Amazon Product Scraper

Product detail for any ASIN, on fifteen Amazon marketplaces. Give it ids or paste product links and get one row per product: title, brand, the price you would pay and the price it is compared against, the star rating and how many ratings it rests on, the bullet points, the full specification tables, every gallery image at full size, and the sibling ASINs when the product is one of a variation set.

A pasted link works without editing. The ASIN is lifted out of the URL, so https://www.amazon.com/dp/B0DJGDC3BD and B0DJGDC3BD are the same input.

The marketplace decides more than the domain

Price, availability and every label on the page follow the marketplace, and so does the currency. currency is read off the page as written rather than inferred from the domain, because a page can quote a second currency and an inferred value would be wrong without ever looking wrong.

Available: com, co.uk, de, fr, it, es, nl, se, pl, ca, com.au, com.mx, com.br, in, co.jp.

Two prices, and why both are here

price is what you would pay. list_price is the struck-through comparison Amazon shows next to it, when there is one. They live in the same block in the markup and are easy to confuse, which is exactly why they are separate columns here rather than one "price" that sometimes means the other thing.

Each has a _text twin holding the figure exactly as displayed. When the number and the text disagree, trust the text: it is what the page said.

Specifications come from four different tables

Amazon uses at least four layouts for the same information and a product carries whichever one its category was built with. All of them are read and merged into a single details mapping, so a laptop's "Screen Size" and a book's "Publisher" arrive the same way.

bullets is the About this item list, cleaned of the "Report an issue" action Amazon puts in the same list.

Images

images is the gallery at full size, largest variant first. These come out of the page's own gallery data rather than the visible img tags, which carry thumbnails only, so the URLs are the ones worth downloading.

Run sizes and throttling

Amazon rate-limits harder than most. Parallel requests defaults to 2 on purpose: low and slow finishes a large batch sooner than fast and blocked. A refused page is retried with a fresh fingerprint rather than parsed into empty fields, and Attempts per product is that budget.

Errors

Code	Meaning
not_found	The page carries Amazon's own 404 marker
blocked	The page came back too small, or as a challenge
not_parsed	A full page arrived with no title in it
fetch_failed	The request could not be completed
bad_input	No ASIN or product URL was given

not_found and blocked are deliberately separate, and not_parsed is separate again. A missing product, a refusal, and a page whose markup moved are three different problems with three different responses, and reporting all three as "no data" would hide which one you have.

Only products that were actually read are billed. A block is not your problem.

Related actors

amazon-buybox for the buying options and who is selling, amazon-offers for every offer on a product, amazon-reviews-scraper for the reviews behind the star rating, amazon-bestsellers for category rankings and amazon-keywords for search-box autocomplete.