RSS News Digest – Keyword Alerts, Duplicates Merged

Every morning, just the new stories you care about – once, even when five sites report them.

Give it RSS / Atom feeds (news sites, ministries, competitors' blogs, release notes) and optional keywords, then schedule it daily. Each run returns:

• only items you have not received before (remembered in a named key-value store in your account)
• only items matching your keywords – with matching that understands Japanese:
  • full / half width: ＡＩ = AI, ｻｰﾊﾞ = サーバ
  • katakana / hiragana and long vowels: サーバー = サーバ = さーば
  • synonyms on one line: AI|人工知能|生成AI
  • English terms match whole words, so AI does not fire on "email" or "available"
• the same story from several feeds merged into one row, with the other sources listed in alsoReportedBy (title similarity, works for Japanese and English)

Pair it with Google Sheets Sync (append mode) or an email / Slack integration and the digest lands where your team reads.

Input

{
  "feedUrls": ["https://www.digital.go.jp/rss/news.xml", "https://example.com/feed"],
  "keywords": ["AI|人工知能|生成AI", "マイナンバー"],
  "excludeKeywords": ["PR"],
  "maxAgeHours": 48
}

Output (one row per story)

{
  "title": "生成AIの新サービスを発表",
  "link": "https://news.example/a/123",
  "published": "2026-10-26T00:00:00.000Z",
  "source": "Example News",
  "summary": "企業向けに…",
  "matchedKeywords": ["AI"],
  "alsoReportedBy": [{ "source": "Other News", "title": "生成ＡＩの新サービス発表、企業向け", "link": "https://other.example/n/9" }]
}

A quiet day returns one row status: "no-new-items". The OUTPUT record has counts (fetched, too old, seen before, not matching, merged) and per-feed errors.

Fair use

• Only the feed URLs you enter are fetched – once per run, at least 1 second apart per site. robots.txt is respected; a disallowed feed is reported and skipped. The tool identifies itself as RssNewsDigest/1.0.
• Only what the feed itself publishes is used: title, link, date and the first 200 characters of the summary. Article pages are not downloaded. Keep the source and link when you republish anything, and follow each site's terms.

Pricing

Pay per event, platform usage included:

Event	Price
Actor start	$0.01 per run
item-delivered – one new story delivered	$0.50 per 1,000 items

Example: 10 feeds checked daily with 20 matching stories a day ≈ 30 × ($0.01 + 20 × $0.0005) = $0.60 per month. Filtered-out, already-seen and merged duplicate items are free.

日本語での説明

指定した RSS・Atom フィードの新着だけを毎日まとめます。キーワードの一致は日本語の表記ゆれに強く、全角半角、カタカナ・ひらがな、長音（サーバー／ｻｰﾊﾞ）を同じものとして扱います。「AI|人工知能」のように同義語もまとめて指定できます。同じニュースが複数のサイトに出たときは1件にまとめ、ほかの出典を並べます。前回までに届けた記事は出しません。記事のページ本体は読まず、フィードに載っているタイトル・リンク・日付・要約の冒頭200文字だけを使います。

料金: 1回 $0.01 ＋ 新着1件 $0.0005（1,000件で $0.50）。絞り込みで外れた記事、既読の記事、まとめた重複は課金しません。