""
  #> description: endpoint for Otel collector
  #> type: string
  OTEL_EXPORTER_OTLP_PROTOCOL: grpc
  #> description: protocol for exporting OTel data
  #> type: string
  DEV: "0"
  #> description: wether to activate dependencies hot reloading or not
  #> type: string
  image: "nvcr.io/nvidia/ace/tokkio-reference-ace-controller"
  #> description: pipeline image repo
  #> type: string
  tag: "5.0.0"
  #> description: pipeline image tag
  #> type: string

tests:
  - name: dev-params1
    app: tests/dev/app.yaml
    params: tests/dev/params1.yaml
    ciTrigger: false
    timeout: 10
    duration: 10
    installPreReqs: true  # Wether to install foundational services
    namespace: default  # Kubernetes namespace
    gpuNodeLabels: ""
    watchAllPods: true # OR set to false and set list of pods to watch below
    watchPods:
    - <pod-name-regex>
    testerPods:  # At least one tester pod is required
    - name: testpod1  # Name of the test pod
      startSignature: <START>  # Signature to look for in the logs indicating start of tests. Regex is accepted
      endSignature: <END>  # Signature to look for in the logs indicating end of tests. Regex is accepted
      errorSignatures:  # Signatures that indicate test failures.  Regex is accepted
      - <REGEX1>
      - <REGEX2>

---
spec:
  - name: ace-controller-deployment
    type: ucf.k8s.app.deployment
    parameters:
      apptype: statefull
      statefulSetServiceName: ace-controller-service
      extraSpecs:
        podManagementPolicy: Parallel


  - name: "ace-controller-container"
    type: ucf.k8s.container
    parameters:
      image:
        repository: $params.image
        tag: $params.tag
        pullPolicy: Always
      command: ["/bin/bash", "-c"]
      args: ["source /opt/scripts/env.sh && /code/entrypoint.sh"]
      env:
        - name: ANIMGRAPH_URL
          value: "http://$egress.animgraph-http.address:$egress.animgraph-http.port"
        - name: ANIMGRAPH_GRPC_URL
          value: "$egress.animgraph-grpc.address:$egress.animgraph-grpc.port"
        - name: REDIS_URL
          value: "redis://$egress.redis.address:$egress.redis.port"
        - name: A2F_GRPC_URL
          value: "$egress.a2f-grpc.address:$egress.a2f-grpc.port"
        - name: RIVA_SERVER_URL
          value: "$egress.riva-speech.address:$egress.riva-speech.port"
        - name: DEV
          value: $params.DEV
        - name: OTEL_SDK_DISABLED
          value: $params.OTEL_SDK_DISABLED
        - name: OTEL_SERVICE_NAME
          value: $params.OTEL_SERVICE_NAME
        - name: OTEL_EXPORTER_OTLP_ENDPOINT
          value: $params.OTEL_EXPORTER_OTLP_ENDPOINT
        - name: OTEL_EXPORTER_OTLP_PROTOCOL
          value: $params.OTEL_EXPORTER_OTLP_PROTOCOL
        - name: OTEL_PYTHON_LOGGING_AUTO_INSTRUMENTATION_ENABLED
          value: 'true'
        - name: IMAGE_NAME
          value: $params.image
        - name: IMAGE_TAG
          value: $params.tag
      ports:
      - containerPort: 8000
        name: http
      startupProbe:
        tcpSocket:
          port: http
        initialDelaySeconds: 20
        failureThreshold: 30
        periodSeconds: 10
      livenessProbe:
        tcpSocket:
          port: http
        initialDelaySeconds: 20
        periodSeconds: 20
        timeoutSeconds: 5
      readinessProbe:
        tcpSocket:
          port: http
        initialDelaySeconds: 20
        periodSeconds: 20
        timeoutSeconds: 5

  - name: app-storage
    type: ucf.k8s.pvc
    parameters:
      spec:
        storageClassName: mdx-local-path
        accessModes: [ReadWriteOnce]
        resources:
          requests:
            storage: 5Gi

  - name: app-storage-volume
    type: ucf.k8s.volume
    parameters:
      persistentVolumeClaim:
        claimName: ace-controller-app-storage

  - name: app-storage-mount
    type: ucf.appspec.defaultVolumeMount
    parameters:
      name: app-storage-volume
      mountPath: /code

  - name: restartPolicy
    type: ucf.k8s.restartPolicy
    parameters:
      policy: Always # Always / OnFailure / Never

  - name: podSecurityContext
    type: ucf.k8s.podSecurityContext
    parameters:
      runAsGroup: 1000
      runAsUser: 1000

  - name: ace-controller-service
    type: ucf.k8s.service
    parameters:
      ports:
      - port: 8000
        protocol: TCP
        name: http-api
      type: ClusterIP

  - name: ace-controller-metrics
    type: ucf.crd.podMonitor
    parameters:
      portName: metrics
      path: /metrics
  
  - name: cm-dependencies
    type: ucf.appspec.restartPodOnConfigChanges
    parameters:
      # Add dependency on all configmaps detected in the microservice
      addAll: true