API

Overview

RuntimeDog provides a simple JSON API for programmatic access to runtime data. The API returns static JSON and does not require authentication.

Endpoint

Returns a JSON object containing all runtimes with their metadata and scores.

Response Format

{
  "version": "1.0",
  "generated": "2026-01-18T00:00:00.000Z",
  "count": 10,
  "runtimes": [
    {
      "id": "wasmtime",
      "name": "Wasmtime",
      "tagline": "Fast, secure WebAssembly runtime",
      "type": "wasm",
      "execution": "aot",
      "interface": "cli",
      "languages": ["Rust", "C", "C++", ...],
      "isolation": "process",
      "maturity": "production",
      "performance": {
        "cold_start_ms": 1,
        "memory_mb": 5,
        "startup_overhead_ms": 0.5
      },
      "score": 88,
      "license": "Apache-2.0",
      "website": "https://wasmtime.dev",
      "github": "https://github.com/...",
      "docs": "https://docs.wasmtime.dev"
    },
    ...
  ]
}

Fields

Usage Example

// Fetch all runtimes
const res = await fetch('https://runtimedog.com/api/runtimes.json');
const data = await res.json();

// Filter by type
const wasmRuntimes = data.runtimes.filter(r => r.type === 'wasm');

// Sort by score
const topRated = data.runtimes.sort((a, b) => b.score - a.score);

🏠 Local LLM API

Dedicated endpoints for local LLM tools and stack recommendations.

GET /api/local-llm.json

Returns all local LLM tools (launchers, engines, formats, backends).

{
  "count": 25,
  "runtimes": [
    {
      "id": "ollama",
      "name": "Ollama",
      "role": "launcher",
      "localFitScore": 95,
      "backends": ["cuda", "metal", "rocm", "cpu"],
      "formats": ["gguf"],
      "install": { "mac": "brew install ollama", ... },
      ...
    },
    ...
  ]
}

GET /api/local-stacks.json

Returns pre-configured stacks by hardware target (NVIDIA/Mac/CPU/AMD).

{
  "count": 4,
  "targets": ["nvidia", "mac", "cpu", "amd"],
  "stacks": [
    {
      "target": "nvidia",
      "description": "NVIDIA GPU users...",
      "bands": [
        {
          "vram_gb": 8,
          "label": "8GB VRAM (RTX 3060/3070)",
          "recipes": [
            {
              "name": "Beginner",
              "launcher": "ollama",
              "engine": "llama.cpp",
              "formats": ["gguf"],
              "quant_hint": "Q4_K_M",
              "install_steps": ["curl ...", "ollama pull ..."],
              "notes": "7Bモデル快適"
            },
            ...
          ]
        },
        ...
      ]
    },
    ...
  ]
}

Score Philosophy

RuntimeDog scores different item classes using different criteria. This ensures fair comparison within each category.

Runtime Score (0-100)

For runtimes (Node.js, Wasmtime, Docker, Firecracker, etc.), the score reflects:

• Performance — Cold start time, memory footprint, startup overhead
• Isolation — Security boundary strength (process/sandbox/VM)
• Maturity — Production readiness, ecosystem, documentation

AI Tool Score (0-100)

For AI tools (Ollama, LM Studio, vLLM, etc.), the score reflects:

• Install Ease — One-command install, minimal dependencies
• Backend Support — GPU/CPU coverage (CUDA, Metal, ROCm, CPU)
• Operability — API compatibility, model management, updates

Formats & Backends

Model formats (GGUF, ONNX, SafeTensors) and GPU backends (CUDA, Metal, Vulkan) are not scored. These are infrastructure components—choosing one depends on your hardware and toolchain, not a quality ranking.

Notes

• No authentication required
• No rate limits (please be reasonable)
• Data is updated periodically
• CORS enabled for browser access