RealtySync Universal — AI Real Estate Extractor

Extract property listings from any Czech or Slovak real estate website — boutique agencies, property managers, developer sites, and non-standard portals that no pre-built scraper covers.

What does this Actor do?

This Actor uses Claude AI (LLM) to understand and parse listing pages from arbitrary real estate websites:

1. Loads the page with Playwright (handles JavaScript-rendered sites and cookie banners)
2. Extracts structured data via multiple layers: JSON-LD, dataLayer, page text (trafilatura), image URLs, and GPS from embedded maps
3. Sends to Claude AI with a forced JSON schema — the LLM identifies and normalizes property details
4. Outputs standardized listings matching the RealtySync schema (price, area, GPS, contact, images)

If the page already has JSON-LD with real estate data, the LLM step is skipped entirely (fast path).

Why use the Universal extractor?

The top 3-4 portals (Sreality, Bezrealitky, iDnes) cover ~80% of the Czech market. The remaining ~20% is spread across hundreds of boutique agency websites, developer landing pages, and regional portals. These sites have unique HTML structures that break traditional scrapers. The Universal extractor handles them all — point it at any listing page and get normalized output.

Target audience:

• Property databases that need the "long tail" of exclusive agency listings
• Market analysts covering niche segments (luxury, new builds, commercial)
• Cross-portal deduplication pipelines that need full market coverage

How to use the Universal extractor

1. Create a free Apify account at apify.com if you don't have one.
2. Get an Anthropic API key at console.anthropic.com — the LLM extraction costs ~$0.001-0.003 per page (billed by Anthropic separately from this Actor's PPE charge).
3. Open this Actor in the Apify Console and click "Start".
4. Add listing page URLs and paste your Anthropic API key.
5. Click Run and download standardized listings from the Output tab.

Input example

{
    "startUrls": [
        { "url": "https://www.bythos.cz/byty/prodej" },
        { "url": "https://www.luxury-reality.cz/nabidka" }
    ],
    "anthropicApiKey": "sk-ant-...",
    "maxResults": 50,
    "fetchDetailPages": true
}

Output example

{
    "listing_id": "universal-a1b2c3d4",
    "title": "Prodej bytu 4+kk 120 m², Praha 1",
    "property_type": "APARTMENT",
    "transaction_type": "SALE",
    "price_czk": 18500000,
    "currency": "CZK",
    "locality": "Praha 1 - Josefov",
    "area_m2": 120.0,
    "latitude": 50.0900,
    "longitude": 14.4200,
    "seller_name": "Luxury Reality s.r.o.",
    "seller_phone": "+420 222 333 444",
    "url": "https://www.luxury-reality.cz/nabidka/byt-4kk-praha-1",
    "images": ["https://..."],
    "source": "universal"
}

Input parameters

Parameter	Type	Required	Default	Description
startUrls	array	Yes	—	URLs of listing/index pages to scrape
anthropicApiKey	string	Yes	—	Your Anthropic API key for Claude AI extraction
maxResults	integer	No	100	Maximum total listings to extract
maxPagesPerSite	integer	No	5	Maximum pagination pages per start URL
fetchDetailPages	boolean	No	true	Follow links to detail pages for richer data (GPS, full specs)

How much will it cost?

This Actor uses Pay-Per-Event pricing at $0.01 per listing (higher than portal-specific scrapers due to LLM processing).

Volume	Cost (Actor PPE)	Est. Anthropic LLM cost	Total
10 listings	$0.10	~$0.01-0.03	~$0.13
50 listings	$0.50	~$0.05-0.15	~$0.65
100 listings	$1.00	~$0.10-0.30	~$1.30

Note: Anthropic API costs are billed separately by Anthropic to your account. Pages with JSON-LD data skip the LLM call entirely, reducing Anthropic costs.

Platform costs (compute) are billed separately by Apify based on your plan.

Tips

• Use the dedicated portal scrapers (Sreality, Bezrealitky, iDnes) for those sites — they're 3x cheaper and faster.
• Set fetchDetailPages: true for best data quality — detail pages usually have GPS coordinates and full specs.
• Start with maxResults: 10 to verify the Actor handles your target site before scaling up.
• Pages with structured data (JSON-LD) are processed instantly without LLM calls — check your site first.

Is it legal to scrape these websites?

This Actor extracts publicly available data from public web pages. Web scraping of publicly accessible information is generally permitted under applicable law. Always review the target site's Terms of Service and applicable regulations (including GDPR for EU data) before use. This Actor does not bypass authentication, paywalls, or CAPTCHAs. Use responsibly and respect rate limits.