Deploy your NeuroFace from Developer Hub

⏱ ~15 min ★ main activity

This is the core workshop activity. You scaffold a personal NeuroFace instance from Developer Hub, watch the pipeline deploy to a spoke, and test PPE detection — the first end-to-end proof that hub GitOps, spoke workloads, and gateway routing work together.

Open Developer Hub

  1. Open the OCP Console on the hub cluster

  2. Application menu (grid icon) → Red Hat Developer Hub

  3. Or navigate directly:

https://developer-hub.apps.cluster.example.com

  1. Log in with your workshop credentials (guest (register first) / Welcome123!)

Browse the AI Computer Vision template

Deployment Flow

  1. Click Create…​ in the left navigation

  2. Select the AI Computer Vision at the Edge template

  3. Review the parameters — the template is defined in the pattern Git repository:

# charts/all/developer-hub/files/software-templates/ai-computer-vision.yaml (excerpt)
metadata:
  name: ai-computer-vision
  title: AI Computer Vision at the Edge
spec:
  parameters:
    - required: [owner, spoke, hubClusterDomain]
      properties:
        owner:
          title: Owner
          ui:field: OwnerPicker
        spoke:
          title: Target spoke
          enum: [east, west]
        hubClusterDomain:
          title: Hub cluster domain
        maasEndpoint:
          default: https://ai-gateway.apps.__HUB_CLUSTER_DOMAIN__/v1
  steps:
    - id: fetch
      action: fetch:template
    - id: publish
      action: publish:gitlab
    - id: register
      action: catalog:register

Create your NeuroFace instance

This step creates real resources. Use your assigned workshop user as owner.

  1. Select the AI Computer Vision at the Edge template

  2. Fill in:

    • Owner: guest (register first)

    • Target spoke: east or west

    • Hub cluster domain: domain without apps. prefix (pre-filled from pattern)

    • MaaS chat endpoint: https://ai-gateway.apps.cluster.example.com/v1 (default)

  3. Click Create and watch the scaffolder log

The scaffold creates namespace neuroface-guest (register first) and GitLab repo ws-workshop/neuroface-guest (register first).

What gets deployed on the spoke

# ai-cv-skeleton — namespace and MaaS wiring (excerpt)
metadata:
  name: neuroface-${{ values.owner }}
  namespace: neuroface-${{ values.owner }}
env:
  NEUROFACE_CHAT_API_KEY: from ExternalSecret neuroface-ai-gateway-key (platform key)
# charts/all/spoke-neuroface/templates/skupper-connector.yaml
apiVersion: skupper.io/v2alpha1
kind: Connector
metadata:
  name: neuroface-app-east
  namespace: service-interconnect
spec:
  routingKey: neuroface-app-east
  host: neuroface-spoke-gateway-istio.neuroface.svc.cluster.local
  port: 8080

Verify deployment on the spoke

east-login guest (register first)
oc get pods -n neuroface-guest (register first)
neuroface-spoke-status guest (register first)

Check pod readiness

oc get pods -n neuroface-guest (register first) -o custom-columns=NAME:.metadata.name,READY:.status.containerStatuses[0].ready,STATUS:.status.phase
Expected output
NAME                              READY   STATUS
neuroface-backend-xxx             true    Running
neuroface-frontend-xxx            true    Running

Test your NeuroFace URL

curl -sk -o /dev/null -w "NeuroFace user route: HTTP %{http_code}\n" \
  "https://neuroface.apps.cluster.example.com/user/guest (register first)/"
Expected output
NeuroFace user route: HTTP 200

Inspect the Skupper connector on the spoke

east-login guest (register first)
oc get connector -n service-interconnect
oc get listener -n service-interconnect
If you get HTTP 503, pods may still be starting. Re-run oc get pods -n neuroface-guest (register first).

Test PPE detection in the UI

  1. Open https://neuroface.apps.cluster.example.com/user/guest (register first)/

  2. Upload a photo and observe PPE bounding boxes

  3. Try the chat panel (LLM via AI Gateway (Kuadrant, legacy))

  4. Check the cluster badge in the UI header — EAST (green), WEST (yellow), or HUB (blue fallback when the spoke API is unreachable)

What you learned

  • Developer Hub scaffolds GitLab repos, DevSpaces workspaces, and catalog entries

  • Each user gets namespace neuroface-userN on the chosen spoke

  • NeuroFace chat uses AI Gateway; PPE uses shared YOLO inference on the spoke

  • Skupper Connectors on the spoke tunnel traffic back to hub Gateway API routes

Next

Continue with Module 03 — OpenShift AI step-by-step.