Agriaffaires / MachineryZone

Search or browse either marketplace. Combine free text with the typed filters the source supports natively — category, manufacturer, model, advert type, year of construction, hour meter, price band, seller country, region and department, professional or private, with a photo, posted this week — and sort by price, year, hours, model, manufacturer or listing date. 50 rows a page. 🔴 Depth is capped by the source at 200 pages = 10,000 rows per query and page 201 silently serves page 1, so narrow rather than page; the filters are measured to bite hard (22,115 rows -> 6,787 by a year and price window -> 947 by one manufacturer).

One listing in full, by id. Returns the machine (manufacturer, model, year of construction, condition, availability, hour meter, engine power), the price with its currency — cross-checked against the page's own schema.org product data and reported if the two disagree — every technical property the seller typed with the source's own label, the raw number, the real unit and the printed unit label, the seller's own description, the full gallery, the category breadcrumb and the business behind the advert. An auction's price row is returned as `auction_start_price`, never as an asking price.

The marketplace's own category tree with the numeric ids every other action takes and the source's own live counts. Agriaffaires publishes some 320 categories under its machinery families, MachineryZone its own set; ids are language-invariant, so the id you get here works on all 43 hosts. 🔴 The tree lives on exactly one page (the homepage); result pages load it over AJAX and do not contain it, so this action reads the homepage for the hierarchy and the source's own facet list for the ids, and says in `meta` how many it could match.

The source's own live filter vocabulary and counts for a query: which categories and manufacturers have stock and how much, which conditions, gearboxes, drive and equipment values exist for that category, and the real minimum and maximum of every numeric property it offers. 🔴 This is the action that keeps you out of this source's worst behaviour — a value it does not recognise does not error, it answers HTTP 200 with zero rows and a total of 0 — so take manufacturer ids and property values from here rather than guessing them.

The site's own autocomplete for a typed prefix, in two buckets: whole search phrases people use, and categories with their numeric ids and live counts. Use it to turn what a user typed into values the `category_id` and `query` filters actually accept.

The dealer directory — the supplier side of the marketplace. Company name, activity, country, town, department, how many further addresses the group has, the manufacturer ranges they carry and how many adverts they have online, 50 a page. Searchable by company name, filterable by country, and ordered by the source's own sort values — company name, or how many adverts a dealer has online.

One dealer's public business profile by id: company name, activity, registration number where the source prints one, how long they have been on the marketplace, their own website, their follower count, the manufacturer ranges they carry and the breakdown of their stock by category with live counts. Contact people, phone numbers and e-mail addresses are deliberately not returned.

One dealer's actual stock, paged — the same row shape as `search`. Unlike some machinery marketplaces, this source's dealer pages carry REAL listings rather than login-walled placeholders (measured: 50 genuine rows with distinct ids, prices and years, over 13 pages for one dealer), so they are returned.

The marketplace's own PRICE OBSERVATORY — what a make, model and year actually sells for on this marketplace: minimum, maximum and average asking price plus how many listings the figure rests on. Call it with nothing for the rubric list, with `rubric_id` for its makes, with `rubric_id` + `make_id` for its models, and with `model` + `year` on top for the price band itself. 🔴 This sub-system keeps its own id space: rubric 510 and make 4039 are the farm tractor and John Deere here, while `search` calls the same two 702 and 7607 — and giving it a search id answers an EMPTY result rather than an error, so always take the ids from this action.

The geography the filters accept, from the source itself: the countries this engine will take, then a country's regions and a region's departments with the ids `search` wants. 🔴 It exists because an area id the source does not know does not error — it answers a different, wrong result set — so the only safe ids are the ones it hands out itself.