Douyin Search and Hot List Data Assistant

Problem

Data analysis work often requires extracting Douyin video titles, authors, descriptions, play counts, likes, and comments from search results or hot lists. Plain HTTP calls are not always enough because the search page depends on mobile rendering and a dynamic layout, while hot list data may come from a separate public endpoint. Without a stable scraping flow, it is hard for an agent to turn a natural-language request into reusable structured results.

How it works

The skill follows a simple path: natural-language query, keyword extraction, browser scraping or API call, and structured output. For search, it uses Playwright in mobile mode to load douyin.com/search/ and extracts fields such as title, description, author, play count, likes, and comments from the rendered page. For hot lists, it calls the public endpoint douyin.com/aweme/v1/web/hot/search/list/ without login. It can map Chinese requests such as “search for seafood videos” to the keyword “seafood”, then run the search and export results as JSON or CSV.

Limitations

This skill is intended for learning and small-scale research, not high-frequency bulk scraping. Search requires Playwright and Chromium; Douyin may present captchas, and repeated requests can trigger rate limits or account restrictions, so request delays should be used. If the page structure changes, empty search results may require checking selectors, network connectivity, and delay settings. If the hot list endpoint changes, the request path or field mapping may need updating.