Wander Search & Extract Listings

Site wander.comTask search-and-extract-listings-d6fgrcVersion v1Updated Jul 13, 2026Category vacation-rentals

Discover Wander vacation-rental listings (via sitemap/locations) and extract a property's complete metadata plus the full image/video gallery from its detail page using server-rendered JSON-LD and the RSC medias arrays — read-only, no auth, no anti-bot. This skill was captured from a live agent session on wander.com and publishes here verbatim, exactly as an agent receives it.

NoteSelectors and URL schemes drift as sites change. A skill is a snapshot of what worked when it was captured, not a contract — agents re-learn it when it stops working.

Purpose

Discover vacation-rental listings across wander.com and extract the complete metadata and full image gallery from any listing's detail page. Returns structured property data — name, slug, geo-coordinates, address, bedroom/bathroom/occupancy counts, bed configuration, amenities, check-in/out times, nightly-rate range, 90-day daily pricing & availability, local recommendations — plus the entire media gallery (every room photo, hero shot, floor plan, and video, with all resolution variants). Read-only; never books, never submits payment. Wander explicitly welcomes AI agents (robots.txt sets Content-Signal: ai-train=yes, search=yes, ai-input=yes and ships an /llms.txt plus per-property markdown mirrors), so the optimal path is a plain browserless_agent goto of the SSR surfaces — no UI interaction, no authentication, and no anti-bot stealth required. Parse the JSON-LD and RSC medias in-page and return a compact projection.

When to Use

  • Building a catalog/index of every Wander property (3,397 property URLs are enumerable from the sitemap).
  • Extracting one listing's full detail record — all metadata + the complete photo/video gallery — for ingestion, comparison, or display.
  • Monitoring pricing and 90-day availability for specific properties.
  • Filtering/browsing listings by category (Beachfront, Mountain, Urban, Desert, Wine, Ski).
  • Anywhere you'd otherwise script a browser against Wander — the HTTP/markdown surfaces are faster, cheaper, and structurally stable.

Workflow

Wander's web UI is a Next.js (App Router) app, but it server-renders a complete schema.org VacationRental JSON-LD block and embeds the full property data (including the entire media gallery) inside the RSC flight payload of the detail-page HTML. It also publishes an LLM-friendly markdown mirror for every property. The recommended method is to fetch these SSR surfaces via browserless_agent goto (parsing in-page) — no UI interaction is needed for either search or full extraction. Honor the published rate limit of 1 request/second (/llms.txt).

Step 1 — Discover listings (pick a search surface)

  • Full enumeration (best for "all listings"): goto https://www.wander.com/sitemap.xml and read it with a text command (~6,011 <loc> entries; filter for /property/ → ~3,397 listing URLs). The slug is the last path segment, e.g. wander-june-lake.
  • Browse + category filter: goto https://www.wander.com/locations — server-rendered with ~3,281 /property/{slug} links and the six categories (Beachfront, Mountain, Urban, Desert, Wine, Ski). Also lists destinations/travel guides.
  • Query search (destination / dates / guests): the search form on / and /locations ("Where? / When? / Who?") drives the same /locations result set client-side; for programmatic use, enumerate from the sitemap and filter on the extracted metadata (geo, category, bedrooms, availability) rather than driving the form.

Dedupe slugs; each maps to https://www.wander.com/property/{slug}.

Step 2 — Load the detail page

{
  "method": "goto",
  "params": {
    "url": "https://www.wander.com/property/{slug}",
    "waitUntil": "load",
    "timeout": 45000
  }
}

A live property returns 200 with a ~700 KB HTML document (a real browser follows any 308 for legacy/renamed slugs automatically; see gotchas). Parse Steps 3–4 in-page with a single evaluate command so the ~700 KB HTML never crosses the wire — the browserless_function/evaluate text return caps at ~200 KB, so project down to the fields below rather than returning raw HTML.

Step 3 — Extract structured metadata from the JSON-LD block

Inside the Step 2 evaluate, parse the single <script type="application/ld+json">…</script> block (@type: VacationRental). It is the authoritative metadata source and is present on every live property page:

  • identifier → slug, name, url, description
  • latitude, longitude, address (streetAddress, addressLocality, addressRegion, postalCode, addressCountry)
  • checkinTime, checkoutTime
  • containsPlace.numberOfBedrooms, .numberOfBathroomsTotal, .occupancy.value (sleeps), .floorSize.value (sq ft), .numberOfRooms
  • containsPlace.bed[] → [{typeOfBed, numberOfBeds}]
  • containsPlace.amenityFeature[] → list of {name, value} (e.g. wifi:true, hotTub:true, petsAllowed:false, parkingType:"Free", poolType:"Outdoor")
  • image[] → 10 representative images only (a subset of the full gallery — do NOT treat this as the complete gallery)

Step 4 — Extract the FULL media gallery from the RSC payload

The complete gallery is embedded in the detail HTML's RSC flight data, split into one "medias":[ … ] array per room/section (Showcase, Exterior, Kitchen, each Bedroom, etc.). Do this union+dedup in-page inside the same evaluate (over document.documentElement.outerHTML) and return only the projected media list — never ship the raw ~700 KB payload back. To get every image/video:

  1. Unescape the flight data: replace the literal two-character sequence \" with " across the HTML.
  2. Find every occurrence of "medias":[, bracket-match each array to its closing ], and JSON.parse each.
  3. Union all arrays and dedupe by assetId — this is the complete gallery. (Nearby/"similar" property cards on the page do NOT contribute medias arrays, so the union is scoped to the current listing. Validate by confirming the JSON-LD image[] asset IDs are a subset of your union — they always are.)

Each media object contains:

  • assetId, id, type (ROOM, HERO, FLOOR_PLAN, HERO_ANIMATED_VIDEO, HOME_TOUR_VIDEO, HOME_SNEAK_PEEK_VIDEO), order, width, height
  • Resolution URLs (webp + png): fullResUrl (/{assetId}/fullres.webp), highResUrl (2440), mediumResUrl (1200), lowResUrl (640), lowestResUrl (320)

Sort by order; pick fullResUrl (or highResUrl) for the canonical image URL. Filter type to ROOM/HERO/FLOOR_PLAN for still photos, or keep the video types separately. Typical galleries hold ~60–150 unique media.

Step 5 — Enrich with the markdown mirror (pricing, availability, local recs)

The .md is a plain-text mirror. Two equivalent ways to read it:

  • Simplest — a goto to the .md URL followed by a text command:
    { "method": "goto", "params": { "url": "https://www.wander.com/property/{slug}.md", "waitUntil": "load", "timeout": 45000 } },
{ "method": "text", "params": { "selector": "body" } }
  • Or a browserless_function that navigates the origin first, then runs a same-origin fetch in-page (a bare fetch has no egress until the page is navigated — page.goto the origin, then page.evaluate):
    export default async function ({ page }) {
  await page.goto("https://www.wander.com/", { waitUntil: "load" });
  const md = await page.evaluate(async () =>
    (await fetch("/property/{slug}.md")).text(),
  );
  return { data: md, type: "text/plain" };
}

A compact (~14 KB) markdown document adding fields not in the JSON-LD:

  • Nightly rate range (USD, before taxes/fees), Next Available date, Available Days count
  • Daily Pricing (next 90 days) — one line per date: Jun 8, 2026: $822 (Available) / (Not Available)
  • Local recommendations — Dining, Activities & Attractions, Local Services (curated blurbs)
  • Additional Information notices (shipping, occupancy, hot-tub, pets, smoking) and Guest Services
  • Property ID (the internal CMS id, e.g. cm7j97um9067tl0x1ychp5t5u) and Hero Image URL

Step 6 — Merge and emit

Combine the JSON-LD metadata, the deduped media gallery, and the markdown pricing/availability/recs into one record (see Expected Output).

Browser fallback (only if the SSR parse ever breaks — not observed)

If the in-page parse ever fails, drive the UI in one browserless_agent call (a plain call is sufficient — no proxy/stealth arg needed):

  1. { "method": "goto", "params": { "url": "https://www.wander.com/property/{slug}", "waitUntil": "load", "timeout": 45000 } } then { "method": "waitForTimeout", "params": { "time": 2500 } }.
  2. { "method": "text", "params": { "selector": "head" } } to read the JSON-LD, then click the photo grid to open the gallery modal and { "method": "snapshot" } to harvest image refs from the urlMap (confirm the grid ref via snapshot if it misses). Note the modal repeats images across section tabs (the "X photos" counter, e.g. "115", double-counts; unique is ~59–64) — dedupe by asset ID. The HTML/RSC parse in Steps 3–4 is strictly better and avoids this.

Site-Specific Gotchas

  • No anti-bot, no auth, no proxy needed. Pre-run probe and direct testing both confirmed: a plain load (no residential proxy, no stealth) returns 200 with the full SSR payload for both /property/{slug} and .md. The successful runs set no proxy arg on the browserless_agent call. Wander actively invites agents (Content-Signal: ai-train=yes, search=yes, ai-input=yes).
  • Respect the published rate limit: 1 request/second (/llms.txt). Contact is ai@wander.com.
  • wander.com → www.wander.com (308). Always use the www. host to avoid an extra redirect hop.
  • JSON-LD image[] is only 10 representative images — NOT the full gallery. The complete gallery (often 60–150 media) lives only in the "medias" arrays of the RSC payload. Using image[] alone silently undercounts photos ~6–15×.
  • The gallery is split across many medias arrays (one per room/section), not a single array. Taking only the first/largest array undercounts (e.g. Wander June Lake's first array has 8, but the union of all 16 section arrays is 64). Always union + dedupe by assetId.
  • RSC flight data is backslash-escaped. Inside self.__next_f.push(...), quotes appear as the literal \". Unescape (\" → ") before regex/JSON parsing the medias objects.
  • The naïve "image has a 2440.webp variant" heuristic is unreliable for gallery membership: only the few hero images are preloaded at 2440 (Wander June Lake had 59 with a 2440 variant, but Wander Tulum had only 4). Use the medias arrays, not resolution-tier guessing.
  • The browser gallery's photo counter double-counts. The modal shows a "Showcase" tab plus per-room tabs that repeat the same images; the displayed total (e.g. "115") is larger than the ~59–64 unique assets. Dedupe by asset ID.
  • Some property slugs 308-redirect or are stale. The example slugs in /llms.txt are illustrative and can drift — wander-big-sur-cliffs currently 308-redirects. Trust the sitemap//locations for live slugs; follow redirects or skip on non-200.
  • property_id (markdown) ≠ assetId/identifier. The markdown "Property ID" is an internal CMS id (cm…); the JSON-LD identifier is the URL slug; assetId values are per-image. Keep them distinct.
  • .md daily pricing reflects rates even on unavailable dates. Each day line carries a price plus an (Available)/(Not Available) flag — a price does not imply bookability; read the flag.
  • Image asset URLs are unauthenticated and hotlinkable on assets.wander.com (/{assetId}/{fullres|2440|1200|640|320}.{webp|png}). No signing/expiry observed.
  • sitemap.xml and .md/HTML sometimes return empty over a flaky proxy hop; a plain browserless_agent call with no proxy arg is reliable and faster here since there's no anti-bot to evade.

Expected Output

A single merged record per listing. media[] is the full deduped gallery; images[] is the convenience list of still-photo URLs.

{
  "success": true,
  "search": {
    "surface": "sitemap.xml",
    "total_property_urls": 3397,
    "categories": ["Beachfront", "Mountain", "Urban", "Desert", "Wine", "Ski"],
    "sample_slugs": [
      "wander-june-lake",
      "wander-tulum-jungle-allure",
      "wander-gulf-shores"
    ]
  },
  "listing": {
    "slug": "wander-june-lake",
    "name": "Wander June Lake",
    "url": "https://www.wander.com/property/wander-june-lake",
    "property_id": "cm7j97um9067tl0x1ychp5t5u",
    "description": "Escape to Wander June Lake, where modern luxury meets nature's breathtaking beauty…",
    "latitude": 37.764228,
    "longitude": -119.103513,
    "address": {
      "streetAddress": "19 Willow Ave June Lake",
      "addressLocality": "June Lake",
      "addressRegion": "California",
      "postalCode": "93529",
      "addressCountry": "US"
    },
    "bedrooms": 3,
    "bathrooms": 2.5,
    "sleeps": 6,
    "square_feet": 1519,
    "beds": [
      { "type": "King", "count": 2 },
      { "type": "Twin", "count": 4 }
    ],
    "amenities": {
      "wifi": true,
      "hotTub": true,
      "fireplace": true,
      "ac": true,
      "balcony": true,
      "washerDryer": true,
      "outdoorGrill": true,
      "instantBookable": true,
      "selfCheckinCheckout": true,
      "smokingAllowed": false,
      "petsAllowed": false,
      "parkingType": "Free",
      "internetType": "Free"
    },
    "checkin_time": "16:00:00",
    "checkout_time": "10:00:00",
    "nightly_rate_range_usd": [822.15, 1583.4],
    "next_available": "2026-06-09",
    "available_days_next_90": 58,
    "daily_pricing": [
      { "date": "2026-06-08", "price_usd": 822, "available": true },
      { "date": "2026-06-14", "price_usd": 822, "available": false }
    ],
    "local_recommendations": {
      "dining": ["Silver Lake Resort Cafe", "Balanced Rock Grill & Cantina"],
      "activities": ["Oh Ridge Campground", "Rush Creek Trailhead"],
      "services": ["Silver Lake Resort"]
    },
    "media_count": 64,
    "media_types": {
      "ROOM": 56,
      "HERO": 3,
      "FLOOR_PLAN": 2,
      "HOME_TOUR_VIDEO": 1,
      "HERO_ANIMATED_VIDEO": 1,
      "HOME_SNEAK_PEEK_VIDEO": 1
    },
    "media": [
      {
        "assetId": "576411499854040933",
        "type": "ROOM",
        "order": 0,
        "width": 3264,
        "height": 2448,
        "fullResUrl": "https://assets.wander.com/576411499854040933/fullres.webp",
        "highResUrl": "https://assets.wander.com/576411499854040933/2440.webp"
      }
    ],
    "images": [
      "https://assets.wander.com/576411499854040933/fullres.webp",
      "https://assets.wander.com/576411630200424914/fullres.webp"
    ],
    "image_count": 61
  },
  "error_reasoning": null
}

Failure shape (e.g. stale/redirected slug):

{
  "success": false,
  "slug": "wander-big-sur-cliffs",
  "http_status": 308,
  "error_reasoning": "Slug 308-redirected (renamed/removed); resolve a live slug from sitemap.xml or /locations."
}

Call it

GET https://production-sfo.browserless.io/skills?token=TOKEN-HERE&domain=wander.com&task=search-and-extract-listings-d6fgrc