Platform And Field Mapping

Whale serves cross-platform public content data. Source platforms have different raw structures, but the external query layer stays unified: consumers query by platform, content type, time, keyword, semantics, and structured fields without knowing how each platform is collected.

Section	Description
Platform scope	Frozen platform enums for the first version
Content types	External unified `content_type` values
Common fields	Stable cross-platform fields
Platform cases	Typical platform data shapes
Metric semantics	How views, reads, likes, comments, and other metrics map
Extension fields	How platform-specific fields are preserved

Initial Platform Scope

Group	Platforms
Chinese content platforms	`douyin`, `kuaishou`, `bilibili`, `wechat_channels`, `xiaohongshu`, `weibo`, `wechat_article`, `zhihu`
Chinese marketplace case	`xianyu`
Global content platforms	`reddit`, `medium`, `linkedin`, `instagram`, `tiktok`, `youtube`
Discovery and web	`google_search`, `google_news`, `google_trends`, `website`, `rss`, `news`

Availability depends on data-product authorization, collection coverage, and compliance policy. API responses use source_platform as the stable platform enum.

Content Types

content_type	Meaning	Platform examples
`post`	Social post, feed item, short content	Weibo, Xiaohongshu, Reddit
`article`	Long article, public account article, column	Medium, WeChat Articles, Zhihu
`video`	Video content, possibly with description, transcript, or audio transcription	YouTube, TikTok, Bilibili, WeChat Channels
`comment`	Comment, reply, thread item	Reddit, Weibo, Xiaohongshu
`product`	Product or marketplace listing	Xianyu
`job_posting`	Job posting	LinkedIn
`web_page`	Web page, news page, RSS article	website, news, rss

Common Fields

Field	Meaning
`document_id`	Whale document ID used for details, versions, and snapshots
`dataset_id`	Data product or data domain, such as social, ecommerce, or recruitment
`source_platform`	Platform enum
`external_content_id`	Stable source-platform content ID
`canonical_url`	Canonical content URL
`content_type`	Unified content type
`author`	Author, account, channel, seller, or company identity
`content`	Title, body summary, URL, publish time, and other standard content
`metrics`	Current projection of engagement and distribution metrics
`parts`	Transcripts, OCR, chapters, specs, comment slices, and other multi-part content
`facts`	Filterable structured facts such as price, company, location, and skills
`platform_payload`	Platform-specific fields

Platform Cases

Platform	External content shape	Key fields	Notes
YouTube	`video`	Title, description, channel, publish time, views, likes, comments, transcript or audio transcription	If a video has no body, `parts.transcript` or `parts.audio_transcript` can still be searched
TikTok / Douyin / Kuaishou	`video`	Description, author, topics, plays, likes, comments, shares	Short videos may have no title; searchable text often comes from description, OCR, or transcription
Bilibili / WeChat Channels	`video`	Title, description, creator/account, plays, likes, favorites, danmaku	Platform-specific metrics remain in extension metrics
Xiaohongshu / Instagram	`post`	Title/body, images, OCR, author, likes, favorites, comments	Image text can be indexed as `parts.ocr`
Weibo / Reddit	`post` or `comment`	Body, author, publish time, reposts, comments, likes	Suitable for topic, event, and sentiment tracking
Medium / WeChat Article / Zhihu	`article`	Title, body, author, publish time, reads/likes/comments	Long text is chunked for search; detail APIs return full content
Xianyu	`product`	Product title, price, want count, area, status, seller information	Price, status, and seller sold/onsale counts are useful snapshot trends
LinkedIn	`job_posting`	Job title, company, location, salary, skills, work mode	Structured fields enter `facts` for filtering and aggregation
Google / website / RSS / news	`web_page` or `article`	Title, summary, URL, source site, publish time	Suitable for web discovery, news tracking, and topic monitoring

Metric Semantics

Standard metric	Meaning	Notes
`view_count`	Views, reads, or impressions	Names vary by platform; externally means “viewed”
`play_count`	Video plays	Preferred for video platforms
`like_count`	Likes	Common across social platforms
`comment_count`	Comments	Common across posts, videos, and articles
`share_count`	Shares	Empty when unsupported by the platform
`collect_count`	Favorites, saves, or wants	Similar but not identical across Xiaohongshu, Bilibili, Xianyu
`repost_count`	Reposts	Used by platforms such as Weibo
`author_follower_count`	Author follower snapshot	Author state, not body content

Time-varying metrics are stored as snapshots. Consumers that need trends should read document snapshots instead of relying only on the current projection returned by document detail.

Extension Fields

Stable cross-platform fields enter the standard model first. Fields that cannot be normalized reliably stay in platform_payload or platform_metrics. Examples:

Bilibili danmaku and coins.
Xianyu seller credit, Zhima credit, product status.
LinkedIn application count, job level, remote work mode.
Google Trends interest and regional distribution.

Usage Guidance

For cross-platform search, start with content_type, source_platform, published_at, and keyword/semantic query.
For platform-specific analysis, read facts, platform_payload, and platform_metrics.
For trends, read snapshots sorted by snapshot_at.
For full body, transcripts, or specs, load document content, parts, and facts.