{
  "id": "tool:tensorrt",
  "_kind": "Tool",
  "_file": "domain/tools/tools-testing-frontend-aiml.yaml",
  "_cluster": "domain",
  "attributes": {
    "displayName": "TensorRT",
    "homepageUrl": "https://developer.nvidia.com/tensorrt",
    "kind": "other",
    "description": "NVIDIA's SDK for high-performance deep-learning inference on GPU. Optimises\ntrained models through layer fusion, precision calibration (FP16/INT8),\nand kernel auto-tuning to minimise latency and maximise throughput.\nIntegrates with PyTorch via torch-tensorrt, ONNX import, and the Triton\nInference Server backend ecosystem.\n"
  },
  "outgoingEdges": [
    {
      "from": "tool:tensorrt",
      "to": "language:cpp",
      "kind": "belongs_to_language"
    },
    {
      "from": "tool:tensorrt",
      "to": "skill-area:model-serving",
      "kind": "tool_used_by",
      "attributes": {}
    },
    {
      "from": "tool:tensorrt",
      "to": "skill-area:model-optimisation",
      "kind": "tool_used_by",
      "attributes": {}
    },
    {
      "from": "tool:tensorrt",
      "to": "skill-area:model-serving",
      "kind": "used_for"
    },
    {
      "from": "tool:tensorrt",
      "to": "skill-area:ai-evaluation",
      "kind": "used_for"
    },
    {
      "from": "tool:tensorrt",
      "to": "tool:vllm",
      "kind": "alternative_to",
      "attributes": {
        "comparison": "ML inference — TensorRT is GPU model optimization; vLLM is LLM-focused with PagedAttention"
      }
    },
    {
      "from": "tool:tensorrt",
      "to": "tool:triton-inference",
      "kind": "alternative_to",
      "attributes": {
        "comparison": "NVIDIA inference — TensorRT optimizes models; Triton serves them"
      }
    },
    {
      "from": "tool:tensorrt",
      "to": "tool:onnx-runtime",
      "kind": "alternative_to",
      "attributes": {
        "comparison": "ML inference — TensorRT is NVIDIA GPU-only; ONNX Runtime is cross-platform"
      }
    }
  ],
  "incomingEdges": [
    {
      "from": "specialization:gpu-programming",
      "to": "tool:tensorrt",
      "kind": "uses_tool"
    },
    {
      "from": "tool:vllm",
      "to": "tool:tensorrt",
      "kind": "alternative_to",
      "attributes": {
        "comparison": "ML inference — vLLM is LLM-focused with PagedAttention; TensorRT is GPU model optimization"
      }
    },
    {
      "from": "tool:triton-inference",
      "to": "tool:tensorrt",
      "kind": "alternative_to",
      "attributes": {
        "comparison": "NVIDIA inference — Triton serves models; TensorRT optimizes them"
      }
    },
    {
      "from": "tool:onnx-runtime",
      "to": "tool:tensorrt",
      "kind": "alternative_to",
      "attributes": {
        "comparison": "ML inference — ONNX Runtime is cross-platform; TensorRT is NVIDIA GPU-only"
      }
    }
  ]
}