openai-webmcp-challenge.pages.dev

{
  "name": "get_cluster_topology",
  "kind": "answer",
  "impl": "imperative",
  "description": "Get current LLM cluster nodes, replica counts, latencies and VRAM usage.",
  "inputSchema": {
    "type": "object",
    "properties": {
      "filterType": {
        "type": "string",
        "description": "Filter by node type: Inference, Gateway, Cache, Database"
      }
    }
  }
}

{
  "name": "scale_service_replicas",
  "kind": "act",
  "impl": "imperative",
  "description": "Scale the replica count of a designated node to absorb traffic spikes.",
  "inputSchema": {
    "type": "object",
    "properties": {
      "nodeId": {
        "type": "string",
        "description": "Node ID (e.g. llm-deepseek-r1, llm-qwen-7b)"
      },
      "replicas": {
        "type": "number",
        "description": "Target replica count (1 to 16)"
      }
    },
    "required": [
      "nodeId",
      "replicas"
    ]
  }
}

{
  "name": "optimize_gpu_memory",
  "kind": "act",
  "impl": "imperative",
  "description": "Evict stale prefix KV caches and defragment GPU memory.",
  "inputSchema": {
    "type": "object",
    "properties": {
      "targetNodeId": {
        "type": "string",
        "description": "Node ID or \"all\" to defragment entire cluster"
      }
    }
  }
}

{
  "name": "run_inference_benchmark",
  "kind": "act",
  "impl": "imperative",
  "description": "Trigger an automated throughput and latency stress test across the cluster.",
  "inputSchema": {
    "type": "object",
    "properties": {
      "concurrency": {
        "type": "number",
        "description": "Simulated request concurrency"
      }
    },
    "required": [
      "concurrency"
    ]
  }
}