Skip to main content

Model Market

Backend API​

Model Market does not have a dedicated backend module. Its catalog is static frontend data in src/data/modelCatalog.js, so there is no backend catalog API to document. The page only calls shared cluster and deployment APIs.

List Lens-managed clusters​

List Lens-managed clusters, optionally including draft records that have not yet been finalized.

GET /api/cluster/clusters

Request:

  • Query params:
    • includeDrafts (boolean, optional): include draft cluster records that have not been finalized yet.

This endpoint has no request body.

Response:

Returns a ClustersResponse object with an items array of ClusterRecord objects.

{
  "items": [
    {
      "id": "cluster-123",
      "name": "lab-cluster",
      "description": "Shared Intel GPU cluster",
      "createdAt": "2026-09-20T10:15:00+00:00",
      "ready": true,
      "proxy": {
        "mode": "auto",
        "httpProxy": null,
        "httpsProxy": null,
        "noProxy": null
      },
      "llmDRef": "v0.9.0",
      "llmDBenchmarkRef": "main",
      "llmDRepoPath": "/opt/prism/checkouts/llm-d",
      "llmDBenchmarkRepoPath": "/opt/prism/checkouts/llm-d-benchmark",
      "deploymentSource": {
        "resolved_from": "cluster-registry",
        "ref": "v0.9.0"
      },
      "benchmarkSource": {
        "resolved_from": "cluster-registry",
        "ref": "main"
      },
      "draft": false
    }
  ]
}

Get a cluster overview​

Get the live Kubernetes overview for a Lens-managed cluster, including node, GPU, cached image, and component health summaries.

GET /api/cluster/overview

Request:

  • Query params:
    • clusterId (string, required): the Lens-managed cluster id to inspect.

This endpoint has no request body.

Response:

Returns an OverviewResponse with the cluster record, fetch timestamp, and a KubernetesSummary snapshot.

{
  "cluster": {
    "id": "cluster-123",
    "name": "lab-cluster",
    "description": "Shared Intel GPU cluster",
    "createdAt": "2026-09-20T10:15:00+00:00",
    "ready": true,
    "proxy": {
      "mode": "auto",
      "httpProxy": null,
      "httpsProxy": null,
      "noProxy": null
    },
    "llmDRef": "v0.9.0",
    "llmDBenchmarkRef": "main",
    "llmDRepoPath": "/opt/prism/checkouts/llm-d",
    "llmDBenchmarkRepoPath": "/opt/prism/checkouts/llm-d-benchmark",
    "deploymentSource": {
      "resolved_from": "cluster-registry",
      "ref": "v0.9.0"
    },
    "benchmarkSource": {
      "resolved_from": "cluster-registry",
      "ref": "main"
    },
    "draft": false
  },
  "fetchedAt": "2026-09-23T09:30:00+00:00",
  "kubernetes": {
    "components": {
      "intelDevicePlugin": {
        "installed": true,
        "ready": 1,
        "desired": 1,
        "status": "ready"
      },
      "monitoring": {
        "installed": true,
        "ready": 5,
        "desired": 5,
        "status": "ready"
      }
    },
    "nodes": [
      {
        "name": "worker-1",
        "version": "v1.30.2",
        "osImage": "Ubuntu 22.04.5 LTS",
        "ready": true,
        "schedulingDisabled": false,
        "cpu": 16.0,
        "cpuUsagePercent": 25.0,
        "memoryBytes": 68719476736,
        "memoryUsagePercent": 50.0,
        "gpu": true,
        "gpuCount": 2,
        "gpuUsagePercent": 80.0,
        "vramBytes": 103079215104,
        "vramUsagePercent": 40.0,
        "role": "worker",
        "cachedImages": [
          "ghcr.io/llm-d/llm-d-xpu:v0.9.0"
        ],
        "gpus": [
          {
            "id": "gpu-0",
            "index": 0,
            "name": "GPU #0",
            "pciAddress": "0000:03:00.0",
            "computeUsagePercent": 80.0,
            "vramTotalBytes": 51539607552,
            "vramUsedBytes": 21474836480,
            "vramUsagePercent": 41.6,
            "vramReadThroughputBytes": 10000,
            "vramWriteThroughputBytes": 20000,
            "vramTotalThroughputBytes": 30000,
            "vramBandwidthPercent": 15.0
          }
        ],
        "diskBytes": 1099511627776,
        "diskUsedBytes": 274877906944,
        "diskUsagePercent": 25.0
      }
    ],
    "namespaces": [
      {
        "name": "standard-qwen3",
        "deployments": [
          {
            "name": "qwen3-vllm",
            "desired": 1,
            "ready": 1,
            "available": 1
          }
        ],
        "ready": 1
      }
    ],
    "hardware": {
      "nodes": 1,
      "cpuCores": 16.0,
      "cpuUsagePercent": 25.0,
      "memoryBytes": 68719476736,
      "memoryUsagePercent": 50.0,
      "gpuNodes": 1,
      "gpuCount": 2,
      "gpuUsagePercent": 80.0,
      "vramBytes": 103079215104,
      "vramUsagePercent": 40.0,
      "diskBytes": 1099511627776,
      "diskUsagePercent": 25.0,
      "allocatedGpuCount": 1,
      "availableGpuCount": 1,
      "usableGpuCount": 1,
      "configuredGpuLimit": null
    }
  }
}

Start a Standard Kubernetes Service deployment run​

Start a Standard Kubernetes Service deployment run using Lens's allowlisted vLLM contract.

note

One-click deployment no longer offers a Standard mode; it deploys through Manual (Guide templates) or Agentic. This endpoint remains available to API and MCP callers that want the plain allowlisted vLLM contract without a Guide template, and is documented here because it is still the deployment contract reached from the Model Market domain.

POST /api/v1/deployments/standard-vllm-runs

Request:

  • JSON body fields:
    • model (string, required)
    • deployment_name (string, required)
    • description (string, optional)
    • cluster_session_id (string, required, 36 chars)
    • replicas (integer, required)
    • tensor_parallel_size (integer, required)
    • storage_type ("local-cache" | "pvc" | "model-cache", optional)
    • storage_volume_id (string, optional)
    • model_cache_path (string, optional)
    • pvc_name (string, optional)
    • max_model_len (integer, optional)
    • gpu_memory_utilization (number, optional)
    • max_num_seqs (integer, optional)
    • max_num_batched_tokens (integer, optional)
    • enable_prefix_caching (boolean, optional)
    • enforce_eager (boolean, optional)
    • vllm_arguments (array<object>, optional; each item uses name and value string keys)
    • use_cluster_proxy (boolean, optional)
    • http_proxy (string, optional)
    • https_proxy (string, optional)
    • no_proxy (string, optional)
{
  "model": "Qwen/Qwen3-8B",
  "deployment_name": "qwen3-8b-standard",
  "description": "Model Market launch",
  "cluster_session_id": "123e4567-e89b-12d3-a456-426614174000",
  "replicas": 2,
  "tensor_parallel_size": 2,
  "storage_type": "local-cache",
  "model_cache_path": "/home/prism/.cache",
  "max_model_len": 8192,
  "gpu_memory_utilization": 0.85,
  "max_num_seqs": 128,
  "max_num_batched_tokens": 4096,
  "enable_prefix_caching": true,
  "enforce_eager": false,
  "vllm_arguments": [
    {
      "name": "enable-auto-tool-choice",
      "value": "true"
    }
  ],
  "use_cluster_proxy": true,
  "http_proxy": "http://proxy.example:911",
  "https_proxy": "http://proxy.example:911",
  "no_proxy": "localhost,.svc"
}

Response:

Returns a queued DeploymentRun. The source_configurations entry is the immutable DeployableConfiguration generated from the allowlisted Standard request.

{
  "id": "run-123",
  "status": "queued",
  "source_configurations": [
    {
      "schema_version": "deployable-configuration.v1",
      "type": "baseline",
      "format": "helm",
      "provider_ref": "baseline-vllm",
      "checksum": "sha256:1d1c9c5b7f4d",
      "content": {
        "model": {
          "name": "Qwen/Qwen3-8B"
        },
        "decode": {
          "replicaCount": 2,
          "tensorParallelSize": 2,
          "maxModelLen": 8192
        },
        "runtime": {
          "image": "ghcr.io/llm-d/llm-d-xpu:v0.9.0",
          "imageMode": "pull-existing",
          "modelSource": "auto-cache",
          "mountPath": "/home/prism/.cache",
          "environment": {
            "HTTP_PROXY": "http://proxy.example:911",
            "HTTPS_PROXY": "http://proxy.example:911",
            "NO_PROXY": "localhost,.svc,127.0.0.1,.cluster.local"
          }
        },
        "customParameters": [
          {
            "target": "decode",
            "kind": "argument",
            "name": "gpu-memory-utilization",
            "value": "0.85"
          },
          {
            "target": "decode",
            "kind": "argument",
            "name": "enable-prefix-caching",
            "value": "true"
          }
        ]
      },
      "provenance": {
        "source": "standard-kubernetes-service.v1",
        "deployment_name": "qwen3-8b-standard",
        "description": "Model Market launch",
        "storage_type": "local-cache",
        "namespace_policy": {
          "prefix": "standard-"
        }
      }
    }
  ],
  "cases": [],
  "failure_policy": "continue",
  "provenance": {
    "deployment_source": {
      "kind": "standard-kubernetes-service"
    },
    "cluster_session_id": "123e4567-e89b-12d3-a456-426614174000",
    "deployment_name": "qwen3-8b-standard",
    "description": "Model Market launch",
    "model_market": {
      "deployment_name": "qwen3-8b-standard",
      "description": "Model Market launch"
    }
  },
  "created_at": "2026-09-23T09:35:00+00:00",
  "started_at": null,
  "finished_at": null
}

Model Market does not expose its own catalog API; see Clusters and Deployments for the full backend-facing module pages.

See the full request/response schemas in the FastAPI OpenAPI schema (regenerate with npm run docs:api).