Naver Cafe Search Scraper

Collect public metadata from Naver's Cafe search tab. The Actor does not sign in, open private posts, or bypass Cafe membership controls. It reads only information that Naver exposes on the public search-results page: article title and canonical URL, snippet, Cafe identity, date label, thumbnail, ranking, query, and scrape provenance.

Input

Use query for one search or queries for a batch. Duplicate and blank queries are removed. maxItems is a global output limit shared fairly across queries. maxPages limits traversal for every query, maxConcurrency bounds browser parallelism, and maxRequestRetries bounds retry cost. Direct access is the default QA configuration; an authorized Apify Proxy configuration can be supplied when the target blocks the execution environment.

{
  "queries": ["인공지능", "로봇"],
  "maxItems": 20,
  "maxPages": 2,
  "maxConcurrency": 2,
  "proxyConfiguration": { "useApifyProxy": false }
}

Output

Each dataset item is one public Cafe article search result. id is query-aware so the same article can legitimately appear for two requested queries, while articleKey remains stable across queries. position is global output order, queryPosition is rank within the query, and page records the Naver result page. url and link are canonical public article URLs with transient tracking tokens removed. articleId and cafeSlug are parsed from that URL. Optional source-backed fields are omitted rather than filled with guesses.

{
  "id": "naver-cafe:example:123:q:%EC%9D%B8%EA%B3%B5%EC%A7%80%EB%8A%A5",
  "articleKey": "example:123",
  "articleId": "123",
  "cafeSlug": "example",
  "position": 1,
  "queryPosition": 1,
  "page": 1,
  "title": "Public Cafe article title",
  "url": "https://cafe.naver.com/example/123",
  "snippet": "Public search snippet",
  "cafeName": "Example Cafe",
  "cafeUrl": "https://cafe.naver.com/example",
  "dateRaw": "2일 전",
  "publishedAt": "2026-08-28T00:00:00.000Z",
  "query": "인공지능",
  "sourceDomain": "search.naver.com",
  "scrapedAt": "2026-08-30T00:00:00.000Z"
}

Relative Korean date labels are converted against scrape time and are approximate. The original label is preserved in dateRaw.

Reliability and responsible use

The Actor uses Naver's current ssc=tab.cafe.all route and real page offsets, canonicalizes article identity, stops repeated result pages, checks HTTP/content type and visible challenge markers, and fails closed on selector drift. It never stores CAPTCHA or access-denial pages as data. Limits are reserved before asynchronous writes, preventing concurrent queries from exceeding maxItems.

Run locally with apify run --purge --input-file qa-inputs/local-search.json. Use npm test, npm run check, apify validate-schema, and node validate-datasets.js for verification. Website markup and access policy may change; keep request volumes modest and comply with Naver's terms, applicable law, and the privacy expectations of Cafe authors.