Catalog & content

Catalog sources

AskMerra only recommends what is in your catalog, so a complete, fresh catalog is the most important input.

Four ways to import

Method	Best for	Freshness
File upload (CSV, JSON, XML)	Getting started, occasional updates	When you upload
Feed URL	Shops that already publish a product feed (e.g. for Google Shopping)	Automatic, every 1–6 h depending on plan
Push API	Custom platforms, ERPs, real-time stock	Real time
Platform apps	Magento, Shopify	Coming soon

Formats

CSV

Comma, semicolon, tab or pipe separated (detected automatically). UTF-8 recommended; files that are not valid UTF-8 are read as Windows-1252. First row = column names.

products.csv

id;name;price;sale_price;currency;in_stock;category;brand;url;image
LUM-SE-01;Hydra Glow Serum 30 ml;39;32;EUR;yes;Face > Serums;Lumière;https://shop.ro/p/hydra-glow;https://shop.ro/img/hydra.jpg

JSON

An array of objects, or an object with a products (or items) array. Keys follow the same aliases as CSV columns.

{ "products": [
  { "id": "LUM-SE-01", "name": "Hydra Glow Serum 30 ml", "price": 39, "sale_price": 32,
    "in_stock": true, "category": "Face > Serums", "brand": "Lumière",
    "skin_type": "dry, sensitive" }
] }

XML (Google Merchant Center)

RSS 2.0 (rss > channel > item) or Atom (feed > entry) feeds with g: fields. The same feed you send to Google Shopping works as is.

Field mapping

After you upload a file or add a feed, AskMerra suggests a mapping from your columns to its fields. Review it before importing; columns you map to attribute:<name> (e.g. attribute:skin_type) become searchable product attributes; common descriptive columns (color, size, material, ingredients…) are suggested as attributes automatically. Recognised aliases:

Field	Recognised column names (examples)	Notes
`external_id`	id, product_id, item_id, offer_id, variant_id	Required. Your stable product id, used to update and to match product pages.
`name`	name, title, product_name, denumire	Required.
`sku`	sku, mpn, product_code, code, reference	Shown to your add-to-cart handler; matched for product-page context.
`parent_external_id`	item_group_id, parent_id, parent_sku	Groups variants (sizes, colors) of the same product.
`description`	description, body_html, descriere	HTML is converted to text.
`url`	url, link, product_url	Product page; product cards link here.
`image_urls`	image, image_link, additional_image_link, images	Several columns allowed; the first image is used on cards.
`price`	price, pret, regular_price	Number; “19,99 lei” and “€19.99” are understood.
`sale_price`	sale_price, special_price, discount_price, pret_promotional	Shown with the regular price struck through.
`currency`	currency, moneda	ISO code (EUR, RON…); defaults to the source or shop currency.
`availability`	availability, in_stock, stock_status, disponibilitate	“in stock”, “yes”, “1”, “instock”… Out-of-stock products are never recommended.
`stock_qty`	stock, quantity, qty, stoc	Optional quantity.
`categories`	category, categories, product_type, google_product_category	“Face > Serums” paths and several columns allowed.
`brand`	brand, manufacturer, vendor, marca
`locale`	locale, language, lang	For multi-language catalogs; one row per product and language.

Sync behaviour

Products are matched by external_id (+ locale). Re-importing updates existing products; nothing is duplicated.
Full sync (default for files and feeds): products missing from the latest file are deactivated (never deleted) so they stop being recommended.
Price and stock changes apply immediately and do not trigger re-enrichment. Changes to name, description or attributes re-run enrichment for that product.
Rows with errors (missing id or name, invalid price) are skipped and listed in the import report; the rest is imported.
Files up to 25 MB. Larger catalogs: use a feed URL or the Push API.

Enrichment

After import, each new or changed product is analysed by AI to extract what shoppers ask about: use cases, skin/hair types, concerns, key ingredients or specs, routine steps and safety notes. The result is embedded for semantic search. Large imports are processed in batches at a lower cost; the Catalog page shows progress and lets you inspect each product’s enrichment.

Better input, better answers

Detailed descriptions and attributes (ingredients, materials, compatibility, sizes) make the biggest difference. Out-of-stock flags must be accurate, because the assistant never recommends unavailable products.

Need real-time updates from your backend? See the Catalog Push API.