douyin-search

name	douyin-search
description	Use Playwright to search Douyin (抖音) by keyword and extract video metadata (title, author, likes, duration, publish time, URL). Optionally write results to a Feishu (Lark) spreadsheet via lark-cli. Requires valid Douyin cookies for authenticated search. Use when user mentions "搜索抖音", "抖音搜索", "douyin search", "抓取抖音视频", "抖音爬虫", or asks to search/scrape videos from Douyin with a specific keyword.
metadata	{"author":"wen.yuan"}

Douyin Keyword Search Scraper

Search Douyin by keyword, extract video metadata, optionally export to Feishu.

Prerequisites

• playwright npm package installed (npm i playwright)
• Valid Douyin cookies (sessionid, etc.) — cookies expire and must be refreshed
• lark-cli on PATH (only if exporting to Feishu)

Usage

Run the bundled script:

node scripts/douyin_search.js <keyword> [options]

Options:

• --target N — max videos to collect (default: 50)
• --output PATH — JSON output file (default: douyin_search_<keyword>.json)
• --feishu — also write results to a Feishu spreadsheet
• --cookies PATH — path to a JSON file containing cookie array (see Cookie Config below)
• --headless — run in headless mode (default: headed)

Examples:

node scripts/douyin_search.js openclaw --feishu
node scripts/douyin_search.js "AI编程" --target 30 --output ai_videos.json
node scripts/douyin_search.js 短剧 --headless --feishu

Cookie Config

Create a JSON file (e.g. cookies.json) with the required Douyin cookies:

[
  { "name": "sessionid", "value": "<your_value>", "domain": ".douyin.com", "path": "/" },
  { "name": "sessionid_ss", "value": "<your_value>", "domain": ".douyin.com", "path": "/" },
  { "name": "ttwid", "value": "<your_value>", "domain": ".douyin.com", "path": "/" }
]

Minimum required cookies: sessionid, sessionid_ss. Include ttwid, passport_csrf_token, sid_guard if available.

If no --cookies flag is provided, the script checks for douyin_cookies.json in the current working directory.

Output

JSON format

Each entry contains:

• title — video title/description
• author — creator display name
• likes — like count (raw string, e.g. "3.8万", "183")
• duration — video length (e.g. "07:40")
• time — relative publish time (e.g. "2周前", "12小时前")
• videoId — Douyin video ID
• url — full video URL

Feishu spreadsheet

Columns: 视频标题 | 作者 | 点赞数 | 时长 | 发布时间 | 链接 Sheet title: 抖音搜索 - <keyword> (<date>)

Notes

• DOM selectors (.VDYK8Xd7, .z2lFLtJ0, .ckopQfVu, .dW_QmDH1) use obfuscated class names that may change with Douyin updates. If extraction returns empty results, inspect the page structure and update selectors in scripts/douyin_search.js.
• Scroll pagination: the script scrolls the search results page to load more videos. Rate-limited with 2-3.5s random delays.
• The script runs in headed mode by default for debugging visibility.