{
  "id": "tool:triton-inference",
  "_kind": "Tool",
  "_file": "domain/tools/tools-testing-frontend-aiml.yaml",
  "_cluster": "domain",
  "attributes": {
    "displayName": "Triton Inference Server",
    "homepageUrl": "https://github.com/triton-inference-server/server",
    "kind": "other",
    "description": "NVIDIA's open-source inference serving platform that hosts models from\nTensorRT, ONNX Runtime, PyTorch, TensorFlow, and vLLM backends behind a\nunified gRPC/HTTP API. Supports dynamic batching, model ensembles,\nconcurrent model execution, and Kubernetes-native deployment with Prometheus\nmetrics out of the box.\n"
  },
  "outgoingEdges": [
    {
      "from": "tool:triton-inference",
      "to": "language:cpp",
      "kind": "belongs_to_language"
    },
    {
      "from": "tool:triton-inference",
      "to": "skill-area:model-serving",
      "kind": "tool_used_by",
      "attributes": {}
    },
    {
      "from": "tool:triton-inference",
      "to": "skill-area:llm-infrastructure",
      "kind": "tool_used_by",
      "attributes": {}
    },
    {
      "from": "tool:triton-inference",
      "to": "skill-area:model-serving",
      "kind": "used_for"
    },
    {
      "from": "tool:triton-inference",
      "to": "skill-area:ai-evaluation",
      "kind": "used_for"
    },
    {
      "from": "tool:triton-inference",
      "to": "tool:vllm",
      "kind": "alternative_to",
      "attributes": {
        "comparison": "ML inference — Triton serves multiple model frameworks; vLLM is LLM-specific"
      }
    },
    {
      "from": "tool:triton-inference",
      "to": "tool:tensorrt",
      "kind": "alternative_to",
      "attributes": {
        "comparison": "NVIDIA inference — Triton serves models; TensorRT optimizes them"
      }
    },
    {
      "from": "tool:triton-inference",
      "to": "tool:onnx-runtime",
      "kind": "alternative_to",
      "attributes": {
        "comparison": "ML inference — Triton is a serving platform; ONNX Runtime is an inference engine"
      }
    }
  ],
  "incomingEdges": [
    {
      "from": "specialization:ml-inference-serving",
      "to": "tool:triton-inference",
      "kind": "uses_tool"
    },
    {
      "from": "tool:vllm",
      "to": "tool:triton-inference",
      "kind": "alternative_to",
      "attributes": {
        "comparison": "ML inference — vLLM is LLM-specific; Triton serves multiple model frameworks"
      }
    },
    {
      "from": "tool:tensorrt",
      "to": "tool:triton-inference",
      "kind": "alternative_to",
      "attributes": {
        "comparison": "NVIDIA inference — TensorRT optimizes models; Triton serves them"
      }
    },
    {
      "from": "tool:onnx-runtime",
      "to": "tool:triton-inference",
      "kind": "alternative_to",
      "attributes": {
        "comparison": "ML inference — ONNX Runtime is an inference engine; Triton is a serving platform"
      }
    }
  ]
}