bugrahans's picture
Upload 4 files
3814b7e verified
Raw History Blame Contribute Delete
24.2 kB
from __future__ import annotations
import json
import os
import re
import time
import urllib.parse
from typing import Any, Dict, List, Optional, Tuple
import requests
from smolagents.tools import Tool
SCRAPE_DO_SIGNUP_URL = (
"https://dashboard.scrape.do/sign-up?utm_source=huggingface&utm_medium=google-maps-scraper"
)
MAPS_BASE = "https://api.scrape.do/plugin/google/maps"
SEARCH_ENDPOINT = MAPS_BASE + "/search"
PLACE_ENDPOINT = MAPS_BASE + "/place"
REVIEWS_ENDPOINT = MAPS_BASE + "/reviews"
# From the Scrape.do docs: every Maps call costs 10 credits.
CREDITS_PER_CALL = 10
PLACES_PER_PAGE = 20
MAX_SEARCH_PAGES = 6 # Google Maps stops at ~120 places.
REVIEWS_PER_CALL = 20
MAX_REVIEWS = 100
# (label, gl, hl, google_domain)
COUNTRIES: List[tuple] = [
("United States", "us", "en", "google.com"),
("United Kingdom", "gb", "en", "google.co.uk"),
("Germany", "de", "de", "google.de"),
("France", "fr", "fr", "google.fr"),
("Spain", "es", "es", "google.es"),
("Italy", "it", "it", "google.it"),
("Netherlands", "nl", "nl", "google.nl"),
("Turkey", "tr", "tr", "google.com.tr"),
("Canada", "ca", "en", "google.ca"),
("Australia", "au", "en", "google.com.au"),
("India", "in", "en", "google.co.in"),
("Japan", "jp", "ja", "google.co.jp"),
("Brazil", "br", "pt", "google.com.br"),
("Mexico", "mx", "es", "google.com.mx"),
]
COUNTRY_BY_GL = {gl: (label, gl, hl, domain) for label, gl, hl, domain in COUNTRIES}
SORT_OPTIONS = {
"relevance": None,
"newest": "newestFirst",
"highest": "ratingHigh",
"lowest": "ratingLow",
}
SHORT_LINK_HOSTS = ("maps.app.goo.gl", "goo.gl", "g.co")
# ------------------------------------------------------------------ helpers
def _locale(country: Optional[str]) -> Tuple[str, str, str]:
gl = (country or "us").strip().lower()
_, gl, hl, domain = COUNTRY_BY_GL.get(gl, ("", gl, "en", "google.com"))
return gl, hl, domain
def normalize_map_center(value: Optional[str]) -> Tuple[Optional[str], Optional[str]]:
"""Turn '41.0082,28.9784' or '@41.0082,28.9784,14z' into Google's `ll` format."""
if not value or not str(value).strip():
return None, None
parts = str(value).strip().lstrip("@").replace(" ", "").split(",")
try:
lat, lng = float(parts[0]), float(parts[1])
except (IndexError, ValueError):
return None, "Map center ignored: use the format 41.0082,28.9784"
if not (-90 <= lat <= 90 and -180 <= lng <= 180):
return None, "Map center ignored: those coordinates are out of range."
zoom = 14
if len(parts) > 2:
match = re.match(r"(\d+(?:\.\d+)?)", parts[2])
if match:
zoom = int(float(match.group(1)))
zoom = max(1, min(zoom, 21))
return f"@{lat},{lng},{zoom}z", None
def parse_place_ids(text: str) -> Dict[str, str]:
"""Pull whatever place identifiers a Google Maps link or pasted ID contains."""
decoded = urllib.parse.unquote(text or "").strip()
ids: Dict[str, str] = {}
# data_id, e.g. 0x89c259606c152b2d:0xe230af1b18e542a7 (inside `!1s...` in place links).
# Its second half, read as a number, is the place's CID.
match = re.search(r"(0x[0-9a-fA-F]{4,}):(0x[0-9a-fA-F]{4,})", decoded)
if match:
ids["data_id"] = f"{match.group(1)}:{match.group(2)}".lower()
ids["data_cid"] = str(int(match.group(2), 16))
match = re.search(r"[?&]cid=(\d{5,})", decoded)
if match and "data_cid" not in ids:
ids["data_cid"] = match.group(1)
match = re.search(r"(?:place_id[:=]|query_place_id=|placeid=)([A-Za-z0-9_-]{20,})", decoded)
if match:
ids["place_id"] = match.group(1)
else:
match = re.search(r"\b(ChIJ[A-Za-z0-9_-]{15,})", decoded)
if match:
ids["place_id"] = match.group(1)
if not ids and re.fullmatch(r"\d{10,20}", decoded):
ids["data_cid"] = decoded
return ids
def _expand_short_link(url: str) -> Optional[str]:
"""Follow a maps.app.goo.gl link's redirects without downloading the final page."""
current = url
for _ in range(6):
try:
response = requests.get(
current,
allow_redirects=False,
timeout=15,
headers={"User-Agent": "Mozilla/5.0 (compatible; maps-link-resolver)"},
)
except requests.exceptions.RequestException:
return None
location = response.headers.get("Location")
if not location:
return current if current != url else None
location = urllib.parse.urljoin(current, location)
# EU visitors get bounced through a consent page that carries the real link.
if "consent.google" in location:
params = urllib.parse.parse_qs(urllib.parse.urlparse(location).query)
if params.get("continue"):
return params["continue"][0]
current = location
if parse_place_ids(current):
return current
return current
def _is_url(text: str) -> bool:
return text.lower().startswith(("http://", "https://"))
# --------------------------------------------------------------------- tool
class ScrapeDoGoogleMapsTool(Tool):
name = "google_maps_scraper"
description = """
Scrape Google Maps. Three modes:
- 'search': find businesses for a query such as 'dentists in Kadikoy Istanbul'
and get up to 120 places with name, category, rating, review count, phone,
website, address, opening hours and GPS coordinates. Good for lead lists.
- 'place': full details for one business: hours, rating breakdown, popular
times, service options, photos, menu link and review highlights.
- 'reviews': a business's user reviews with rating, date, text, reviewer
and owner replies, sortable by newest or rating, filterable by keyword.
For 'place' and 'reviews', pass a Google Maps link, a place ID (ChIJ...),
a data ID (0x...:0x...), a CID, or just the business name.
Powered by Scrape.do's Google Maps API, which handles proxies and CAPTCHAs.
"""
output_type = "string"
def __init__(self, api_token: Optional[str] = None) -> None:
self._api_token = api_token
self.inputs = {
"mode": {
"type": "string",
"description": "'search' to find businesses, 'place' for one business's details, 'reviews' for its user reviews.",
},
"query": {
"type": "string",
"description": (
"For 'search': what to look for, ideally with a place, e.g. 'coffee shops in Austin TX'. "
"For 'place' and 'reviews': a Google Maps link, place ID, data ID, CID or business name."
),
},
"country": {
"type": "string",
"description": "Two-letter country code such as 'us', 'gb', 'de', 'tr'. Sets language and Google domain. Defaults to 'us'.",
"nullable": True,
},
"pages": {
"type": "integer",
"description": "'search' only: result pages to fetch (1-6, 20 places each, 10 credits each). Defaults to 1.",
"nullable": True,
},
"max_reviews": {
"type": "integer",
"description": "'reviews' only: how many reviews to collect (1-100, 20 per 10-credit call). Defaults to 20.",
"nullable": True,
},
"sort_by": {
"type": "string",
"description": "'reviews' only: 'relevance', 'newest', 'highest' or 'lowest'. Defaults to relevance.",
"nullable": True,
},
"keyword": {
"type": "string",
"description": "'reviews' only: only return reviews that mention this word, e.g. 'parking'.",
"nullable": True,
},
"location": {
"type": "string",
"description": "'search' only: map center as 'latitude,longitude' (optionally ',zoom'), e.g. '41.0082,28.9784'.",
"nullable": True,
},
}
super().__init__()
# ------------------------------------------------------------ plumbing
def _resolve_token(self) -> Optional[str]:
if self._api_token:
return self._api_token.strip()
env_token = os.getenv("SCRAPE_DO_API_TOKEN")
return env_token.strip() if env_token else None
@staticmethod
def _missing_token() -> Dict[str, Any]:
return {
"status": "error",
"errorMessage": "Missing Scrape.do API token.",
"hint": f"Get a free token at {SCRAPE_DO_SIGNUP_URL}",
}
@staticmethod
def _as_error(data: Dict[str, Any]) -> Dict[str, Any]:
result = {"status": "error", "errorMessage": data.get("_error", "Unknown error")}
if data.get("_auth"):
result["hint"] = f"Check your token or create one at {SCRAPE_DO_SIGNUP_URL}"
return result
def _get(self, url: str, params: Dict[str, Any], retries: int = 1) -> Dict[str, Any]:
"""GET a Scrape.do endpoint. Retries once on the transient 5xx errors the docs mention."""
token = str(params.get("token") or "")
def redact(text: str) -> str:
return text.replace(token, "***") if token else text
full_url = url + "?" + urllib.parse.urlencode(params, quote_via=urllib.parse.quote_plus)
last: Dict[str, Any] = {"_error": "Request failed."}
for attempt in range(retries + 1):
try:
response = requests.get(full_url, timeout=120)
except requests.exceptions.RequestException as exc:
last = {"_error": redact(f"Request failed: {exc}")}
if attempt < retries:
time.sleep(1.5)
continue
return last
if response.status_code == 401:
return {"_error": "Scrape.do rejected the token (HTTP 401).", "_auth": True}
try:
data = response.json()
except ValueError:
data = None
if response.ok and isinstance(data, dict):
return data
message = ""
if isinstance(data, dict):
message = str(data.get("message") or data.get("error") or "")
if not message:
message = (response.text or "").strip()[:200] or "no details"
last = {
"_error": redact(f"Scrape.do returned HTTP {response.status_code}: {message}"),
"_status": response.status_code,
}
if response.status_code in (500, 502, 503, 504) and attempt < retries:
time.sleep(1.5)
continue
return last
return last
def _find_by_name(
self, token: str, name: str, locale: Tuple[str, str, str], ll: Optional[str]
) -> Tuple[Optional[Dict[str, Any]], Optional[str]]:
gl, hl, domain = locale
params: Dict[str, Any] = {"token": token, "q": name, "gl": gl, "hl": hl, "google_domain": domain}
if ll:
params["ll"] = ll
data = self._get(SEARCH_ENDPOINT, params)
if "_error" in data:
return None, data["_error"]
if isinstance(data.get("place_results"), dict):
return data["place_results"], None
results = data.get("local_results") or []
if results:
return results[0], None
return None, f"Google Maps found no place for “{name}”."
def _resolve_place(
self, token: str, ref: str, locale: Tuple[str, str, str]
) -> Dict[str, Any]:
"""Work out place IDs from a link, an ID, or a business name.
Returns {"ids": {...}, "credits": int, "note": str|None} or {"error": str}.
"""
text = (ref or "").strip()
if not text:
return {"error": "Paste a Google Maps link, a place ID, or a business name."}
host = urllib.parse.urlparse(text).netloc.lower() if _is_url(text) else ""
if host in SHORT_LINK_HOSTS:
expanded = _expand_short_link(text)
if not expanded:
return {
"error": "Couldn't open that short link. Open it in your browser, then copy "
"the full link from the address bar and paste that instead."
}
text = expanded
ids = parse_place_ids(text)
if ids:
return {"ids": ids, "credits": 0, "note": None}
ll = None
if _is_url(text):
decoded = urllib.parse.unquote(text)
if "/maps/search/" in decoded or "/maps/dir/" in decoded:
return {
"error": "That's a search or directions link, not a single place. Click one "
"business in Google Maps and copy that link, or use the Find businesses tab."
}
match = re.search(r"/maps/place/([^/@?]+)", decoded)
if not match:
return {"error": "Couldn't find a place in that link. Try the business name instead."}
name = urllib.parse.unquote_plus(match.group(1)).strip()
coords = re.search(r"@(-?\d+\.\d+),(-?\d+\.\d+),(\d+(?:\.\d+)?)z", decoded)
if coords:
ll = f"@{coords.group(1)},{coords.group(2)},{int(float(coords.group(3)))}z"
else:
name = text
place, error = self._find_by_name(token, name, locale, ll)
if error:
return {"error": error}
ids = {key: str(place[key]) for key in ("place_id", "data_id", "data_cid") if place.get(key)}
if not ids:
return {"error": f"Found “{place.get('title')}” but Google gave no ID for it."}
note = (
f"Looked up “{name}” and used the first match: {place.get('title')}"
+ (f", {place.get('address')}" if place.get("address") else "")
+ f" ({CREDITS_PER_CALL} credits)."
)
return {"ids": ids, "credits": CREDITS_PER_CALL, "note": note}
# --------------------------------------------------------------- search
def search_places(
self,
query: str,
country: str = "us",
pages: int = 1,
location: Optional[str] = None,
) -> Dict[str, Any]:
token = self._resolve_token()
if not token:
return self._missing_token()
if not query or not query.strip():
return {"status": "error", "errorMessage": "Type what you're looking for, e.g. 'dentists in Kadikoy Istanbul'."}
gl, hl, domain = _locale(country)
pages = max(1, min(int(pages or 1), MAX_SEARCH_PAGES))
base: Dict[str, Any] = {"token": token, "q": query.strip(), "gl": gl, "hl": hl, "google_domain": domain}
warnings: List[str] = []
ll, ll_warning = normalize_map_center(location)
if ll:
base["ll"] = ll
if ll_warning:
warnings.append(ll_warning)
places: List[Dict[str, Any]] = []
seen: set = set()
credits = 0
pages_fetched = 0
maps_url = None
for page_index in range(pages):
data = self._get(SEARCH_ENDPOINT, dict(base, start=page_index * PLACES_PER_PAGE))
if "_error" in data:
if page_index == 0:
return self._as_error(data)
# Deep pages sometimes answer 502 when Google has nothing more.
warnings.append(f"Stopped after page {page_index}: {data['_error']}")
break
credits += CREDITS_PER_CALL
pages_fetched += 1
if page_index == 0:
maps_url = (data.get("search_metadata") or {}).get("google_maps_url")
batch = data.get("local_results")
if not batch and isinstance(data.get("place_results"), dict):
batch = [data["place_results"]] # The query matched one exact place.
batch = batch if isinstance(batch, list) else []
for item in batch:
key = item.get("place_id") or item.get("data_id") or (item.get("title"), item.get("address"))
if key in seen:
continue
seen.add(key)
item = dict(item)
item["page"] = page_index + 1
item["position"] = len(places) + 1
places.append(item)
if len(batch) < PLACES_PER_PAGE:
break
result: Dict[str, Any] = {
"status": "success",
"query": query.strip(),
"country": gl,
"google_maps_url": maps_url,
"places": places,
"pages_fetched": pages_fetched,
"credits_used": credits,
}
if warnings:
result["warnings"] = warnings
return result
# ---------------------------------------------------------------- place
def place_details(self, place: str, country: str = "us") -> Dict[str, Any]:
token = self._resolve_token()
if not token:
return self._missing_token()
locale = _locale(country)
resolved = self._resolve_place(token, place, locale)
if "error" in resolved:
return {"status": "error", "errorMessage": resolved["error"]}
ids, credits = resolved["ids"], resolved["credits"]
gl, hl, domain = locale
base = {"token": token, "gl": gl, "hl": hl, "google_domain": domain}
attempts = []
if ids.get("place_id"):
attempts.append({"place_id": ids["place_id"]})
if ids.get("data_cid"):
attempts.append({"data_cid": ids["data_cid"]})
if not attempts:
return {"status": "error", "errorMessage": "No usable place ID found."}
data: Dict[str, Any] = {}
for identifier in attempts:
data = self._get(PLACE_ENDPOINT, dict(base, **identifier))
if "_error" not in data:
credits += CREDITS_PER_CALL
break
if "_error" in data:
return self._as_error(data)
details = data.get("place_results")
if not isinstance(details, dict) or not details:
return {"status": "error", "errorMessage": "Google Maps returned no details for this place."}
result: Dict[str, Any] = {
"status": "success",
"place": details,
"google_maps_url": (data.get("search_metadata") or {}).get("google_maps_url"),
"credits_used": credits,
}
if resolved.get("note"):
result["note"] = resolved["note"]
return result
# -------------------------------------------------------------- reviews
def reviews(
self,
place: str,
country: str = "us",
max_reviews: int = 20,
sort_by: Optional[str] = None,
keyword: Optional[str] = None,
) -> Dict[str, Any]:
token = self._resolve_token()
if not token:
return self._missing_token()
locale = _locale(country)
resolved = self._resolve_place(token, place, locale)
if "error" in resolved:
return {"status": "error", "errorMessage": resolved["error"]}
ids, credits = resolved["ids"], resolved["credits"]
gl, hl, domain = locale
warnings: List[str] = []
fallback_info: Optional[Dict[str, Any]] = None
# The reviews endpoint needs a data ID or place ID. A bare CID has
# neither, so look the place up first.
if not ids.get("data_id") and not ids.get("place_id"):
data = self._get(
PLACE_ENDPOINT,
{"token": token, "gl": gl, "hl": hl, "google_domain": domain, "data_cid": ids.get("data_cid", "")},
)
if "_error" in data:
return self._as_error(data)
credits += CREDITS_PER_CALL
fallback_info = data.get("place_results") or {}
for key in ("place_id", "data_id"):
if fallback_info.get(key):
ids[key] = str(fallback_info[key])
if not ids.get("data_id") and not ids.get("place_id"):
return {"status": "error", "errorMessage": "Couldn't find this place's review ID."}
max_reviews = max(1, min(int(max_reviews or 20), MAX_REVIEWS))
base: Dict[str, Any] = {"token": token, "gl": gl, "hl": hl, "google_domain": domain}
if ids.get("data_id"):
base["data_id"] = ids["data_id"]
else:
base["place_id"] = ids["place_id"]
sort_key = (sort_by or "").strip()
sort_value = SORT_OPTIONS.get(sort_key.lower(), sort_key if sort_key in SORT_OPTIONS.values() else None)
if sort_value:
base["sort_by"] = sort_value
if keyword and keyword.strip():
base["query"] = keyword.strip()
collected: List[Dict[str, Any]] = []
place_info: Optional[Dict[str, Any]] = None
topics: List[Dict[str, Any]] = []
next_token: Optional[str] = None
pages_fetched = 0
while len(collected) < max_reviews and pages_fetched < 12:
params = dict(base, num=min(REVIEWS_PER_CALL, max_reviews - len(collected)))
if next_token:
params["next_page_token"] = next_token
data = self._get(REVIEWS_ENDPOINT, params)
if "_error" in data:
if pages_fetched == 0:
return self._as_error(data)
warnings.append(f"Stopped after {len(collected)} reviews: {data['_error']}")
break
credits += CREDITS_PER_CALL
pages_fetched += 1
if pages_fetched == 1:
place_info = data.get("place_info")
topics = data.get("topics") or []
batch = data.get("reviews") or []
for review in batch:
if len(collected) >= max_reviews:
break
review = dict(review)
review["position"] = len(collected) + 1
collected.append(review)
next_token = (data.get("pagination") or {}).get("next_page_token")
if not next_token or not batch:
break
if not place_info and fallback_info:
place_info = {
key: fallback_info.get(key) for key in ("title", "address", "rating", "reviews", "type")
}
result: Dict[str, Any] = {
"status": "success",
"place_info": place_info or {},
"topics": topics,
"reviews": collected,
"reviews_fetched": len(collected),
"place_ids": ids,
"pages_fetched": pages_fetched,
"credits_used": credits,
}
if resolved.get("note"):
result["note"] = resolved["note"]
if warnings:
result["warnings"] = warnings
return result
# -------------------------------------------------------------- forward
def forward(
self,
mode: str,
query: str,
country: Optional[str] = None,
pages: Optional[int] = None,
max_reviews: Optional[int] = None,
sort_by: Optional[str] = None,
keyword: Optional[str] = None,
location: Optional[str] = None,
) -> str:
mode_key = (mode or "").strip().lower()
if mode_key in ("search", "find", "businesses"):
result = self.search_places(query, country or "us", pages or 1, location)
elif mode_key in ("place", "details", "place_details"):
result = self.place_details(query, country or "us")
elif mode_key in ("reviews", "review"):
result = self.reviews(query, country or "us", max_reviews or 20, sort_by, keyword)
else:
result = {"status": "error", "errorMessage": "mode must be 'search', 'place' or 'reviews'."}
return json.dumps(result, ensure_ascii=False, indent=2)