AI Integration Reference

Agent Navigation Map

GearVerify exposes 6 primary tools accessible to AI agents. Each endpoint returns real-time hardware diagnostic data rendered in the browser via WebGPU — no server involvement.

GPU Diagnostic Output Schema

The GPU lab exports the following structured object when a test completes. This is the data passed to the certificate engine and can be captured by observing window.GpuLab.hardware after initialisation.

{
  "vendor":          "NVIDIA",
  "description":     "RTX 3080",           // WebGL UNMASKED_RENDERER
  "deviceId":        "0x2206",             // PCI HW Device ID (silicon-level)
  "architecture":    "ampere",             // WebGPU adapter.info.architecture
  "vram":            "10GB GDDR6X",
  "vramGB":          10,
  "release":         "Sep 2020",
  "age":             "5 Years",
  "ageYears":        5,
  "benchmarkScore":  26085,                // PassMark G3D Mark
  "benchmarkName":   "PassMark G3D Mark",

  // ─── Stress Test Results (post-run) ──────────────────
  "gpu_integrity":   "pass",               // "pass" | "fail" | "degraded"
  "vram_artifacts":  false,
  "tdr_events":      0,                    // Timeout Detection & Recovery count
  "avg_fps":         87,
  "peak_load_pct":   98.4,
  "est_temp_c":      72,
  "duration_s":      600,
  "bios_spoof_risk": false,               // true if DevID ≠ reported model
  "validation_id":   "A7F3X9K"            // Random alphanumeric cert ID
}

CPU Diagnostic Output Schema

Accessible via window.CpuLab.hardware after initialisation.

{
  "name":              "Intel Core i7-12700K",
  "vendor":            "Intel",
  "cores":             12,
  "threads":           20,
  "architecture":      "Alder Lake",
  "release":           "Oct 2021",

  // ─── Stress Test Results ─────────────────────────────
  "cpu_integrity":     "pass",               // "pass" | "fail" | "throttling"
  "thermal_throttle":  false,
  "avg_ops":           4820,                 // million ops/sec
  "peak_load_pct":     97.2,
  "est_temp_c":        68,
  "duration_s":        300,
  "validation_id":     "B2M8P4Q"
}

Display Diagnostic Output Schema

{
  "resolution_native":  "3840x2160",         // screen.width × devicePixelRatio
  "resolution_css":     "1920x1080",         // CSS viewport pixels
  "device_pixel_ratio": 2,
  "colour_depth":       24,
  "refresh_rate_hz":    144,
  "hdr_capable":        true,
  "touch_points":       0,
  "display_integrity":  "pass"
}

Certificate Data Payload

Passed to ReportEngine.shareReport('GPU', data) to generate the A4 PNG certificate.

{
  "model":       "RTX 3080",
  "vendor":      "NVIDIA",
  "vram":        "10GB GDDR6X",
  "release":     "Sep 2020",
  "deviceId":    "0x2206",            // Silicon Signature (DevID)
  "avgMetric":   "87",                 // Avg FPS (GPU) or Avg OPS (CPU)
  "load":        98,                   // percentage
  "temp":        72,                   // °C estimated
  "duration":    600,                  // seconds
  "history":     [82, 87, 91, 88, ...], // FPS timeline array
  "validationId": "A7F3X9K"
}

WebGPU Requirements

GearVerify's GPU lab requires the following hardware-level browser API access to function. These are automatically requested on page load with no user interaction needed.

• 🖥️

  navigator.gpu — WebGPU API

  Required for compute shader execution (WGSL), adapter info enumeration, and GPU compute pipeline creation. Available in Chrome 113+, Edge 113+. The lab falls back to WebGL 2 if unavailable.

• 🔑

  adapter.requestAdapterInfo() — Adapter Metadata

  Reads vendor, architecture, device, and description from the GPU adapter. Used for silicon signature detection and BIOS spoof checking. Requires no special permission prompt.

• 🧮

  WEBGL_debug_renderer_info — WebGL Extension

  Retrieves the unmasked GPU renderer string (e.g. "NVIDIA GeForce RTX 3080/PCIe/SSE2"). Used as the primary hardware identification method on all browsers. No user prompt required.

• 📱

  window.devicePixelRatio — DPR API

  Used by the Display Lab to calculate the true native resolution from CSS viewport pixels. Multiplied by screen.width and screen.height to get physical pixel dimensions.

• ⚡

  Web Workers — Parallel CPU Stress

  The CPU Lab spawns multiple Worker threads equal to navigator.hardwareConcurrency to simulate max-load CPU conditions for thermal and throughput testing.

• 🎨

  OffscreenCanvas — Certificate Rendering

  The report engine uses Canvas 2D API to render the A4 certificate at 150 DPI (1240×1754px). The result is exported as a PNG data URL for download. No server upload occurs.

AI & Crawler Meta Signals

All GearVerify tool pages emit the following machine-readable signals to assist AI crawlers, LLM training pipelines, and automated agent frameworks.

<!-- AI Agent Capability Signal -->
<meta name="ai-agent-capability"
      content="browser-based-hardware-diagnostics">

<!-- Canonical URL -->
<link rel="canonical" href="https://gearverify.com/gpu/">

<!-- AXO Semantic Manifest (JSON article feed) -->
<link rel="alternate" type="application/json"
      href="/content.json">

<!-- JSON-LD: SoftwareApplication -->
<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "SoftwareApplication",
  "name": "GearVerify Hardware Diagnostic Suite",
  "applicationCategory": "Hardware Utility",
  "operatingSystem": "WebBrowser",
  "offers": {
    "@type": "Offer",
    "price": "0",
    "priceCurrency": "USD"
  }
}
</script>

Crawlability & Discoverability

GearVerify is fully crawlable by all major search engines and AI bots.