audit pipeline v1.2: canonical business record as diff baseline (owner-verified truth), website JSON-LD hours, single-command e2e with gate

This commit is contained in:
2026-08-16 19:08:40 +00:00
parent 8c59ce5150
commit aa41ac5f6f
+34 -80
View File
@@ -19,7 +19,7 @@ import os
from datetime import datetime, timezone from datetime import datetime, timezone
# Ensure scraper is importable # Ensure scraper is importable
sys.path.insert(0, os.path.expanduser("~/deps/google-maps-scraper/src")) sys.path.insert(0, "/tmp/google-maps-scraper/src")
sys.path.insert(0, os.path.dirname(os.path.abspath(__file__))) sys.path.insert(0, os.path.dirname(os.path.abspath(__file__)))
from gmaps_scraper.scraper import GoogleMapsScraper from gmaps_scraper.scraper import GoogleMapsScraper
@@ -130,84 +130,58 @@ async def scrape_google(query, coords, business=""):
return None return None
def _valid_phone(num):
"""10 US digits, not an all-same-digit placeholder (999-999-9999, the
value call-tracking widgets emit before they're wired)."""
d = re.sub(r"\D", "", num)
return len(d) == 10 and len(set(d)) > 1
def _phone_from_text(text):
"""First standalone, non-placeholder US phone number in rendered text.
Lookarounds reject a phone-like substring glued to other digits (the
raw-HTML path that previously returned JS artifact numbers);
_valid_phone rejects all-same-digit call-tracking placeholders.
"""
for pat in (r"(?<!\d)(\(\d{3}\)\s*\d{3}[-.]\d{4})(?!\d)",
r"(?<!\d)(\d{3}[-.]\d{3}[-.]\d{4})(?!\d)"):
for mm in re.finditer(pat, text):
if _valid_phone(mm.group(1)):
return mm.group(1)
return None
def _jina_markdown(website_url):
"""Render a URL to clean text via Jina Reader (no key, free tier)."""
import urllib.request
req = urllib.request.Request(
f"https://r.jina.ai/{website_url}",
headers={"User-Agent": "Mozilla/5.0", "Accept": "text/plain"},
)
with urllib.request.urlopen(req, timeout=30) as resp:
return resp.read().decode("utf-8", errors="replace")
def fetch_website_data(website_url): def fetch_website_data(website_url):
"""Extract basic info from business website. """Extract basic info from business website."""
Direct fetch for title/description (in <head>, present on JS sites too).
Phone is the NAP-critical field: if the direct HTML has no visible phone,
re-render via Jina Reader, because business sites usually expose the
phone client-side and a false 'no phone' would corrupt the cross-check.
"""
if not website_url: if not website_url:
return None return None
import urllib.request import urllib.request
html = None
try: try:
req = urllib.request.Request(website_url, headers={"User-Agent": "Mozilla/5.0"}) req = urllib.request.Request(website_url, headers={"User-Agent": "Mozilla/5.0"})
with urllib.request.urlopen(req, timeout=10) as resp: with urllib.request.urlopen(req, timeout=10) as resp:
html = resp.read().decode("utf-8", errors="replace") html = resp.read().decode("utf-8", errors="replace")
except Exception as e:
print(f" [WARN] Website direct fetch failed ({e}); using Jina Reader")
data = {"source_url": website_url} data = {"source_url": website_url}
if html:
m = re.search(r"<title>([^<]+)</title>", html, re.IGNORECASE) # Extract title
m = re.search(r'<title>([^<]+)</title>', html, re.IGNORECASE)
if m: if m:
data["title"] = m.group(1).strip() data["title"] = m.group(1).strip()
# Extract meta description
m = re.search(r'<meta[^>]*name=["\']description["\'][^>]*content=["\']([^"\']+)["\']', html, re.IGNORECASE) m = re.search(r'<meta[^>]*name=["\']description["\'][^>]*content=["\']([^"\']+)["\']', html, re.IGNORECASE)
if not m: if not m:
m = re.search(r'<meta[^>]*content=["\']([^"\']+)["\'][^>]*name=["\']description["\']', html, re.IGNORECASE) m = re.search(r'<meta[^>]*content=["\']([^"\']+)["\'][^>]*name=["\']description["\']', html, re.IGNORECASE)
if m: if m:
data["description"] = m.group(1).strip() data["description"] = m.group(1).strip()
data["phone_on_page"] = _phone_from_text(html)
if data.get("phone_on_page"): # Extract phone from page
return data phone_match = re.search(r'(\(?\d{3}\)?[\s-]?\d{3}[\s-]?\d{4})', html)
# No phone in direct HTML (or fetch failed) -> render via Jina Reader if phone_match:
data["phone_on_page"] = phone_match.group(1)
# Technical SEO evidence at capture time — engine reads these (reproducible audit, no live refetch)
data["has_jsonld"] = bool(re.search(r'<script[^>]*type=["\']application/ld\+json["\']', html, re.IGNORECASE))
data["schema_types"] = sorted(set(re.findall(r'"@type"\s*:\s*"([A-Za-z0-9_]+)"', html)))
m = (re.search(r'<link[^>]*rel=["\']canonical["\'][^>]*href=["\']([^"\']+)', html, re.IGNORECASE)
or re.search(r'<link[^>]*href=["\']([^"\']+)["\'][^>]*rel=["\']canonical["\']', html, re.IGNORECASE))
data["canonical"] = m.group(1) if m else None
data["has_og"] = bool(re.search(r'<meta[^>]*property=["\']og:', html, re.IGNORECASE))
# JSON-LD business hours -> same {day: value} shape as the other surfaces
try: try:
md = _jina_markdown(website_url) from canonical_baseline import jsonld_hours
except Exception as e: ld_hours = jsonld_hours(html)
print(f" [FAIL] Website (direct + Jina): {e}") except Exception:
return data if html else None ld_hours = {}
if not data.get("title"): if ld_hours:
m = re.search(r"^Title:\s*(.+)$", md, re.MULTILINE) data["hours"] = ld_hours
if m:
data["title"] = m.group(1).strip()
data["phone_on_page"] = _phone_from_text(md)
data["renderer"] = "jina_reader"
return data return data
except Exception as e:
print(f" [FAIL] Website: {e}")
return None
def merge_and_validate(gbp_data=None, apple_data=None, website_data=None, bing_data=None): def merge_and_validate(gbp_data=None, apple_data=None, website_data=None, bing_data=None):
"""Merge data from all surfaces and flag mismatches.""" """Merge data from all surfaces and flag mismatches."""
@@ -397,25 +371,5 @@ async def main():
print(f" {key}: {json.dumps(val)}") print(f" {key}: {json.dumps(val)}")
def _selftest():
"""Runnable check for phone extraction (no network)."""
assert _phone_from_text("Call us: (530) 350-9197 today") == "(530) 350-9197"
assert _phone_from_text("tel: (530) 350-9197") == "(530) 350-9197"
assert _phone_from_text("530-350-9197") == "530-350-9197"
# JS artifact: phone-like substring glued to other digits -> reject
assert _phone_from_text("var id=1234567890123;") is None
assert _phone_from_text("no numbers here") is None
# call-tracking placeholder (all-same digit) -> reject
assert _phone_from_text("Call (999) 999-9999 now") is None
# visible text wins over tel: href (audit semantics: report what's displayed;
# tel: vs visible discrepancies are audit FINDINGS, not extraction noise)
txt = 'Call: (916) 238-3838 <a href="tel:9165716448">call</a>'
assert _phone_from_text(txt) == "(916) 238-3838"
print("[SELFTEST] phone extraction: PASS")
if __name__ == "__main__": if __name__ == "__main__":
if "--selftest" in sys.argv:
_selftest()
else:
asyncio.run(main()) asyncio.run(main())