Smart Article Extractor

Give me static article URLs and I return structured article text, metadata, and a deterministic content hash.

Extracts title, author, published date, text, word count, language, canonical URL, source URL, and whether text was truncated. Author and date may be empty when page metadata is missing. Extraction uses HTML paragraphs and metadata; it does not use AI or a browser.

Input

{"articleUrls": ["https://blog.python.org/2024/10/python-3130-final-released/"], "maxChars": 20000}

Accepts 1–50 URLs per run. Exact duplicates after trimming are processed once. maxChars is 1,000–100,000, default 20,000. Invalid URLs become individual error records; malformed input objects fail validation.

Output

Useful records have status: "ok", article fields, contentHash, input, url, finalUrl, sourceUrl, actorVersion, and contractId. Failed URLs have status: "error" and an error message. The hash describes the returned text, including truncation; it is a 16-character SHA-256 prefix for routine deduplication, not a security identifier.

Summary and failures

Read the SUMMARY key-value record for submitted and deduplicated counts, processed/succeeded/failed counts, actually charged units, status, settings, and unprocessed URLs. The legacy run-summary key contains the same record.

A mixed batch preserves useful articles and reports partial_success. A batch with no usable article fails after saving its error records and summary. Storage failures prevent charging. Billing failures stop further work, preserve existing results, and fail the run. Spending limits stop further extraction and identify remaining URLs.

Pricing

$0.001 per useful article saved ($1 per 1,000). A usable title and nonempty text are required. Failed or empty extractions are not charged. Platform usage is included in the public pay-per-event price. Owner development runs incur platform costs separately. This does not promise settled creator revenue or profit.

Limitations

Static, publicly accessible article HTML only. No crawling, JavaScript rendering, HTML meta-refresh redirect handling, login, paywall bypass, or proxy fallback. Use the final article URL for HTML redirect pages. Not intended for article index pages. Paragraph extraction is heuristic and may miss lists, tables, captions, or unusual layouts. Metadata accuracy depends on the source. Respect the publisher's access and reuse conditions.

API and agent use

Start a run using the Apify API Actor ID 5SmjtjmTos1ONDyV7 with the input above. Read article records from its default dataset and SUMMARY from its default key-value store. The Actor output exposes both result links. Keep API credentials in a secret store.