Collecting engine metrics

# InferenceClass for the L4 shape on GKE, validated serving Qwen2.5-0.5B.
#
# One NVIDIA L4 on a g2-standard-8. The GPU is declared as a DRA device: the
# scheduler matches a ModelDeployment's nodeSelector against this capacity, then
# DRA binds the physical GPU to the serving pod. A 0.5B model uses a sliver of
# this L4 - the shape is shared with the larger L4 examples, not sized for it.
apiVersion: modelplane.ai/v1alpha1
kind: InferenceClass
metadata:
  name: gke-l4-1x-g2
spec:
  description: "GKE g2-standard-8, 1x NVIDIA L4"
  provisioning:
    provider: GKE
    gke:
      machineType: g2-standard-8
      diskSizeGb: 100
      accelerator:
        type: nvidia-l4
        count: 1
  devices:
  - name: gpu
    claim: DRA
    driver: gpu.nvidia.com
    deviceClassName: gpu.nvidia.com
    count: 1
    attributes:
      architecture: { string: Ada Lovelace }
    capacity:
      # The L4's real usable VRAM as the NVIDIA DRA driver reports it, not the
      # nominal 24GB.
      memory: { value: "23034Mi" }

# GKE InferenceCluster with one L4 node pool. Replace the project ID before
# applying. No clusterSelector targets it; the ModelDeployment matches on device
# capacity alone, so it lands here or on any other compatible cluster.
#
# Modelplane installs an in-cluster Prometheus on this cluster (the monitoring
# namespace) with open PodMonitor discovery - the metrics section of the example
# scrapes the engine through it.
apiVersion: modelplane.ai/v1alpha1
kind: InferenceCluster
metadata:
  name: gke-l4-single
  labels:
    modelplane.ai/cloud: gke
    modelplane.ai/region: us-central
spec:
  cluster:
    source: GKE
    gke:
      project: my-gcp-project  # Replace with your GCP project ID.
      region: us-central1
  nodePools:
  - name: gpu-l4
    className: gke-l4-1x-g2
    nodeCount: 1
    zones:
    - us-central1-a
    minNodeCount: 0
    maxNodeCount: 4

# Qwen2.5-0.5B-Instruct served on a single NVIDIA L4 by vLLM, validated end to
# end on GKE.
#
# A 0.5B dense model is the smallest useful serving shape: one Standalone vLLM
# pod, no ModelCache, weights pulled straight from Hugging Face. It barely touches
# the L4's VRAM, so the flags here are about behavior, not fit:
#
#   --max-model-len=16384   caps the context; the default 32K KV cache is wasteful
#                           for a model this size and a demo this small.
#   --served-model-name     the id clients pass as "model" in OpenAI requests.
#
# vLLM exposes Prometheus metrics at /metrics on its serving port (:8000) with no
# extra flag, which is what the example's PodMonitor scrapes.
#
# No --port or --host: Modelplane's routing expects the engine on its default
# :8000 with a /health probe, and passes args through verbatim.
apiVersion: modelplane.ai/v1alpha1
kind: ModelDeployment
metadata:
  name: qwen2-5-0-5b
  namespace: ml-team
spec:
  # One replica, matched to any compatible InferenceCluster by device capacity.
  replicas: 1
  engines:
  - name: qwen2-5-0-5b
    members:
    # A single self-contained vLLM pod. The container named "engine" is the
    # inference server; its image and args pass through verbatim.
    - role: Standalone
      nodeSelector:
        devices:
        - name: gpu
          count: 1
          selectors:
          # The model needs almost no VRAM; >=20Gi simply pins it to the L4 pool
          # this example provisions (the L4 reports ~23Gi). DRA evaluates this CEL
          # against the InferenceClass device, then against the GPU's
          # ResourceSlice when it binds the claim.
          - cel: |
              device.capacity["gpu.nvidia.com"].memory.compareTo(quantity("20Gi")) >= 0
      template:
        spec:
          containers:
          - name: engine
            image: vllm/vllm-openai:v0.23.0
            args:
            - "--model=Qwen/Qwen2.5-0.5B-Instruct"
            - "--served-model-name=qwen2.5-0.5b"
            - "--max-model-len=16384"

# Exposes the qwen2-5-0-5b deployment's endpoints as a single OpenAI-compatible
# URL. Modelplane labels each composed ModelEndpoint with the deployment name, so
# this selector reaches every replica. Read the public address from status.address:
#   kubectl get ms qwen2-5-0-5b -n ml-team -o jsonpath='{.status.address}'
apiVersion: modelplane.ai/v1alpha1
kind: ModelService
metadata:
  name: qwen2-5-0-5b
  namespace: ml-team
spec:
  endpoints:
  - selector:
      matchLabels:
        modelplane.ai/deployment: qwen2-5-0-5b