OpenOP OEG integrations for QoD

This deployment relies on a `.env` file that is loaded at runtime by Docker Compose.

```

GROQ_API_KEY=your-secret-key

OpenOP_ISI_BACKEND=http://example-backend

OpenOP_BACKEND_OAI=http://example-backend

```

## Run

    environment:

      MCP_HOST: 0.0.0.0

      MCP_PORT: 8004

      OpenOP_Backend: http://api:8081

      OpenOP_ISI_BACKEND: ${OpenOP_ISI_BACKEND}

      OpenOP_BACKEND_OAI: ${OpenOP_BACKEND_OAI}

    ports:

      - "8004:8004"

    networks:

# Kubernetes Deployment (ai2)

This directory contains Kubernetes manifests and a deployment script for running:

- `mcp_module` (service: `mcp-service`, port `8004`)

- `ai_agent` (service: `ai-agent-service`, port `9013`)

The default namespace is `ai2`.

## Files

- `namespace.yaml`: Creates namespace `ai2`.

- `configmap.yaml`: Non-sensitive runtime configuration.

- `secret.yaml`: Sensitive runtime configuration (`GROQ_API_KEY`).

- `mcp-deployment.yaml` / `mcp-service.yaml`: MCP Deployment and NodePort Service (`32004`).

- `ai-agent-deployment.yaml` / `ai-agent-service.yaml`: AI Agent Deployment and NodePort Service (`32013`).

- `deploy.sh`: Applies resources in order and waits for deployments to become available.

## Prerequisites

- Kubernetes cluster with `kubectl` configured.

- Access to pull container images from:

  - `labs.etsi.org:5050/oop/code/ai2/mcp_module:latest`

  - `labs.etsi.org:5050/oop/code/ai2/ai_agent:latest`

- A valid `GROQ_API_KEY`.

## Configure Secrets

Set the API key in `secret.yaml`:

```yaml

stringData:

  GROQ_API_KEY: "<your-key>"

```

Alternative (CLI):

```bash

kubectl create secret generic ai2-secrets \

  --from-literal=GROQ_API_KEY=<your-key> \

  -n ai2 \

  --dry-run=client -o yaml | kubectl apply -f -

```

## Deploy

From this directory:

```bash

chmod +x deploy.sh

./deploy.sh

```

Optional namespace override:

```bash

NAMESPACE=ai2 ./deploy.sh

```

## Verify

```bash

kubectl get ns ai2

kubectl get all -n ai2

kubectl get configmap ai2-config -n ai2

kubectl get secret ai2-secrets -n ai2

```

Check readiness and logs:

```bash

kubectl rollout status deployment/mcp-deployment -n ai2

kubectl rollout status deployment/ai-agent-deployment -n ai2

kubectl logs deployment/mcp-deployment -n ai2

kubectl logs deployment/ai-agent-deployment -n ai2

```

Health endpoints:

- MCP: `http://<node-ip>:32004/health`

## Test API Locally

Forward the AI Agent service to localhost:

```bash

kubectl port-forward svc/ai-agent-service 9013:9013 -n ai2

```

In a separate terminal, test endpoint `http://127.0.0.1:9013/groq-mcp` with `Content-Type: application/json`.

Create QoD session:

```bash

curl -X POST "http://127.0.0.1:9013/groq-mcp" \

  -H "Content-Type: application/json" \

  -d '{"query":"Create QoD session for device IP 12.1.0.20 application server IP 198.51.100.10 with QoS qos-e and duration 60 seconds"}'

```

Get QoD session:

```bash

curl -X POST "http://127.0.0.1:9013/groq-mcp" \

  -H "Content-Type: application/json" \

  -d '{"query":"Get  QoD session information for 33f6392b-57d0-42a3-a9fd-b4dcce617fce"}'

```

Delete QoD session:

```bash

curl -X POST "http://127.0.0.1:9013/groq-mcp" \

  -H "Content-Type: application/json" \

  -d '{"query":"Delete QoD session 33f6392b-57d0-42a3-a9fd-b4dcce617fce"}'

```

## Deployment Order

`deploy.sh` enforces this sequence:

1. Namespace

2. ConfigMap + Secret

3. MCP Deployment + Service

4. AI Agent Deployment + Service

The AI Agent Pod also includes an init container that waits for `http://mcp-service:8004/health` before starting.

## Cleanup

```bash

kubectl delete namespace ai2

```

apiVersion: apps/v1

kind: Deployment

metadata:

  name: ai-agent-deployment

  namespace: ai2

  labels:

    app: ai-agent

    component: ai-agent

spec:

  replicas: 1

  selector:

    matchLabels:

      app: ai-agent

  template:

    metadata:

      labels:

        app: ai-agent

        component: ai-agent

    spec:

      initContainers:

        - name: wait-for-mcp

          image: curlimages/curl:latest

          command: ["sh", "-c"]

          args:

            - |

              until curl -f http://mcp-service:8004/health; do

                echo "Waiting for MCP service to be ready..."

                sleep 2

              done

              echo "MCP service is ready!"

      containers:

        - name: ai-agent

          image: labs.etsi.org:5050/oop/code/ai2/ai_agent:latest

          imagePullPolicy: IfNotPresent

          ports:

            - containerPort: 9013

              name: http

              protocol: TCP

          env:

            - name: APP_HOST

              valueFrom:

                configMapKeyRef:

                  name: ai2-config

                  key: APP_HOST

            - name: APP_PORT

              valueFrom:

                configMapKeyRef:

                  name: ai2-config

                  key: APP_PORT

            - name: MCP_SERVER_URL

              valueFrom:

                configMapKeyRef:

                  name: ai2-config

                  key: MCP_SERVER_URL

            - name: GROQ_API_KEY

              valueFrom:

                secretKeyRef:

                  name: ai2-secrets

                  key: GROQ_API_KEY

            - name: GROQ_MODEL_NAME

              valueFrom:

                configMapKeyRef:

                  name: ai2-config

                  key: GROQ_MODEL_NAME

          livenessProbe:

            httpGet:

              path: /health

              port: 9013

            initialDelaySeconds: 10

            periodSeconds: 10

            timeoutSeconds: 5

            failureThreshold: 3

          readinessProbe:

            httpGet:

              path: /health

              port: 9013

            initialDelaySeconds: 5

            periodSeconds: 5

            timeoutSeconds: 3

            failureThreshold: 3

 No newline at end of file

apiVersion: v1

kind: Service

metadata:

  name: ai-agent-service

  namespace: ai2

  labels:

    app: ai-agent

    component: ai-agent

spec:

  type: NodePort

  ports:

    - port: 9013

      targetPort: 9013

      nodePort: 32013

      protocol: TCP

      name: http

  selector:

    app: ai-agent