feat(nodes): GPU detection and K8s node hardware info

ushadow/backend/src/models/kubernetes.py

@@ -114,6 +114,10 @@ class KubernetesNode(BaseModel):
     external_ip: Optional[str] = Field(None, description="External IP address")
     hostname: Optional[str] = Field(None, description="Hostname")
 
+    # GPU capacity (from extended resources)
+    gpu_capacity_nvidia: Optional[int] = Field(None, description="NVIDIA GPU count from nvidia.com/gpu")
+    gpu_capacity_amd: Optional[int] = Field(None, description="AMD GPU count from amd.com/gpu")
+
     # Taints and labels
     taints: List[Dict[str, str]] = Field(default_factory=list, description="Node taints")
     labels: Dict[str, str] = Field(default_factory=dict, description="Node labels")

ushadow/backend/src/models/unode.py

@@ -36,6 +36,16 @@ class UNodeType(str, Enum):
     KUBERNETES = "kubernetes"  # Kubernetes cluster
 
 
+class GPUDevice(BaseModel):
+    """A single GPU device detected on the node."""
+    vendor: str  # "nvidia" or "amd"
+    index: int = 0
+    model: str = ""
+    vram_mb: Optional[int] = None
+    cuda_version: Optional[str] = None
+    rocm_version: Optional[str] = None
+
+
 class UNodeCapabilities(BaseModel):
     """Capabilities of a u-node."""
     can_run_docker: bool = True
@@ -45,6 +55,11 @@ class UNodeCapabilities(BaseModel):
     available_memory_mb: int = 0
     available_cpu_cores: float = 0
     available_disk_gb: float = 0
+    # GPU details (additive, backward-compatible defaults)
+    gpu_count: int = 0
+    gpu_devices: List[GPUDevice] = Field(default_factory=list)
+    gpu_model: Optional[str] = None
+    gpu_vram_mb: Optional[int] = None
 
 
 class UNodeBase(BaseModel):

ushadow/backend/src/services/kubernetes_manager.py

@@ -283,6 +283,12 @@ async def list_nodes(self, cluster_id: str) -> List["KubernetesNode"]:
                         "effect": taint.effect
                     })
 
+                # Parse GPU extended resources
+                gpu_nvidia_raw = capacity.get("nvidia.com/gpu")
+                gpu_amd_raw = capacity.get("amd.com/gpu")
+                gpu_capacity_nvidia = int(gpu_nvidia_raw) if gpu_nvidia_raw else None
+                gpu_capacity_amd = int(gpu_amd_raw) if gpu_amd_raw else None
+
                 k8s_node = KubernetesNode(
                     name=node.metadata.name,
                     cluster_id=cluster_id,
@@ -296,6 +302,8 @@ async def list_nodes(self, cluster_id: str) -> List["KubernetesNode"]:
                     memory_capacity=capacity.get("memory"),
                     cpu_allocatable=allocatable.get("cpu"),
                     memory_allocatable=allocatable.get("memory"),
+                    gpu_capacity_nvidia=gpu_capacity_nvidia,
+                    gpu_capacity_amd=gpu_capacity_amd,
                     roles=roles,
                     internal_ip=internal_ip,
                     external_ip=external_ip,

ushadow/frontend/src/components/kubernetes/ClusterNodeList.tsx

@@ -0,0 +1,159 @@
+import { useState } from 'react'
+import { ChevronDown, ChevronRight, Server, RefreshCw } from 'lucide-react'
+import { kubernetesApi, KubernetesNode } from '../../services/api'
+
+interface ClusterNodeListProps {
+  clusterId: string
+  clusterStatus: string
+  nodeCount?: number
+}
+
+function formatMemory(memStr?: string): string {
+  if (!memStr) return '?'
+  const ki = parseInt(memStr.replace('Ki', ''))
+  if (!isNaN(ki)) return `${(ki / 1024 / 1024).toFixed(1)} Gi`
+  if (memStr.endsWith('Gi')) return memStr
+  return memStr
+}
+
+function GpuBadge({ node }: { node: KubernetesNode }) {
+  const nvidia = node.gpu_capacity_nvidia
+  const amd = node.gpu_capacity_amd
+  if (!nvidia && !amd) return null
+
+  const parts: string[] = []
+  if (nvidia) parts.push(`${nvidia}x NVIDIA`)
+  if (amd) parts.push(`${amd}x AMD`)
+
+  return (
+    <span
+      className="inline-flex items-center px-2 py-0.5 text-xs font-medium rounded-full bg-accent-100 dark:bg-accent-900/30 text-accent-700 dark:text-accent-300"
+      data-testid={`k8s-node-gpu-${node.name}`}
+    >
+      {parts.join(', ')} GPU
+    </span>
+  )
+}
+
+function K8sNodeCard({ node }: { node: KubernetesNode }) {
+  return (
+    <div
+      className="p-3 bg-neutral-50 dark:bg-neutral-800/50 rounded-lg border border-neutral-200 dark:border-neutral-700"
+      data-testid={`k8s-node-card-${node.name}`}
+    >
+      <div className="flex items-center justify-between mb-2">
+        <div className="flex items-center gap-2">
+          <Server className="h-4 w-4 text-neutral-400" />
+          <span className="text-sm font-medium text-neutral-900 dark:text-neutral-100 truncate">
+            {node.name}
+          </span>
+        </div>
+        <div className="flex items-center gap-2">
+          <GpuBadge node={node} />
+          <span
+            className={`px-2 py-0.5 text-xs rounded-full font-medium ${
+              node.ready
+                ? 'bg-success-100 dark:bg-success-900/30 text-success-700 dark:text-success-300'
+                : 'bg-danger-100 dark:bg-danger-900/30 text-danger-700 dark:text-danger-300'
+            }`}
+            data-testid={`k8s-node-status-${node.name}`}
+          >
+            {node.status}
+          </span>
+        </div>
+      </div>
+
+      <div className="grid grid-cols-2 gap-x-4 gap-y-1 text-xs text-neutral-600 dark:text-neutral-400">
+        {node.roles.length > 0 && (
+          <div>
+            <span className="text-neutral-400 dark:text-neutral-500">Roles: </span>
+            {node.roles.join(', ')}
+          </div>
+        )}
+        <div>
+          <span className="text-neutral-400 dark:text-neutral-500">CPU: </span>
+          {node.cpu_capacity || '?'}
+        </div>
+        <div>
+          <span className="text-neutral-400 dark:text-neutral-500">Mem: </span>
+          {formatMemory(node.memory_capacity)}
+        </div>
+        {node.kubelet_version && (
+          <div>
+            <span className="text-neutral-400 dark:text-neutral-500">Kubelet: </span>
+            {node.kubelet_version}
+          </div>
+        )}
+        {node.os_image && (
+          <div className="col-span-2 truncate">
+            <span className="text-neutral-400 dark:text-neutral-500">OS: </span>
+            {node.os_image}
+          </div>
+        )}
+      </div>
+    </div>
+  )
+}
+
+export default function ClusterNodeList({ clusterId, clusterStatus, nodeCount }: ClusterNodeListProps) {
+  const [expanded, setExpanded] = useState(false)
+  const [nodes, setNodes] = useState<KubernetesNode[]>([])
+  const [loading, setLoading] = useState(false)
+  const [error, setError] = useState<string | null>(null)
+  const [loaded, setLoaded] = useState(false)
+
+  const handleToggle = async () => {
+    if (!expanded && !loaded) {
+      setLoading(true)
+      setError(null)
+      try {
+        const response = await kubernetesApi.listNodes(clusterId)
+        setNodes(response.data)
+        setLoaded(true)
+      } catch (err: any) {
+        setError(err.response?.data?.detail || 'Failed to load nodes')
+      } finally {
+        setLoading(false)
+      }
+    }
+    setExpanded(!expanded)
+  }
+
+  if (clusterStatus !== 'connected') return null
+
+  return (
+    <div className="mt-4 pt-4 border-t border-neutral-200 dark:border-neutral-700">
+      <button
+        onClick={handleToggle}
+        className="flex items-center gap-2 text-sm font-medium text-neutral-700 dark:text-neutral-300 hover:text-neutral-900 dark:hover:text-neutral-100 transition-colors w-full"
+        data-testid={`cluster-nodes-toggle-${clusterId}`}
+      >
+        {loading ? (
+          <RefreshCw className="h-4 w-4 animate-spin" />
+        ) : expanded ? (
+          <ChevronDown className="h-4 w-4" />
+        ) : (
+          <ChevronRight className="h-4 w-4" />
+        )}
+        Nodes{nodeCount != null && ` (${nodeCount})`}
+      </button>
+
+      {expanded && (
+        <div
+          className="mt-3 space-y-2"
+          data-testid={`cluster-nodes-list-${clusterId}`}
+        >
+          {error && (
+            <p className="text-xs text-danger-600 dark:text-danger-400">{error}</p>
+          )}
+          {!error && nodes.map((node) => (
+            <K8sNodeCard key={node.name} node={node} />
+          ))}
+          {!error && loaded && nodes.length === 0 && (
+            <p className="text-xs text-neutral-500">No nodes found</p>
+          )}
+        </div>
+      )}
+    </div>
+  )
+}

ushadow/frontend/src/pages/ClusterPage.tsx

@@ -35,6 +35,17 @@ interface UNode {
     available_memory_mb: number
     available_cpu_cores: number
     available_disk_gb: number
+    gpu_count?: number
+    gpu_devices?: Array<{
+      vendor: string
+      index: number
+      model: string
+      vram_mb?: number
+      cuda_version?: string
+      rocm_version?: string
+    }>
+    gpu_model?: string
+    gpu_vram_mb?: number
   }
   metadata?: {
     last_metrics?: {
@@ -102,6 +113,9 @@ interface LeaderInfo {
       available_memory_mb: number
       available_cpu_cores: number
       available_disk_gb: number
+      gpu_count?: number
+      gpu_model?: string
+      gpu_vram_mb?: number
     }
     services?: string[]
     manager_version?: string
@@ -793,6 +807,40 @@ export default function ClusterPage() {
                 )}
               </div>
 
+              {/* GPU Info */}
+              {node.capabilities?.can_run_gpu && (
+                <div
+                  className="flex flex-wrap items-center gap-2 text-xs mb-4"
+                  data-testid={`node-gpu-info-${node.hostname}`}
+                >
+                  <span className="inline-flex items-center px-2 py-0.5 rounded-full bg-accent-100 dark:bg-accent-900/30 text-accent-700 dark:text-accent-300 font-medium">
+                    {(node.capabilities.gpu_count || 1) > 1
+                      ? `${node.capabilities.gpu_count}x GPU`
+                      : 'GPU'}
+                  </span>
+                  {node.capabilities.gpu_model && (
+                    <span className="text-neutral-600 dark:text-neutral-300">
+                      {node.capabilities.gpu_model}
+                    </span>
+                  )}
+                  {node.capabilities.gpu_vram_mb && (
+                    <span className="text-neutral-500 dark:text-neutral-400">
+                      {(node.capabilities.gpu_vram_mb / 1024).toFixed(0)} GB VRAM
+                    </span>
+                  )}
+                  {node.capabilities.gpu_devices?.[0]?.cuda_version && (
+                    <span className="text-neutral-500 dark:text-neutral-400">
+                      CUDA {node.capabilities.gpu_devices[0].cuda_version}
+                    </span>
+                  )}
+                  {node.capabilities.gpu_devices?.[0]?.rocm_version && (
+                    <span className="text-neutral-500 dark:text-neutral-400">
+                      ROCm {node.capabilities.gpu_devices[0].rocm_version}
+                    </span>
+                  )}
+                </div>
+              )}
+
               {/* Deployed Services */}
               {getNodeDeployments(node.hostname).length > 0 && (
                 <div className="mb-4 border-t border-neutral-200 dark:border-neutral-700 pt-3">

ushadow/frontend/src/pages/KubernetesClustersPage.tsx

@@ -6,6 +6,7 @@ import Modal from '../components/Modal'
 import ConfirmDialog from '../components/ConfirmDialog'
 import DeployModal from '../components/DeployModal'
 import DNSManagementPanel from '../components/kubernetes/DNSManagementPanel'
+import ClusterNodeList from '../components/kubernetes/ClusterNodeList'
 
 interface InfraService {
   found: boolean
@@ -621,6 +622,13 @@ export default function KubernetesClustersPage() {
                   )}
                 </div>
 
+                {/* Nodes */}
+                <ClusterNodeList
+                  clusterId={cluster.cluster_id}
+                  clusterStatus={cluster.status}
+                  nodeCount={cluster.node_count}
+                />
+
                 {/* Actions */}
                 <div className="flex justify-between items-center pt-4 border-t border-neutral-200 dark:border-neutral-700 gap-2">
                   <div className="flex gap-2">

ushadow/frontend/src/services/api.ts

@@ -662,6 +662,29 @@ export interface CertificateStatus {
   renewal_time?: string
 }
 
+export interface KubernetesNode {
+  name: string
+  cluster_id: string
+  status: string
+  ready: boolean
+  kubelet_version?: string
+  os_image?: string
+  kernel_version?: string
+  container_runtime?: string
+  cpu_capacity?: string
+  memory_capacity?: string
+  cpu_allocatable?: string
+  memory_allocatable?: string
+  gpu_capacity_nvidia?: number
+  gpu_capacity_amd?: number
+  roles: string[]
+  internal_ip?: string
+  external_ip?: string
+  hostname?: string
+  taints: Array<{ key: string; value: string; effect: string }>
+  labels: Record<string, string>
+}
+
 export const kubernetesApi = {
   addCluster: (data: { name: string; kubeconfig: string; context?: string; namespace?: string; labels?: Record<string, string> }) =>
     api.post<KubernetesCluster>('/api/kubernetes', data),
@@ -745,6 +768,10 @@ export const kubernetesApi = {
     api.get<{ certificates: CertificateStatus[]; total: number }>(
       `/api/kubernetes/${clusterId}/dns/certificates${namespace ? `?namespace=${namespace}` : ''}`
     ),
+
+  // Node operations
+  listNodes: (clusterId: string) =>
+    api.get<KubernetesNode[]>(`/api/kubernetes/${clusterId}/nodes`),
 }
 
 // Service Definition and Deployment types