{
  "recordedAt": "2026-09-07",
  "sourceType": "user_confirmed_project_convention",
  "userDecision": "所有按美元额度和三段价格折算token的数据统一使用一个项目标准，不再称为Kimi mix，也不声称代表Command Code、Ollama或任何平台的实际负载。",
  "standardTokenMix": {
    "cache": 0.975,
    "input": 0.0215,
    "output": 0.0035,
    "sum": 1.0
  },
  "labelZh": "项目统一标准负载：缓存读取97.5% / 普通输入2.15% / 输出0.35%",
  "scope": [
    "美元额度除以三段价格：OpenCode Go、Command Code GOAT、Ollama Pro/Max",
    "按量API混合单价",
    "同池模型按三段价格比派生的额度",
    "官方credits三段费率的模型间换算"
  ],
  "interpretation": [
    "这是固定比较基准，不是任何供应商或用户的实际缓存命中率、输入比例或输出比例。",
    "所有参与比较的渠道使用同一组成，避免为每个平台任意选择有利的负载分布。",
    "平台实测token总量或面板反推值不经过此标准重算。",
    "若未来改标准，应统一重算全部受影响点，不单独调整某一供应商。"
  ],
  "implementation": {
    "singleSource": "data/conventions.json#standardTokenMix",
    "consumers": [
      "scripts/build_adopted.py",
      "scripts/compute.py",
      "scripts/build_html.py",
      "scripts/plot_quotas.py"
    ]
  },
  "supersedesWording": [
    "本机Kimi MIX",
    "实测agent负载分布作为跨平台事实",
    "observed token mix"
  ]
}