Friends Data API独立数据服务

DATASET · research

事件优先级训练标签

research.event_priority_train_label
研究成品

按事件和标签版本保存人工或规则生成的优先级标签,用于训练与评估事件排序模型。

字段19
关键字段2
更新口径weekly
时间列event_ts
单页上限5000
分页窄筛选
PIT不支持
规模10K-99K

WHAT IT ANSWERS

这张表适合回答什么

  • 训练事件优先级模型
  • 比较标签版本
  • 联接 event.stream 获取事件特征

GRAIN & TIME

一行代表什么

业务粒度
每行代表“事件优先级训练标签”的一条记录;逻辑粒度由 event_id × label_version 共同确定。
时间口径
以 event_ts 为主要查询时间轴;除下方限制特别说明外,时间戳按 UTC、日期按来源市场日历解释。
更新节奏
通常每周随上游披露或批处理更新;不是实时数据。

KEY FIELDS

先看这些字段

event_id
统一事件标识;可联接 event.stream。
label_version
用于确定该表业务粒度或默认顺序的 label version 字段。

RECOMMENDED JOINS

怎样和其他表关联

event.streamon event_id

用 event_id 回到统一事件流和事件分类。

HOW TO QUERY SAFELY

查询方法与已知限制

先用 catalog 确认当前契约;可按 catalog 中列出的 filters 进一步缩小范围;单页最多 5000 行,该表不提供稳定 cursor,结果过大时必须缩小标的、日期或其他筛选。

MACHINE CONTRACT

筛选、排序与分页

UNVERSIONED
默认排序
event_ts, event_id, label_version
唯一键
—
读取模式
canonical_final_active
最大时间窗
未单独限制

可用 filters

event_id→ event_id · string
label_version→ label_version · string
source→ source · string

POINT IN TIME

可复现读取

NOT AVAILABLE

该资产没有 PIT 契约。普通 rows 端点仍按当前公开读模式返回数据;不要把它表述成历史时点快照。

COPY-READY CALLS

怎么调用

同一数据集可用 HTTP、Python SDK、CLI 或 MCP;Key 始终只通过环境变量/Bearer Header 传递。

READ ONLY
HTTP
curl --fail-with-body -H "Authorization: Bearer $FRIENDS_DATA_API_KEY" "https://api.alphahubs.uk/api/v1/data/datasets/research/event_priority_train_label/rows?limit=100&source=SOURCE_VALUE&start=2026-08-01&end=2026-08-01"

FIELD DICTIONARY

19 个返回字段

descriptionSource 可在 JSON 中区分仓库注释、标准术语表和名称推导说明。

#字段ClickHouse 类型JSON可空角色说明
1event_idStringstringnofilter, default_order, warehouse_primary_keyStable identifier of the normalized event.标准术语表
2event_tsUTC_datetimeDateTime64(3, 'UTC')string / date-timenotime, default_orderUTC timestamp of the normalized event.标准术语表
3label_versionLowCardinality(String)stringnofilter, default_order, warehouse_primary_keySource-normalized field: label version.名称推导
4labelNullable(Int16)integeryesvalueSource-normalized field: label.名称推导
5label_sourceLowCardinality(String)stringnovalueSource-normalized field: label source.名称推导
6sourceLowCardinality(String)stringnofilter, warehouse_primary_key, provenanceNormalized upstream source identifier.标准术语表
7providerLowCardinality(String)stringnoprovenanceUpstream data provider identifier.标准术语表
8source_urlStringstringnoprovenanceURL of the upstream evidence or record when distributable.标准术语表
9source_record_idStringstringnoprovenanceProvider-side identifier for the source record.标准术语表
10raw_payload_hashStringstringnoprovenanceOpaque fingerprint of the normalized source payload for provenance checks.标准术语表
11parser_versionLowCardinality(String)stringnoprovenanceVersion of the parser or normalizer that produced the row.标准术语表
12quality_statusLowCardinality(String)stringnoprovenanceMachine-readable data-quality classification.标准术语表
13license_scopeLowCardinality(String)stringnoprovenanceLicense or redistribution scope attached to the row.标准术语表
14visibility_tierLowCardinality(String)stringnoprovenanceServing visibility tier attached to the row.标准术语表
15ingest_run_idStringstringnoprovenanceIdentifier of the ingestion run that produced the row.标准术语表
16versionUInt64integernoprovenanceMonotonic row version used for correction-aware reads.标准术语表
17is_deletedUInt8integernovalueSoft-delete marker; public serving excludes current tombstones.标准术语表
18created_atUTC_datetimeDateTime64(3, 'UTC')string / date-timenovalueUTC timestamp when the warehouse row was first created.标准术语表
19updated_atUTC_datetimeDateTime64(3, 'UTC')string / date-timenovalueUTC timestamp when the warehouse row was last updated.标准术语表
字典口径

结构来自 Friends 只读身份可见的真实 serving relation;权限、过滤、排序、PIT 与分页来自可执行 DatasetSpec。Schema observed at 2026-10-06T18:08:52Z。