zgarcit.ro logo
zgarcit.ro

list-all-products

Installation

Adds this website's skill for your agents

 

Summary

Enumerate the full zgârcit.ro catalog of Romanian grocery discounts across 9 retail chains by paging the ?page=N listing and parsing the inline Next.js RSC payload (no public API).

FIG. 01
FIG. 02
FIG. 03
SKILL.md
162 lines

List All Products on zgârcit.ro

Purpose

Enumerate the complete catalog of current grocery discounts ("produse") aggregated by zgârcit.ro, a Romanian price-comparison site that tracks weekly promotions across nine retail chains (Kaufland, Lidl, Auchan, Mega Image, Penny, Carrefour, Profi, Selgros, Metro). This skill is read-only and returns a structured list of every deal — product title, store, category, discounted price, original price, loyalty-card requirement, and image — paginated 60 items per page across the full result set (~18,679 products / 312 pages at time of writing).

When to Use

  • "Get me every product / deal currently listed on zgârcit.ro."
  • "List all discounts from Lidl (or Kaufland, Penny, …) on zgârcit.ro."
  • "Export the full zgârcit.ro catalog as JSON."
  • "Find all products matching lapte (milk) and their prices."
  • Building a dataset of Romanian grocery promotions, price tracking, or cross-store comparison.

Workflow

Recommended method: HTTP fetch + RSC parse (no browser needed).

zgârcit.ro is a Next.js App Router site. There is no public JSON API (/api/ is Disallow-ed in robots.txt and returns nothing useful), but every listing page server-renders its product data inline as a React Server Components (RSC) flight payload inside <script>self.__next_f.push(...)</script> tags. You can fetch the plain HTML for any page and parse the products out deterministically — far cheaper and more reliable than driving a browser. Cloudflare fronts the origin, so route the fetch through residential proxies.

  1. Discover the page count. Fetch the homepage and read totalPages from the RSC payload:

    browse cloud fetch "https://zgarcit.ro/" --proxies

    Grep the response body for "totalPages":<N> (escaped as totalPages\":312 in the raw stream) and "currentPage":1. Each page holds 60 products (the last page holds the remainder).

  2. Iterate every page via the ?page=N query param (1-indexed, 1 … totalPages):

    for p in $(seq 1 312); do
  browse cloud fetch "https://zgarcit.ro/?page=$p" --proxies --output "page-$p.html"
done

  3. Parse products from each page's RSC payload. Two component prop blocks per product carry all reliable fields — match both by id and join:

    • Meta block: {"id":"…","category":"…","provider":"…","requiresVendorCard":false,"local":false}
    • Price block: {"id":"…","title":"…","newPrice":0.99,"oldPrice":1.49,"provider":"…","priceTiers":null}
    • Image: the adjacent {"src":"https://bucket.zgarcit.ro/<provider>/<hash>","alt":"<title>"}

    The raw bytes are JSON-string-escaped (every " appears as \"), so regex against the escaped form, e.g. \\"newPrice\\":([0-9.]+). Example extractor logic (Node): collect a {id → {category, requiresVendorCard, local}} map from meta blocks, then walk the price blocks in document order (they map 1:1 and in-order to the image src list) and merge.

  4. Aggregate all pages' products into one list. Expect ≈ (totalPages-1) × 60 + lastPageCount items; the UI header ("18679 produse") gives the authoritative count to validate against.

Optional filters (all are real URL deep-links, combinable with &page=N):

  • Single store: ?providers=Lidl (exact display name, URL-encode spaces, e.g. ?providers=Mega%20Image).
  • Multiple stores: repeat the param — ?providers=Lidl&providers=Kaufland. CSV does not work (?providers=Lidl,Kaufland returns 0 products).
  • Text search: ?search=lapte (NOT ?q= — q is silently ignored and returns the full catalog).
  • Each filtered view recomputes its own totalPages, so re-read it after applying a filter.

Browser fallback

Only needed if the RSC payload format changes. Open https://zgarcit.ro/, dismiss the "Cookie-uri?" dialog (click Accept), and read the product grid. Pagination controls are top-right (1 2 … 312); the left sidebar has store checkboxes ("Magazine") and sort controls ("Sortare"). Navigate pages by editing ?page=N in the URL rather than clicking, since the grid is virtualized. Extract each card's title, the two stacked prices (discounted in orange, struck-through original), the RON/kg unit price, and the validity date (e.g. "9 iunie"). Do not use the in-page sort buttons as a deep-link — sorting is client-side state only and never appears in the URL.

Site-Specific Gotchas

  • No JSON API. /api/ is Disallow-ed in robots.txt and is not the data source. All product data is in the server-rendered RSC flight payload (self.__next_f). Don't waste time hunting for a REST/GraphQL endpoint — there isn't one.
  • Cloudflare in front. Origin is a Railway-hosted Next.js app behind Cloudflare. Plain fetches work but use --proxies (residential) to avoid edge challenges. Responses are zstd/gzip encoded; browse cloud fetch decodes them for you.
  • Raw bytes are double-escaped. Inside the HTML, the RSC JSON has every " written as \". Regex/parse against the escaped form (\\"title\\":\\"…\\"), not clean JSON.
  • 60 products per page, fixed. Pagination is strictly ?page=N, 1-indexed. There is no page-size parameter — you must walk all pages.
  • ?search= not ?q=. The search param is search. q= is silently ignored and returns the unfiltered full catalog (a 312-page false positive — easy to mistake for "no filter applied").
  • Multi-store is repeated params, not CSV. ?providers=A&providers=B works; ?providers=A,B returns 0 results.
  • Store names are exact display strings. Use Mega Image (URL-encoded Mega%20Image), Kaufland, Lidl, Auchan, Penny, Carrefour, Profi, Selgros, Metro. The canonical list is in /sitemap.xml.
  • Sorting is not URL-addressable. Clicking "Preț crescător" / "Discount descrescător" etc. changes only client-side state; the URL stays /. If you need sorted output, sort the extracted list yourself.
  • oldPrice can be null and priceTiers is usually null but can be an array for quantity/loyalty-tier pricing (e.g. some Lidl Plus deals). Handle both.
  • requiresVendorCard: true marks deals that need the chain's loyalty card (e.g. ~23 of the 61 Lidl deals require Lidl Plus). local: true marks region-specific deals.
  • Unit price (RON/kg) and validity date are rendered as display text in the card DOM, not as clean fields in the price-block props — derive them from the rendered HTML if needed, or compute unit price from newPrice + the quantity in the title.
  • Catalog is volatile — deals refresh weekly (sitemap changefreq: daily), so totalPages and the exact product set drift day to day. Always re-read totalPages at run time rather than hardcoding 312.
  • A separate /peco page lists fuel (PECO) prices — out of scope for "products"; the homepage grid is the grocery catalog.

Expected Output

A JSON object with the catalog total and a flat product array. Each product:

{
  "totalProducts": 18679,
  "totalPages": 312,
  "pageSize": 60,
  "scrapedAt": "2026-06-03T15:30:00Z",
  "products": [
    {
      "id": "n5ibm29azfh31d3",
      "title": "Demibaghetă",
      "provider": "Kaufland",
      "category": "Brutărie & patiserie",
      "newPrice": 0.99,
      "oldPrice": 1.49,
      "currency": "RON",
      "requiresVendorCard": false,
      "local": false,
      "priceTiers": null,
      "image": "https://bucket.zgarcit.ro/kaufland/5c8b832810c5b2dae02d5dea5b00d022e2d59ff3e9f04b6d7787403383107576"
    },
    {
      "id": "3d6th7vnqg5yi2q",
      "title": "Covridog",
      "provider": "Kaufland",
      "category": "Brutărie & patiserie",
      "newPrice": 3.19,
      "oldPrice": 3.89,
      "currency": "RON",
      "requiresVendorCard": false,
      "local": true,
      "priceTiers": null,
      "image": "https://bucket.zgarcit.ro/kaufland/3dc0eb8471d45dde811cfac78535a9903d1d78c7d44e9b17efc16048266c84e5"
    }
  ]
}

Variant — a deal requiring a loyalty card with no recorded original price:

{
  "id": "abc123examplex",
  "title": "ORLANDO Salam pentru câini",
  "provider": "Lidl",
  "category": "Hrană animale",
  "newPrice": 5.99,
  "oldPrice": null,
  "currency": "RON",
  "requiresVendorCard": true,
  "local": false,
  "priceTiers": null,
  "image": "https://bucket.zgarcit.ro/lidl/<hash>"
}

When scoped with a filter (e.g. ?providers=Lidl), the same shape is returned but totalProducts/totalPages reflect the filtered subset (e.g. 61 products / 2 pages for Lidl alone).

List All Products on zgârcit.ro · browse.sh