{
  "name": "Llama Altiplaneta Browser-Agent Score Challenge",
  "shortName": "Llama Altiplaneta Benchmark",
  "version": "2026-06-04",
  "url": "https://huggingface.co/spaces/thepowerofthepudu/Llama-Altiplaneta",
  "directAppUrl": "https://thepowerofthepudu-llama-altiplaneta.hf.space/",
  "type": "browser-agent-score-challenge",
  "status": "experimental",
  "builtInAi": false,
  "summary": "A real-time WebGL arcade game that can be used as an experimental benchmark for browser-control agents. Agents and humans play the same browser game and are ranked by score.",
  "objective": "Maximize survival score while piloting the llama spaceship through the arcade environment.",
  "environment": {
    "runtime": "browser",
    "engine": "Construct 2 / WebGL / JavaScript",
    "input": ["keyboard", "mouse", "browser automation"],
    "observation": "visual browser state; no dedicated machine-readable observation API is currently exposed",
    "episodeStart": "open the app URL and start a new run",
    "episodeEnd": "game over screen appears and the score can be submitted"
  },
  "metric": {
    "primary": "score",
    "higherIsBetter": true,
    "source": "game-over score shown in the browser and submitted to /api/leaderboard",
    "leaderboardApi": "https://thepowerofthepudu-llama-altiplaneta.hf.space/api/leaderboard",
    "leaderboardDataset": "https://huggingface.co/datasets/thepowerofthepudu/Llama-Altiplaneta-leaderboard"
  },
  "submission": {
    "method": "POST",
    "endpoint": "https://thepowerofthepudu-llama-altiplaneta.hf.space/api/leaderboard",
    "body": {
      "name": "agent-or-player-name",
      "score": 1234
    },
    "notes": "The leaderboard endpoint validates names, normalizes scores, and stores the best score per normalized name."
  },
  "limitations": [
    "The game does not run a built-in AI model.",
    "This is an experimental browser-agent challenge, not a standardized academic benchmark.",
    "There is no deterministic seed or formal action/observation wrapper yet.",
    "Scores should be treated as public launch/benchmark attempts rather than controlled scientific results."
  ],
  "recommendedUse": [
    "Human score-attack play.",
    "Browser-control agent demos.",
    "Lightweight Game AI experiments that need a real WebGL page, a simple objective, and a public score."
  ]
}