下载URL分别取回，保留同屏版本对应关系

{
  "inputRows": [
    {
      "artifact": "html",
      "method": "screen.getHtml()",
      "returns": "download URL"
    },
    {
      "artifact": "image",
      "method": "screen.getImage()",
      "returns": "download URL"
    },
    {
      "screenRoute": "/booking",
      "pairKey": "booking-reviewed-v1"
    }
  ],
  "referenceRows": [
    {
      "file": "booking.html",
      "mustContain": "<!doctype or <html structure"
    },
    {
      "file": "booking.webp/png",
      "mustMatch": "same screen text and layout"
    },
    {
      "failure": "HTML error response",
      "action": "reject as asset"
    }
  ],
  "invalidRows": [
    {
      "file": "index.html",
      "content": "https://example.invalid/download",
      "error": "URL saved as content"
    },
    {
      "pair": [
        "old HTML",
        "new screenshot"
      ],
      "error": "mismatched version"
    }
  ],
  "notes": [
    "全部输入为作者自制教学资料，不含真实客户或账号。",
    "referenceRows是校验基准，不是目标模型预测或识别结果；targetProduct=not_run。"
  ],
  "files": {
    "input.json": "[\n  {\n    \"artifact\": \"html\",\n    \"method\": \"screen.getHtml()\",\n    \"returns\": \"download URL\"\n  },\n  {\n    \"artifact\": \"image\",\n    \"method\": \"screen.getImage()\",\n    \"returns\": \"download URL\"\n  },\n  {\n    \"screenRoute\": \"/booking\",\n    \"pairKey\": \"booking-reviewed-v1\"\n  }\n]\n",
    "reference.json": "{\n  \"identity\": \"作者验收参考，非模型返回\",\n  \"rows\": [\n    {\n      \"file\": \"booking.html\",\n      \"mustContain\": \"<!doctype or <html structure\"\n    },\n    {\n      \"file\": \"booking.webp/png\",\n      \"mustMatch\": \"same screen text and layout\"\n    },\n    {\n      \"failure\": \"HTML error response\",\n      \"action\": \"reject as asset\"\n    }\n  ]\n}\n",
    "invalid.json": "{\n  \"identity\": \"作者构造负例，非实际错误回包\",\n  \"rows\": [\n    {\n      \"file\": \"index.html\",\n      \"content\": \"https://example.invalid/download\",\n      \"error\": \"URL saved as content\"\n    },\n    {\n      \"pair\": [\n        \"old HTML\",\n        \"new screenshot\"\n      ],\n      \"error\": \"mismatched version\"\n    }\n  ]\n}\n",
    "case.md": "# Stitch导出HTML和截图：下载URL分别取回，保留同屏版本对应关系\n\nStitch导出HTML和截图先分清返回值是下载地址。用相同screen的HTML和图片成对保存，检查内容类型及版本，防止把URL字符串写成index.html。\n\n## 锁定同一屏幕\n\n从项目读取目标screen，记录screenId和版本时间。\n\n不要一个方法用旧引用、另一个方法重新选最新页面。\n\n## 获取两种下载地址\n\n分别await screen.getHtml()和screen.getImage()。\n\n官方SDK返回URL，不是HTML字符串或图片字节。\n\n## 下载并检查类型\n\n在授权环境取回两个URL，将HTML和图片写到不同文件。\n\n若返回错误页或鉴权失败，保存状态并停止，不按扩展名伪装成功。\n\n## 检查HTML内容\n\n确认文件含HTML结构与页面文字，而非一行https链接。\n\n如果下载页面依赖外链资源，另外记录依赖，不默认单文件离线可用。\n\n## 配对截图和源码\n\n给两份文件相同screen键与取得时间，打开检查标题、布局和文案。\n\n图片与HTML不一致时先核对来源，不手工改图掩盖版本差异。\n\n## 处理过期资源\n\n地址失效时从同一目标screen重新读取新下载地址。\n\n不要把旧URL当永久公开资源写进长期教程素材。\n\n## 交付导出清单\n\n保存本地文件、内容类型和屏幕身份，删除日志中不必要的临时敏感参数。\n\n未执行下载时仍为待验证，不称源码已运行。\n\n身份：输入与参考为作者材料；localCli有实测记录时仅证明对应CLI运行，目标云端与模型推理未执行。\n",
    "download.ts": "import { stitch } from \"@google/stitch-sdk\";\nimport {writeFile} from \"node:fs/promises\";\nconst projectId=process.env.STITCH_PROJECT_ID;const screenId=process.env.STITCH_SCREEN_ID;\nif(!projectId||!screenId)throw new Error(\"Set actual bare project/screen IDs in environment\");\nconst screen=await stitch.project(projectId).getScreen(screenId);\nfor(const [kind,url] of [[\"html\",await screen.getHtml()],[\"image\",await screen.getImage()]] as const){\n const response=await fetch(url);if(!response.ok)throw new Error(`download ${kind} ${response.status}`);\n const type=response.headers.get(\"content-type\")??\"\";\n if(kind===\"html\"&&!type.includes(\"text/html\"))throw new Error(\"unexpected HTML content type\");\n if(kind===\"image\"&&!type.startsWith(\"image/\"))throw new Error(\"unexpected image content type\");\n await writeFile(kind===\"html\"?\"screen.html\":\"screen.image\",new Uint8Array(await response.arrayBuffer()));\n}\n"
  }
}