Alfagift (Alfamart) Product, Price & Promo Scraper

Twelve Alfagift scrapers in one. No login, no browser, no captcha, no proxy required.

Alfagift is the online store of Alfamart, Indonesia's largest minimarket chain (~20,000 stores). This actor reads its public catalogue through the same internal API the website itself uses: roughly 10,000 products with live prices, discounts, stock flags and full category paths, plus the brand directory and every promotional campaign currently running.

A full-catalogue run takes 49 HTTP requests and about three minutes for 10,081 records.

---

What you can pull

Mode	What it gives you	Needs input?
search	Products for a keyword, plus Alfagift's own SEO copy and spelling suggestions	keywords
searchSuggestions	Autocomplete suggestions for a seed keyword	keywords
categoryTree	The whole navigable category tree (61 nodes)	—
categoryProducts	Every product in one or more categories	category ids or URLs
fullCatalog	The entire catalogue, ~10,000 products in ten requests	—
productDetail	Full product page: barcode, brand, image gallery, HTML description, discount end date	product ids or URLs
brandDirectory	The A–Z brand directory (~2,400 brands)	—
brandProducts	Every product carried under a brand	brand ids or URLs
newArrivals	The newest-products feed (198 items)	—
recommendations	The homepage recommendation rail (48 items)	—
promoBanners	Live promotional campaigns with artwork and run dates	—
promoProducts	The products inside each live campaign, at promo prices	—

Every mode accepts pasted alfagift.id URLs as well as raw ids, so you can drive a run straight from links in a browser.

Quick start

Cheapest useful run — what Alfamart is promoting right now and at what price:

{ "mode": "promoProducts" }

Price-track a keyword:

{ "mode": "search", "searchKeywords": ["indomie", "susu uht", "kopi kapal api"] }

The whole catalogue:

{ "mode": "fullCatalog" }

Enrich specific products with barcodes and brands:

{
  "mode": "productDetail",
  "productIds": ["13585", "https://alfagift.id/p/indomie-mi-instan-goreng-84-g-754971"]
}

Output

Every record shares one envelope, so a single dataset can hold products, categories, brands and campaigns and still be joined on id:

{
  "item_type": "product",
  "id": "845475",
  "data": {
    "productId": "845475",
    "productName": "Soulyu Stain Lip Tint 03 Courage 4 ml",
    "sku": "A8454750002167",
    "plu": "461719",
    "basePrice": 59000.0,
    "finalPrice": 29900.0,
    "discountPercent": 49.32203389830509,
    "stock": 0,
    "validFlag": false,
    "categoryNameLvl0": "Personal Care",
    "categoryNameLvl1": "Riasan Wajah & Tubuh",
    "categoryNameLvl2": "Liquid Lip",
    "productUrl": "https://alfagift.id/p/soulyu-stain-lip-tint-03-courage-4-ml-845475",
    "isBuyable": false,
    "isDiscounted": true,
    "discountAmount": 29100.0
  },
  "metadata": {
    "scrapedAt": "2026-09-21T14:05:13+00:00",
    "sourceUrl": "https://alfagift.id/promotion-products?id=6aacea5b144aa661216a3007",
    "mode": "promoProducts",
    "rank": 2,
    "parentId": "6aacea5b144aa661216a3007"
  }
}

Alfagift's own fields are passed through verbatim. Four are added: productUrl, isBuyable, isDiscounted and discountAmount.

Read isBuyable before you average anything

Alfagift leaves delisted and out-of-area products in every listing, with validFlag: false and every price field set to 0.0. An average price computed over a raw category is dragged toward zero by rows that are not for sale at all. isBuyable makes that one filter instead of a convention you have to know. The row above is a real example: a live 49%-off promo item that is nonetheless flagged unbuyable and out of stock.

Things worth knowing

Prices are national. Alfagift publishes one price list (seller SAT) to anonymous callers. The API accepts latitude/longitude headers but ignores them — Jakarta, Medan and Jayapura return identical prices. Store-local pricing needs a logged-in member with a chosen store, which this actor does not do.

Some brand letters are broken upstream. Letters C, F, G, V and W return HTTP 500 from Alfagift's own brand index on every attempt, from every address. 21 of 26 letters work (~2,400 brands). The run reports those letters as diagnostic rows rather than silently missing brands — it is Alfagift's index, not the scraper.

Some category pages are unreadable, and the actor works around it. A few listings contain a record Alfagift's gateway cannot serialise, so any page spanning it returns HTTP 500 forever. At the time of writing that was the whole 1,008-product Lifestyle category, which returns nothing to a normal request. This actor re-reads a failed page in smaller slices and recovered 951 of the 1,008 products, with the summary row stating how many it could not reach:

{ "_warning": "poisoned_pages_skipped", "unreadablePages": 1, "approxRowsLost": 10 }

Roughly half the catalogue is not actually for sale. The 2026-09-21 full run returned 10,071 products, of which 4,926 were buyable and 4,749 carried a discount. That ratio is why isBuyable exists.

fullCatalog returned no duplicates. Each product resolves to a single top-level category, so the ten-category walk did not overlap once across 10,071 rows. De-duplicating on data.productId is still a cheap safeguard if Alfagift ever starts cross-listing.

Product ids are numeric — a SKU will not work. /v2/products/A13170001210 returns HTTP 500. Use the numeric productId from a listing row, or paste the /p/ URL and let the actor extract it. The run says so explicitly rather than failing quietly.

Feed ceilings are the server's, not the actor's. newArrivals stops at 198 rows and recommendations at 48, whatever limit you ask for.

What this actor cannot reach. Store locations, stock by store, carts, orders and member data all require a real member login and stay closed. This is a public-data actor by design.

How it gets in

Alfagift has no WAF — no Cloudflare, no DataDome, no TLS fingerprint gate, no IP or geo restriction. Plain HTTP/1.1 from a datacentre address works.

What it has is a custom fingerprint request header that the site's own JavaScript builds by AES-encrypting a FingerprintJS visitor id. Without it every endpoint returns 401 Unauthorized — including the login endpoint itself, which is the giveaway that it is a client check rather than authentication. The actor reproduces the encryption in Python, so no browser is needed.

The full reverse-engineering write-up, the endpoint map, and every decoy and trap measured are in CRAWLING_METHOD.md.

Input reference

Field	Default	Notes
mode	search	See the table above
searchKeywords	—	Keywords or /find/ URLs
categoryIds	—	24-char ids or /c/ URLs
productIds	—	Numeric ids or /p/ URLs
brandIds	—	Numeric ids or /b/ URLs
promotionIds	—	Optional; empty discovers every live campaign
brandLetters	A–Z	Restrict the brand sweep
targets	—	Generic list, routed to whatever the mode needs
maxItems	0 (unlimited)	Hard cap on dataset rows
pageSize	1000	Alfagift enforces no cap; most targets fit in one request
maxPagesPerTarget	50	Safety ceiling
bannerLocation	main_banner_new	The only location the live site uses
includePromoDetail	false	One extra request per banner
requestDelaySecs	0.2	Raise if a long walk ever sees HTTP 429
proxyConfiguration	off	Optional — only buys rate-limit headroom

Local development

cd alfagift-scraper
../.venv/Scripts/python.exe test_local.py          # all 63 checks, live
../.venv/Scripts/python.exe test_local.py search   # one mode

echo '{"mode":"categoryTree"}' > storage/key_value_stores/default/INPUT.json
../.venv/Scripts/python.exe -m src

test_local.py covers the AES gate against a fixed vector, URL/id resolution, the derived fields, the dataset schema (including the id: null diagnostic row that Apify validates batch-wide), the page-index paging regression, and every mode against the live site.

Legal

Reads only data Alfagift serves publicly to any visitor. No login, no account, no personal data, no checkout or cart interaction. robots.txt is User-agent: * / Allow: /.