Article Extractor — Clean Text for LLM & RAG Pipelines

URL in → clean article JSON out. Title, author, publish date, full plain text, top image, language, word count, reading time. No browser, no boilerplate, no nav-menu garbage in your embeddings.

Why this Actor

If you're feeding web articles to an LLM — RAG ingestion, summarization pipelines, news monitoring, research agents — raw HTML is 90% noise. This Actor extracts just the article using a three-layer strategy:

1. JSON-LD (Article / NewsArticle / BlogPosting structured data) — most reliable when present
2. Open Graph / meta tags — title, author, dates, hero image
3. Readability heuristics — semantic containers (<article>, <main>, common content classes) with paragraph-level scoring, nav/footer/aside stripped

Plain-fetch only: fast, cheap, and scales to 1,000 URLs per run.

Input

{ "url": "https://example.com/blog/some-article" }

Batch mode:

{ "urls": ["https://a.com/post-1", "https://b.com/story-2"], "maxChars": 50000 }

Output

{
  "url": "https://example.com/blog/some-article",
  "title": "How We Scaled to 1M Users",
  "author": "Jane Smith",
  "publishedAt": "2026-06-12T09:00:00Z",
  "siteName": "Example Engineering",
  "language": "en",
  "description": "Lessons from scaling...",
  "topImage": "https://example.com/hero.jpg",
  "text": "Full clean plain text of the article...\n\nParagraphs preserved...",
  "wordCount": 1840,
  "readingTimeMinutes": 8,
  "extractionSignals": { "hadJsonLd": true, "hadOgTitle": true }
}

Failed URLs return { url, error } rows — batch runs never fail silently.

Use cases

• RAG / vector-DB ingestion — clean text straight to your embedder
• News + brand monitoring — schedule against a URL feed, pipe to Slack/webhook
• LLM research agents — give your agent a "read this page properly" tool via Apify MCP
• Content archiving — normalized JSON of everything your team publishes or tracks

Limitations

Fetch-only: JavaScript-rendered articles (rare for news/blogs) and hard paywalls are out of scope. For those, pair with a browser-based Actor.

Pricing

Pay-per-result: $0.002 per article ($2 per 1,000). A 100-article batch typically completes in under a minute.

Author

Built by Peak Post — 14 more data + audit Actors on the profile.