gmbscraper

A typed, importable Playwright scraper for Google Maps / Google Business Profile data: place search, place details, identity-based matching, review scraping, and parallel fan-out helpers.

• Search Google Maps and scrape place details (name, category, address, phone, website, rating, review count) without the official (paid, rate-limited) Places API.
• Adaptive quadrant search that recursively splits the map viewport to pull results beyond Google's ~120-per-query cap.
• Confidence-tiered identity matching (phone_and_domain_match > phone_match > domain_match > exact_name_match) for matching a known business/outlet against scraped candidates.
• Review scraping with owner-response filtering, "see more" text expansion, and multilingual review-tab detection.
• Thread-based fan-out helpers (chunk_items, run_chunked) for running multiple browser workers over a batch of items.

Status: pre-1.0 (0.x). Extracted from duplicated scraping code across several internal projects; the API may still move before 1.0.0.

Using this from another project (agent or human)? Read skills/SKILL.md first — it covers the API surface, matching semantics, parallel fan-out, and how to add this as a dependency before PyPI publication.

Install

Not yet published to PyPI. Add it as a uv path dependency — see skills/SKILL.md for the exact pyproject.toml snippet — then:

uv run playwright install --with-deps chromium

Quickstart

from gmbscraper import BusinessIdentity, OutletIdentity, google_maps_browser

with google_maps_browser() as maps:
    # Search + adaptive pagination
    links = maps.search_places_adaptive("plumbers", latitude=-33.87, longitude=151.21)

    # Scrape a single place page
    place = maps.scrape_place(links[0].maps_url, fallback_name=links[0].fallback_name)
    print(place.name, place.rating, place.review_count)

    # Find the Google Business Profile for a known business
    match = maps.search_and_match(
        OutletIdentity(name="Acme Plumbing", phone="0400 000 000", address="1 Main St, Sydney"),
        BusinessIdentity(name="Acme Plumbing Pty Ltd", website="acmeplumbing.com.au"),
    )
    if match:
        print(match.match_reason, match.place_id)

    # Scrape reviews; on a re-sync, pass the ids you already store to stop
    # scrolling at the first known review
    reviews = maps.scrape_reviews(place.maps_url, max_reviews=50)
    newer = maps.scrape_reviews(place.maps_url, known_ids={r.external_id for r in reviews})

Design notes

• Every network-facing function swallows and skips broken selectors rather than raising — a PlaceProfile is returned with whatever fields could be read, since Google's DOM/markup changes without notice and partial data is usually more useful than a hard failure. scrape_reviews is the exception: it raises GoogleReviewsUnavailableError when Google serves a page with no reviews UI at all (throttled/limited view, or the reviews tab never opens), since a caller needs to distinguish "zero reviews" from "couldn't scrape." Its GoogleLimitedViewError subclass marks Google's signed-out limited view, which signals throttling rather than a problem with one place.
• Each place opens in a single page load, and each browser launch gets a random desktop Chrome user agent unless you pass user_agent=.
• There's no retry/backoff or proxy rotation built in. This is a DOM scraper, not a hosted anti-block service — for scraping at a scale where blocking is routine, a paid API (Outscraper, SerpApi, etc.) will be cheaper than building that infrastructure yourself.

Development

uv sync
uv run pytest
uv run ruff check .
uv run ruff format --check .

See skills/references/development.md for what is and isn't covered by the test suite.

License

MIT