Any Website URL to Article Summarizer

Extract and summarize article content from any website URL. Works on news sites, blogs, Wikipedia, Medium, documentation, and more.

Uses extractive summarization (selects the most important sentences from the article itself) — no external AI API key required.

Features

• 📰 Extract articles from any URL — news sites, blogs, Wikipedia, Medium, etc.
• 📝 Automatic extractive summary generation (selects most informative sentences)
• 🔑 Key points — top 5 most important sentences from the article
• 👤 Extracts title, author, publish date, description, and hero image
• 📊 Word count and reading time estimation
• 🔒 No AI API key required — all summarization happens locally
• 🌐 Supports multiple URLs per run
• 💾 Optional: include full cleaned article text in output

How Summarization Works

This actor uses extractive summarization:

1. Article text is cleaned and split into sentences
2. Each sentence is scored by word frequency (TF-style scoring)
3. The top-scoring sentences are selected and returned in their original reading order
4. Position bias boosts early sentences (intros are typically more important)

This approach works across all languages and domains without requiring an LLM or external API.

Input

Field	Type	Default	Description
urls	array	required	List of article URLs to summarize
summaryLength	string	"medium"	short (3 sentences), medium (5), long (8)
includeFullText	boolean	false	Include full cleaned article text in output
maxItems	integer	10	Maximum number of articles to process

Example Input

{
  "urls": [
    "https://en.wikipedia.org/wiki/Artificial_intelligence",
    "https://techcrunch.com/2024/01/01/sample-article/"
  ],
  "summaryLength": "medium",
  "includeFullText": false,
  "maxItems": 10
}

Output

Each dataset record represents one summarized article:

Field	Type	Description
position	integer	Position in results
url	string	Article URL
domain	string	Website domain (e.g. "techcrunch.com")
title	string	Article title
author	string|null	Article author name
publishDate	string|null	Publish date (ISO format or raw string)
description	string|null	Meta description or excerpt
summary	string	Extractive summary of the article
keyPoints	array	Top 5 key sentences from the article
wordCount	integer	Total word count
readingTime	string	Estimated reading time (e.g. "5 min read")
image	string|null	Hero image URL (og:image)
siteName	string|null	Website name (og:site_name)
language	string|null	Document language code
fullText	string	Full cleaned article text (if includeFullText=true)
scrapedAt	string	ISO 8601 scrape timestamp

Example output

{
  "position": 1,
  "url": "https://en.wikipedia.org/wiki/Machine_learning",
  "domain": "en.wikipedia.org",
  "title": "Machine learning - Wikipedia",
  "author": null,
  "publishDate": null,
  "description": "Machine learning (ML) is a field of study...",
  "summary": "Machine learning is a subset of artificial intelligence...",
  "keyPoints": ["Machine learning models are often vulnerable to...", "..."],
  "wordCount": 9653,
  "readingTime": "48 min read",
  "image": null,
  "siteName": null,
  "language": "en",
  "scrapedAt": "2025-08-01T12:00:00.000Z"
}

Dataset Views

• Articles Overview — table with title, author, date, word count, reading time, URL, and summary
• Summaries — focused view showing title, summary, key points, URL, and domain

Technical Notes

• Uses real Google Chrome browser (Playwright) for handling JavaScript-rendered pages
• Fingerprint injection for natural browser behavior
• Article content is extracted using a multi-selector heuristic that prioritizes <article>, [itemprop="articleBody"], and common blog/CMS CSS classes
• Wikipedia [edit] and footnote [1] markers are automatically removed
• Reference sections (.reflist, .references) are removed from Wikipedia pages
• For paywalled articles, only publicly visible content is extracted

License

Apache-2.0