qData 2.3.0：主数据、数据集市与非结构化元数据，三把钥匙解开数据中台落地死结

# master_data_customer.yaml — 客户主数据标准模型
apiVersion: qdata.mdm/v2
kind: MasterDataModel
metadata:
  name: customer-standard
  displayName: 客户主数据标准模型
  version: "1.0.0"
  description: 全企业统一的客户核心属性定义
spec:
  entity:
    codeRule: "CUS-{YYYYMMDD}-{SEQ:4}"  # 编码规则：CUS-20250710-0001
    attributes:
      - name: customer_id
        type: string
        length: 20
        required: true
        unique: true
        displayName: 客户统一编码
      - name: customer_name
        type: string
        length: 100
        required: true
        displayName: 客户名称
      - name: industry
        type: enum
        enumValues: [制造业, 金融, 互联网, 医疗, 教育, 其他]
        required: true
        displayName: 行业分类
      - name: region
        type: string
        length: 50
        displayName: 所属区域
      - name: credit_level
        type: enum
        enumValues: [A, B, C, D]
        displayName: 信用等级
  # 多源映射：将异构系统字段对齐到标准属性
  sourceMappings:
    - sourceSystem: CRM
      mappings:
        - sourceField: customer_id    # CRM 里的字段名
          targetAttribute: customer_id
        - sourceField: company_name
          targetAttribute: customer_name
        - sourceField: sector
          targetAttribute: industry
    - sourceSystem: ERP
      mappings:
        - sourceField: kunnr          # SAP 客户编号
          targetAttribute: customer_id
        - sourceField: name1
          targetAttribute: customer_name
        - sourceField: bztxt
          targetAttribute: industry
  # 入库质量规则
  qualityRules:
    - name: uniqueness_check
      type: unique
      on: [customer_id]
      action: reject                  # 重复编码直接拦截
    - name: completeness_check
      type: notNull
      on: [customer_id, customer_name, industry]
      action: quarantine              # 缺必填项进入隔离区待人工处理

# data_mart_customer.yaml — 客户主题数据集市
apiVersion: qdata.mart/v2
kind: DataMart
metadata:
  name: customer-mart
  displayName: 客户主题集市
  owner: data-governance-team
  version: "1.0.0"
spec:
  subjectDomain: 营销与客户管理
  description: 包含客户主数据、交易汇总、信用评分等指标
  datasets:
    - name: customer_master
      source: masterData:customer-standard   # 引用上面定义的主数据模型
      refresh: daily
    - name: customer_transaction_summary
      source: dw:fact_transaction
      grain: customer_id, month
      refresh: daily
  indicators:
    - name: customer_count
      displayName: 客户总数
      measureType: count
      grain: [region, industry]
      datasetRef: customer_master
    - name: avg_monthly_revenue
      displayName: 客户月均收入
      measureType: avg
      measureField: revenue
      grain: [region, industry, credit_level]
      datasetRef: customer_transaction_summary
     口径说明: "单客户近12个月交易金额均值"
  access:
    roles:
      - name: marketing-analyst
        permissions: [read, subscribe]
      - name: data-engineer
        permissions: [read, write, subscribe]
    subscription:
      enabled: true
      channels: [email, webhook]

"""
演示：通过 qData SDK 为一批合同 PDF 提取元数据并打标签
运行前：pip install qdata-sdk
配置环境变量：QDATA_HOST, QDATA_TOKEN
"""
from qdata_sdk import UnstructuredClient, TagSet

client = UnstructuredClient()

# 1. 批量提取 PDF 元数据
pdf_paths = ["contracts/2024/供应链协议A.pdf",
             "contracts/2024/采购合同B.pdf"]

for path in pdf_paths:
    meta = client.extract_metadata(
        file_path=path,
        extract_text=True,          # 提取关键文本段落
        extract_tags=True,          # 自动推断标签
        ocr_enabled=True            # 扫描件启用 OCR
    )
    print(f"文件: {path}")
    print(f"  关键文本片段: {meta.text_snippets[:2]}")
    print(f"  自动标签: {meta.auto_tags}")

# 2. 定义业务标签体系并手动补充
tag_set = TagSet(name="合同分类体系")
tag_set.add_category("合同类型", values=["采购", "销售", "供应链", "租赁"])
tag_set.add_category("风险等级", values=["低", "中", "高"])
tag_set.add_category("涉及业务域", values=["供应链", "财务", "法务"])
client.publish_tag_set(tag_set)

# 3. 为指定文件打上业务标签
client.tag_file(
    file_path="contracts/2024/供应链协议A.pdf",
    tags={"合同类型": "供应链", "风险等级": "中", "涉及业务域": ["供应链", "法务"]}
)

# 之后在统一目录中搜索：
results = client.search("合同类型=供应链 AND 风险等级=中")
for r in results:
    print(r.file_path, r.tags, r.text_snippets[0][:80])