2026-02-15 — v0.3: Data Cleanup & Deduplication Deduplicate and flatten the raw enrichment output. Price appeared 12+ times, IDs in 5+ objects, promos 4x, images mixed product/UGC/avatars/videos. Now a clean, consumer-friendly structure.
Output : ~1,850 lines/product → ~280 lines (85% reduction), zero data lossPrice : 12+ locations → 4 flat fields (pricepriceYaBankpriceWithoutVatcurrencyRating : 3 sources → 4 fields (ratingratingCountreviewCountratingDistributionSeller : scattered → 6 flat fields (sellerNamesellerLogosellerRatingImages : 60+ mixed items → 3 typed arrays (imagesugcImagesvideosDelivery : 4 overlapping structures → deliverydeliveryAlternativesPromos : 4 copies → 1 array (richest version with descriptions, all entries kept)IDs : scattered across 5+ objects → single top-level setSpecs : dropped redundant "Артикул Маркета" and "Бренд" entries
Removed (dropped blobs — all data preserved in flat fields)
offerSnippetproductPayloadbuyOptioncomposeshopInfoseller*deliveryOptionsdeliveryConsolejsonldOfferjsonldDescriptionjsonldImagejsonldUrlaggregateRatingSearch tracking fields: showUidppmetaisReferralhasVendorUrlsignalssnippetTypetype
clean_product()New top-level fields: modelNamecategoryNameproductSlugoskuSlugarticleNumberbusinessIdfeedIdnavnodeIddepartmentIdpriceYaBankpriceWithoutVatstockCountmaxOrderQuantityplacementTypepaymentTypepaymentMethodsisBnplofferFlagsbenefitBadgefeedOfferIdwarehouseIdsearchPosition
isCrossBorder
2026-02-14 — v0.2: Render-Lazy Enrichment Migration Migrated product enrichment from full SSR page fetch (2.4MB/product) to render-lazy widget API (6 POSTs, ~95KB/product). 96% bandwidth reduction with richer structured data.
Enrichment : 6 render-lazy widget POSTs replace 1 full-page GET per product
ProductPageMetaProductDescriptionDefaultOfferUnifiedShopInfoBlockBreadcrumbsDeliveryConsole
Retries : 3 attempts (down from 5) — wraps entire 6-widget batchEnrichability : Only needs modelIdmarketSkuSession : Exposes csp_nonce
New fields : sellerRatingsellerLogosellerRatingCountstockcomposebuyOptionincludeReviewstests/test_review_widgets.py
aggregateRatingnumGradesnumReviewsnumOffersincludeReviewsSlug extraction dependency — render-lazy uses dummy slug x
2026-02-14 — v0.1: Actor Implementation First working Apify actor. All scraping logic ported from 9 test files into production structure.
Search : Resolve API with resolvePoorRemoteSearchExtraction : 35+ fields per product from zone-data (title, price, SKU, modelId, shopId, vendorId, delivery, etc.)Filters : sortBy, priceFrom/priceTo, categoryId (hid) — passed as resolve API paramsRegion : Cookie-based region override (lr + yandex_gid), defaults to Moscow (213)Enrichment (on by default): Fetches /product--{slug}/{modelId}?sku={skuId}
Brand, description, 10-15 specs, seller name, aggregateRating, 6-level breadcrumbs
37-43 original-quality images, up to 5 reviews with author/date/pros/cons/comment/grade
numGrades vs numReviews distinction, numOffers (seller count)
Specs extraction : CSS-class-free — uses specLinktext<span>Session : curl_cffi with Chrome TLS fingerprint, automatic sk/version extraction from window.stateError handling : CAPTCHA detection, session expiry auto-recovery, user-friendly error messagesProxy : Input proxyUrl > PROXY_URL env > Apify proxy configurationPPE : Single product_scraped
55 fields per enriched product (25 from search + 11 from enrichment + 19 bonus zone-data fields).
