Yandex Maps Scraper

Organisations from Yandex Maps, searched by name, category or address, centred anywhere in the world. Each row carries the rating with its two counts kept apart, the phone numbers, the websites, the social and booking links, the opening hours with the current open state, every category, the amenity features, nearby metro stations, the coordinates, the photo count — and the organisation's own id and page address, which the Place and Reviews Actors take back verbatim.

The contact details are on the search row. Most place directories hold phone numbers and websites back for a per-business lookup; here a whole page of twenty-five organisations arrives with them attached.

Accepted input

queries is required and takes one search term per line — a business name, a category, or an address. A term is never split on a comma. An address returns a toponym row rather than an organisation, which is a real answer with no rating and no page of its own.

latitude and longitude set the centre of the search. Send both or neither. One without the other is silently ignored by Yandex, which then answers from its own default region — with no centre at all, the term coffee measured as returning a cafe in a small town in Krasnodar Krai. They are two named fields rather than one pasted lat,lon string on purpose: Yandex's own viewport parameter is longitude-first, and a transposed pair is how a search for New York quietly returns results in Somalia.

span is the half-width of the search area in degrees, applied to both axes. 0.05 is roughly a city centre. Leave it empty to let Yandex choose.

bounds decides what happens to results outside that area. widen is Yandex's own behaviour and searches beyond it; restrict refuses them and needs a span to restrict to. Measured: a Red Square query with a New York area returns Moscow's Red Square when widening and nothing when restricting.

language picks the language titles, addresses and opening hours come back in: English, Russian, Turkish or Ukrainian. That is a closed set — a locale outside it is refused upstream rather than falling back, so offering more would be offering a failure.

startPage defaults to 0 and is where each term's walk begins.

maxItems defaults to 100 and caps each term separately. 0 means no limit — the walk then continues until a page carries nothing new, which is the only honest end on this surface. Why, is answered under Frequently asked questions.

There is no page-size input. It is always the twenty-five Yandex serves: it answers 26 with a failure rather than clamping, and a smaller page divides one request's cost among fewer rows for nothing.

{
  "queries": ["coffee"],
  "latitude": 40.7484,
  "longitude": -73.9857,
  "span": 0.05,
  "bounds": "widen",
  "language": "en_US",
  "maxItems": 100
}

Response fields

{
  "query": "coffee",
  "page": 0,
  "language": "en_US",
  "total_result_count_estimate": 89,
  "id": "1018907821",
  "type": "business",
  "title": "Pushkin",
  "short_title": "Pushkin",
  "description": "Restaurant",
  "url": "https://yandex.com/maps/org/pushkin/1018907821/",
  "uri": "ymapsbm1://org?oid=1018907821",
  "seoname": "pushkin",
  "is_advert": false,
  "address": "Tverskoy Blvd, 26А",
  "full_address": "Moscow, Tverskoy Blvd, 26А",
  "postal_code": "125009",
  "country": "Russia",
  "latitude": 55.7649,
  "longitude": 37.6039,
  "geo_id": 213,
  "metro": [{ "name": "Пушкинская", "distance": 320 }],
  "categories": [{ "name": "Restaurant" }],
  "features": [{ "name": "Average bill", "value": "2500 ₽" }],
  "phones": [{ "type": "phone", "number": "+7 495 739-00-33" }],
  "websites": ["https://cafe-pushkin.ru"],
  "working_time_text": "Daily, 12:00 AM – 12:00 AM",
  "rating_value": 4.6,
  "rating_count": 15219,
  "review_count": 6001,
  "photo_count": 4210
}

Fields absent from a record are returned as null rather than omitted, so every row has the same shape.

rating_count and review_count are two different numbers and both are published. Yandex counts people who rated separately from people who wrote, and the second is the one Yandex Maps Reviews Scraper pages through. A single column called "reviews" would silently be one of the two.

id and url are both published so nothing ever has to build an address. Either is accepted by the Place and Reviews Actors, which is how the three compose.

type is business for an organisation and toponym for an address or a street. A toponym carries no rating, no phone numbers and no page of its own; searching for an address returns one, and it is a real answer rather than a broken row.

is_advert is true on a paid placement inside the map results. Yandex renders those in the same shape as organic rows with nothing distinguishing them, exactly as it does on the web results.

total_result_count_estimate is named for what it is: an estimate, and it grows as the walk deepens. One six-page walk of the same term reported 89, 114, 169, 227, 284 and 338. Do not compute a page count from it — this Actor does not either.

categories, features, phones, websites, metro, chain and working_status are republished in Yandex's own shape. There is no measured field list to rename them into, and renaming fields nobody has seen populated publishes a permanently-null column and calls it a schema.

Behaviour on partial results

The walk ends on an empty page, or on a page that carried nothing new. There is no "has more" signal here and the stated total grows as you page, so a page of repeats is the only honest stop condition — and it is a real one, because a map page is stable: fetched twice from one pinned route, the first page returned all twenty-five rows in the same order both times.

Rows repeat as the walk deepens and the repeats are dropped here. One measured six-page walk returned 150 rows of which 100 were distinct, so a run that fetched six pages can legitimately save a hundred rows. The run log says how many it skipped.

A term that answers with nothing is skipped with a line in the log and the rest of the list still runs. One coordinate without the other, or restricting without an area size, stops the run before any request is made.

Coverage is deepest in Russia and the CIS. A Moscow restaurant carried 15,219 ratings and 6,001 written reviews while Manhattan coffee shops carried none — that is real sparsity in Yandex's data rather than a field this Actor failed to read.

Frequently asked questions

Why is there no total I can plan a harvest against? Because Yandex's own number is an estimate and it moves as you page: 89, 114, 169, 227, 284 and 338 across one six-page walk of the same query. Dividing it by the page size would ask for pages that do not exist, and treating it as a row count would report a result set nearly twice the size of the one you collected. It is published under a name that says it is an estimate, and the walk stops on what it actually gets.

Why did I get fewer rows than pages × 25? Because deeper pages repeat organisations from earlier ones and the repeats are dropped. That is paging drift rather than resampling — a single page fetched twice is identical, in the same order.

I searched without a centre and got results from somewhere random. That is Yandex's default region answering. Set latitude and longitude; one without the other is ignored rather than refused, which is why this Actor stops the run instead of letting it happen.

Why are latitude and longitude separate fields? Because Yandex's own viewport parameter is longitude-first. A single "lat,lon" string handed through unchanged is how a request for New York returns results in Somalia. Two named fields cannot be transposed by accident.

Why is restrict refusing everything? Because it needs an area to restrict to and the results genuinely fall outside it. Widening is Yandex's own behaviour and is usually what you want; restricting is for when a query would otherwise pull in a famous match from another country.

Can I get more than the languages offered? No. Those four are what Yandex Maps answers; anything else is a failure upstream rather than a fallback to a default, so offering it would be offering a broken run.

Is a Yandex account or API key required? No. No account, session cookie or key of any kind is supplied to the Actor or needed by it.

Related Actors

Yandex Maps Place Scraper takes the id or url from any row here and reads one organisation on its own. It costs a whole request per organisation — twenty-five times a row here — so it is for looking up a specific business rather than for reading an area.

Yandex Maps Reviews Scraper takes the same id and exports the written reviews, with the reviewer's contributor level and Yandex's own per-aspect summary.

Yandex Search Scraper is the other half of this site: web results with paid placements flagged. Unlike the map surface it resamples between identical requests, and its README says so.