All-in-One Weibo Scraper

The Actor covers eight Weibo surfaces behind one mode selector: posts, their comments, their likers, profiles, an account's videos, account search, topic search, and the hot-search board. Pick the mode for the run and fill in the section belonging to it.

{
  "scraperType": "postDetail",
  "mid": "https://m.weibo.cn/detail/5145994595795680",
  "maxItems": 100
}

Modes and what each one needs

scraperType chooses the dataset and decides which of the fields below matters.

• Post — needs mid, a post ID or a link to the post. Returns one row: the text, publish time, client, images and the repost, comment and like counts. A video post also carries its stream URLs and duration.
• Comments — needs mid. Returns the comments on that post, one row each, with Weibo's own total for the thread.
• Likers — needs mid. Returns the accounts that liked the post, walking the liker list page by page. page sets which page it starts from.
• Profile — needs uid, a numeric user ID or a profile link. Returns one row: screen name, bio, location, gender, avatars, verification and its stated reason, and the follower, following and post counts.
• Account videos — needs uid. Returns the videos that account posted, walking the feed on its own cursor until it runs out.
• Account search — needs query. Returns matching accounts, one row each, in a single request.
• Topic search — needs query. Returns matching topics, walking the result pages. page sets which page it starts from.
• Hot search board — needs nothing at all. Returns the board as it stands right now, one row per entry.

maxItems applies to every mode and defaults to 100; 0 disables the bound.

Result fields

Rows differ by mode, and each mode fills only its own columns.

A post row carries mid, text, text_length, created_at, source, pic_ids, reposts_count, comments_count, attitudes_count and user, plus is_video, video_stream_url, video_stream_url_hd and video_duration_seconds. A profile row carries uid, screen_name, description, location, gender, verified, verified_type, verified_reason, followers_count, friends_count, statuses_count, profile_image_url, avatar_hd and status_total_counter.

The list modes — comments, likers, account videos, both searches, the board — pass each entry through exactly as the platform returns it, and add what the row was collected under: mid, uid or query, plus page where the mode pages by one.

count and total_number are not the same thing. count is how many rows came back on that page — a page size. total_number is Weibo's own catalogue total, and only the comment and liker lists carry one. If you want to know whether you have a whole thread, compare against total_number; count will always tell you that you do.

A Weibo video is a post

There is no separate video record on Weibo, and no mode here pretends there is. A video is a post that happens to have a video on it, so the Post mode is how you fetch one, and it returns is_video, the stream URLs and the duration on the post itself. Account videos is the bulk form of the same thing: one account's video feed rather than one video.

Where the paged modes stop

Only account videos paginates by cursor, and it is walked until a page comes back empty, with a cursor that stops moving as the backstop. Likers and topic search page by an ordinal page number and stop the same way, or when Weibo's own total_number has been reached.

Comments, account search and the board are one request each. Comments does return a field called next_cursor_hint, and a hint is precisely what it is: nothing says what to pass it back as and no second page has been seen returned from one, so this Actor does not spend your money guessing.

No mode stops on a has_more flag. On the sibling TikTok feed that flag stayed true through three consecutive empty pages with the cursor climbing 30, 60, 90, 120 — a run that trusted it would never stop, and would pay for the compute until the platform timed it out.

Empty is an answer, and a failure is a failure

A deleted post, an account that has been closed, a post nobody has liked, a search that matches nothing: all of these come back empty rather than as errors. The run succeeds, those rows are not saved, and nothing is charged for them.

A run that genuinely could not reach the backend fails loudly instead, and so does a response in a shape this Actor does not recognise. An empty dataset from this Actor always means "there was nothing there" — it never quietly stands in for a request that went wrong.

Related Actors

Six of the modes also ship on their own, for when a run only ever does one of them: the Weibo Post Scraper, the Weibo Comments Scraper, the Weibo Profile Scraper, the Weibo User Videos Scraper, the Weibo User Search Scraper and the Weibo Hot Search Scraper. Likers and topic search live only here.