{
  "schemaVersion": "1.0",
  "updates": [],
  "companyNews": [],
  "id": "b918",
  "language": "zh",
  "name": "Unstructured",
  "originalName": "Unstructured",
  "library": "tools",
  "category": "模型开发",
  "categoryKey": "模型开发",
  "categoryKeys": [
    "模型开发"
  ],
  "categories": [
    "模型开发"
  ],
  "entryPoints": [],
  "summary": "用开源 Python 组件将 PDF、图像与办公文档拆为结构化元素，为大模型检索和知识库构建准备输入。",
  "introduction": "此条目收录 Unstructured 的开源文档预处理库：按文件类型读取文档，结合相应解析、版面识别与 OCR 组件，把内容转换为带结构的元素，用于大模型检索和知识库流水线。它需要开发环境及按格式安装的依赖。开源库与托管 Transform 服务能力不同，不能把托管服务最新模型的效果直接归给开源版。库默认有轻量遥测，可按官方说明关闭。",
  "capabilities": [],
  "company": {
    "details": {
      "facts": [
        {
          "key": "website",
          "label": "公司官网",
          "value": "https://unstructured.io/"
        },
        {
          "key": "focus",
          "label": "主要业务",
          "value": "提供面向 AI 应用的文档转换与数据预处理技术，同时维护开源库和商业托管产品。"
        }
      ],
      "otherProducts": []
    },
    "name": "Unstructured Technologies, Inc.",
    "originalName": "Unstructured Technologies, Inc.",
    "country": {
      "codes": [
        "US"
      ],
      "names": [
        "美国"
      ],
      "display": "美国",
      "basis": "headquarters",
      "basisLabel": "公司总部"
    }
  },
  "officialUrl": "https://github.com/Unstructured-IO/unstructured",
  "profileUpdatedOn": "2026-10-21",
  "sources": [
    {
      "kind": "product",
      "url": "https://github.com/Unstructured-IO/unstructured"
    },
    {
      "kind": "documentation",
      "url": "https://github.com/Unstructured-IO/unstructured/blob/main/README.md"
    }
  ],
  "news": [],
  "links": {
    "interactive": "products/unstructured--b918.html",
    "reading": "read/zh/products/unstructured--b918.html",
    "markdown": "read/zh/products/unstructured--b918.md",
    "json": "read/zh/products/unstructured--b918.json"
  }
}