# API v2 changelog

Every operation's versions, newest first.

- 2026-11-17, `x/profiles/get` 1.0.0: First version on the API v2, in place of the v1 Twitter Profile Scraper: a profile by its handle or its address, with its counts, verification, affiliation, professional profile and images, read with your own connected X account.
- 2026-11-17, `x/tweets/list` 1.1.0: Two more fields, both null on most tweets: `place`, the place a tweet's author tagged it with, and `author.affiliation`, the organization whose badge the author carries next to its name. `author.website_url` is null, no longer an empty string, when the profile has no link.
- 2026-11-17, `x/tweets/list` 1.0.0: First version on the API v2, in place of the v1 Twitter User Tweets Scraper: a profile by its handle or its address, every tab of its page (posts, replies, reposts, videos, photos), about 20 tweets a page with their counts, media, author, quoted tweet and edits, read with your own connected X account.
- 2026-11-17, `x/tweets/search` 1.1.0: Two more fields, both null on most tweets: `place`, the place a tweet's author tagged it with, and `author.affiliation`, the organization whose badge the author carries next to its name. `author.website_url` is null, no longer an empty string, when the profile has no link.
- 2026-11-16, `lacentrale/listings/check` 2.0.1: source_url is the listing's page on lacentrale.fr, online or not, as for every operation: the address a person opens.
- 2026-11-16, `lacentrale/listings/check` 2.0.0: status_code is gone from the record: a listing online is found in LaCentrale's search and its page is no longer asked, so there is no HTTP status of the page to give. `status` says it all: online, blocked or gone.
- 2026-11-16, `lacentrale/listings/check` 1.0.0: First version on the API v2, in place of the v1 LaCentrale Listing Status Checker: one listing a call, no account, by URL, reference or numeric id. A listing online says when it was published and last updated; one LaCentrale has taken offline says when, for as long as LaCentrale holds it.
- 2026-11-16, `x/tweets/search` 1.0.0: First version on the API v2, in place of the v1 Twitter Search Results Scraper: a search as typed in X's box, its Top or Latest tab, 20 tweets a page with their counts, media, author, quoted tweet and edits, read with your own connected X account, which goes out from the country you use it from.
- 2026-11-15, `realtor/listings/get` 1.1.0: `ids` also holds the two ids realtor.com gives with some listings' source, `z_pals_id` and `z_property_id`, under their own names, null when it gives none.
- 2026-11-15, `realtor/listings/search` 1.2.0: `ids` also holds the two ids realtor.com gives with some listings' source, `z_pals_id` and `z_property_id`, under their own names, null when it gives none.
- 2026-11-13, `google_maps/places/search` 1.3.0: `next` keeps the search as it was written and goes on with `cursor`, opaque, as every paged operation of the API (Google's page tokens): copy `next` as before. `page` still starts at any page.
- 2026-11-13, `google_maps/reviews/list` 1.1.0: `cursor` is opaque, as every cursor of the API: copy it from `next` as before. It holds Google's token sealed with the place and sort it continues, and is refused with another (`invalid_cursor`).
- 2026-11-13, `google_news/articles/search` 2.0.0: `id` is now short: 16 characters, the same for an article's publisher URL in every search and edition. Google's own id (CBMi..., about 250 characters, 44 % of an answer) moves to `ids.article_id`. A search of 100 articles with a few `fields` now fits what an agent reads in one call.
- 2026-11-13, `lacentrale/listings/get` 1.0.0: Get one complete vehicle listing by URL or id, with full equipment, seller comments, warranty details, electric specifications, history preview, maintenance estimates, gallery and public seller contacts. 4 credits per successful listing.
- 2026-11-13, `lacentrale/listings/search` 2.0.0: `next` keeps the search as it was written and goes on with `cursor`, opaque, as every paged operation of the API (Google's page tokens): copy `next` as before. `search` is no longer an option: the ranking context travels inside the cursor. The URL's page still starts at any page.
- 2026-11-13, `lacentrale/listings/search` 1.0.0: First version: 23 standard results per page from LaCentrale’s search API, with vehicle specifications, public seller phones, photo URLs, financing, delivery and warranty data. Continuations retain ranking context. Up to 347 pages per search; 23 credits per nonempty call.
- 2026-11-13, `leboncoin/listings/search` 2.1.0: `next` keeps the search as it was written and goes on with `cursor`, opaque, as every paged operation of the API (Google's page tokens): copy `next` as before. The URL is no longer rewritten with the next page; its `page` still starts at any page.
- 2026-11-13, `leboncoin/listings/search` 2.0.0: `since` is gone: Leboncoin has no date filter, it was ours, and a page is now served exactly as Leboncoin serves it. To keep only the newest listings, a batch filters on `republished_at` and stops early on a search sorted by date, newest first.
- 2026-11-13, `realtor/listings/get` 1.0.0: First version on the API v2: one realtor.com listing as its page shows it read to the end: description and features, agents and their contacts, office, broker and MLS, price and sale history, taxes and values, permits, schools, environmental risks, value estimates, monthly payment, HOA, photos and tours; with the MLS's wish on AI use.
- 2026-11-13, `realtor/listings/search` 1.1.0: `next` keeps the search as it was written and goes on with `cursor`, opaque, as every paged operation of the API (Google's page tokens): copy `next` as before. The URL is no longer rewritten with the next page; its pg-N still starts at any page.
- 2026-11-13, `realtor/listings/search` 1.0.0: First version on the API v2: one page of a realtor.com search, 42 homes for sale or sold, with price, beds, baths, surfaces, address, coordinates, status, photos, tours, open house, the brokerage or builder, the contact line, the MLS and its ids, and the MLS's wish on AI use; a filter realtor.com does not read is refused, never ignored.
- 2026-11-13, `seloger/listings/search` 2.0.0: `next` keeps the search as it was written and goes on with `cursor`, opaque, as every paged operation of the API (Google's page tokens): copy `next` as before. `search` is no longer an option: the criteria SeLoger's page reads for a search by travel time or with no place travel inside the cursor. The URL's `page` still starts at any page.
- 2026-11-12, `google_news/articles/search` 1.0.0: First version on the API v2: one Google News search of up to 100 articles, as Google ranks them, each with its publisher's URL, time, publisher, authors, image and story; by edition, site, the last hour, day or week, or days.
- 2026-11-12, `seloger/listings/get` 1.0.0: First version on the API v2: one seloger.com listing, whole, by its id or address: the full description, every feature, the energy certificates, each line of the price, the co-ownership, the schools nearby, SeLoger's price per m² of the area, and the seller with its phones, SIREN and SIRET.
- 2026-11-12, `seloger/listings/search` 1.2.0: Every page as fresh as SeLoger's search, the first included, and lighter: a search URL is read as SeLoger's own page reads it, then asked of its search service (about 47 KB a page instead of 170 KB). SeLoger's page is kept up to 15 minutes by its CDN: its newest listings came late, and the last of page 1 came again on page 2. `next` is a URL again; `search` only follows a search by travel time or with no place, read from its page.
- 2026-11-12, `seloger/listings/search` 1.1.0: A next page is lighter: `next` now holds `search`, the search as SeLoger reads its URL, and a call with it asks SeLoger's search service, as the site does from page 2 (about 47 KB instead of 170 KB). Same listings, same fields, same price; a URL alone, any page, is read as before.
- 2026-11-12, `seloger/listings/search` 1.0.0: First version on the API v2: one page of a seloger.com search as the page shows it, 30 listings with their previous price and its change, photos with their room, transport nearby, and the seller with its phone numbers, SIREN and SIRET; a filter SeLoger does not read is refused, never ignored.
- 2026-11-12, `websites/contacts/get` 1.1.0: More phones and a few more emails, read as a person reads them: a number written after its label (Tél., Phone:, Telefon...), the mailto and tel: links a site builder keeps in the page's scripts, links written without quotes, a home page that frames another address. On 9,864 company websites of 45 countries, a phone on 71.8 % of the sites read, from 65.1 % (82 % from 71 % in France), and in E.164 on 69.9 %, from 54.8 %: numbers with a 0 kept after their country (+33 0 1...) are given right, and `country` reads 45 countries' numbers. Doctolib is no longer among the profiles. A page heavy with inline images no longer takes seconds to read.
- 2026-11-10, `leboncoin/listings/check` 1.0.0: First version on the API v2, in place of the v1 Leboncoin Listing Status Checker: one listing a call, no account, its state (active, paused, gone) with Leboncoin's status code and the listing's dates.
- 2026-11-10, `leboncoin/listings/get` 1.0.0: First version on the API v2, in place of the v1 Leboncoin Listing Scraper: one listing a call, whole, with the same fields as a listing of leboncoin/listings/search, and how many people saved it (`favorites`).
- 2026-11-10, `websites/contacts/get` 1.0.0: First version on the API v2, in place of the v1 Contact Details Scraper: one site a call, its emails, phones and profiles (25 social networks and directories), each with the page it is on.
- 2026-11-09, `google_maps/places/get` 2.0.1: A well-formed place id that names no place answers place_not_found (404, not charged), as it should; it answered unexpected_answer.
- 2026-11-08, `google_maps/places/get` 2.0.0: One input for the place: `place` takes a name with its city too ("Le Vieux Lyon, Lyon"), as it takes an id, a cid, a feature id or a link; `query` is gone (a call that still sends it answers invalid_options: send the name as `place`).
- 2026-11-02, `leboncoin/listings/search` 1.0.0: First version on the API v2: one page of a leboncoin.fr search as the page shows it, featured copies in their place, with every description and attribute; `next` carries the page's cursor, so a search read page by page neither skips nor repeats a listing.
- 2026-11-02, `leboncoin/phones/get` 1.0.2: A listing no longer online that Leboncoin answers with a 410 reads phone.status none too (it read as site_unavailable). Leboncoin's anti-bot page is recognized by its content whatever its status (403, 404, 202), so a blocked call is never read as a listing without a number.
- 2026-11-02, `leboncoin/phones/get` 1.0.1: A listing without the "Voir le numéro" button, or no longer online, reads phone.status none (free) instead of the error listing_not_found: Leboncoin answers both the same way, and a listing without a number was said removed when it was not.
- 2026-11-02, `leboncoin/phones/get` 1.0.0: First version on the API v2.
- 2026-10-27, `google_maps/places/get` 1.2.0: primary_category_id said it was the main category's id but held Google's search-result type, a family for most places (shopping for a sewing machine store). That value is now result_type. primary_category_id is the main category's real id, which only a place's page gives: places/get fills it, places/search leaves it null.
- 2026-10-27, `google_maps/places/get` 1.1.0: Three fields Google gives on a place's page: rating.distribution (reviews per number of stars), review_topics (the topics drawn from the reviews, with their mentions) and owner_description (what the place says of itself). Found by the preventive loop (lobstrio/platform#19).
- 2026-10-27, `google_maps/places/search` 1.2.0: primary_category_id said it was the main category's id but held Google's search-result type, a family for most places (shopping for a sewing machine store). That value is now result_type. primary_category_id is the main category's real id, which only a place's page gives: places/get fills it, places/search leaves it null.
- 2026-10-27, `google_maps/places/search` 1.1.0: Three fields Google gives on a place's page: rating.distribution (reviews per number of stars), review_topics (the topics drawn from the reviews, with their mentions) and owner_description (what the place says of itself). Found by the preventive loop (lobstrio/platform#19).
- 2026-10-26, `google_maps/places/get` 1.0.0: First version on the API v2.
- 2026-10-26, `google_maps/places/search` 1.0.0: First version on the API v2.
- 2026-10-26, `google_maps/reviews/list` 1.0.0: First version on the API v2.
