add public contact extraction

2026-09-03 11:15:54 +02:00
parent fb89a28f2c
commit 89eb7e07e6
14 changed files with 402 additions and 7 deletions
@@ -0,0 +1,139 @@
+"""Conservative extraction of public business contact signals from approved HTML.
+
+This module never fetches URLs. Callers must provide HTML obtained from the approved
+website scanner and a source URL already associated with the business.
+"""
+from __future__ import annotations
+
+import html as html_lib
+import re
+from html.parser import HTMLParser
+from urllib.parse import unquote, urlparse
+
+MAX_HTML_BYTES = 512 * 1024
+MAX_RESULTS = 100
+EMAIL_RE = re.compile(r"(?<![\w.+-])[A-Z0-9.!#$%&'*+/=?^_`{|}~-]+@[A-Z0-9](?:[A-Z0-9-]{0,61}[A-Z0-9])?(?:\.[A-Z0-9](?:[A-Z0-9-]{0,61}[A-Z0-9])?)+", re.I)
+PHONE_RE = re.compile(r"(?<!\w)(\+?\d[\d ()().-]{6,}\d)(?!\w)")
+SOCIAL_HOSTS = {"facebook.com", "instagram.com", "linkedin.com", "twitter.com", "x.com", "youtube.com", "tiktok.com"}
+FREE_MAIL = {"gmail.com", "googlemail.com", "yahoo.com", "yahoo.co.uk", "hotmail.com", "outlook.com", "live.com", "icloud.com", "aol.com", "proton.me", "protonmail.com", "mail.com"}
+ROLE_NAMES = {"info", "hello", "contact", "sales", "support", "admin", "office", "enquiries", "inquiries", "accounts", "billing", "careers", "hr", "help", "marketing", "bookings", "reception", "service", "customerservice"}
+EXAMPLE_DOMAINS = {"example.com", "example.org", "example.net", "example.test", "invalid", "localhost"}
+
+
+def _clean(value: str) -> str:
+    return re.sub(r"\s+", " ", html_lib.unescape(value or "")).strip()
+
+
+def _decode_obfuscation(value: str) -> str:
+    value = html_lib.unescape(unquote(value or ""))
+    value = re.sub(r"\s*(?:\[|\(|\{|\s)at(?:\]|\)|\}|\s)\s*", "@", value, flags=re.I)
+    value = re.sub(r"\s*(?:\[|\(|\{|\s)dot(?:\]|\)|\}|\s)*", ".", value, flags=re.I)
+    return value
+
+
+def valid_email(value: str) -> bool:
+    value = value.strip().lower()
+    if len(value) > 254 or value.count("@") != 1 or value.split("@", 1)[1] in EXAMPLE_DOMAINS:
+        return False
+    return bool(EMAIL_RE.fullmatch(value)) and not any(x in value for x in ("password", "token", "secret", "credential", "apikey"))
+
+
+def normalize_phone(value: str) -> str:
+    value = value.strip()
+    digits = re.sub(r"\D", "", value)
+    if value.startswith("+") and digits:
+        return "+" + digits
+    return digits
+
+
+class _ContactParser(HTMLParser):
+    def __init__(self):
+        super().__init__(convert_charrefs=True)
+        self.text: list[str] = []
+        self.links: list[tuple[str, str]] = []
+        self.form_fields: list[str] = []
+        self._ignore = 0
+        self._anchor = ""
+        self._in_form = False
+
+    def handle_starttag(self, tag, attrs):
+        attrs = {str(k).lower(): str(v or "") for k, v in attrs}
+        tag = tag.lower()
+        if tag in {"script", "style", "noscript", "template", "svg"}:
+            self._ignore += 1
+        if tag == "a": self._anchor = attrs.get("href", "")
+        if tag == "form": self._in_form = True
+        if self._in_form and tag in {"input", "textarea", "select"}:
+            name = attrs.get("name", "") or attrs.get("id", "")
+            typ = attrs.get("type", "")
+            self.form_fields.append(_clean(" ".join((name, typ, attrs.get("placeholder", "")))))
+
+    def handle_endtag(self, tag):
+        tag = tag.lower()
+        if tag in {"script", "style", "noscript", "template", "svg"} and self._ignore: self._ignore -= 1
+        if tag == "a": self._anchor = ""
+        if tag == "form": self._in_form = False
+
+    def handle_data(self, data):
+        if self._ignore: return
+        if data.strip(): self.text.append(data)
+        if self._anchor: self.links.append((self._anchor, data))
+
+
+def _record(kind, value, label, source_url, confidence, classification="unknown", *, suppressed=False, provenance="visible_text"):
+    return {"kind": kind, "value": value, "label": label[:200], "classification": classification,
+            "confidence": round(max(0.0, min(1.0, confidence)), 2), "source_url": source_url,
+            "public_business": True, "mx_status": "unknown", "suppressed": bool(suppressed),
+            "do_not_contact": bool(suppressed), "provenance": provenance}
+
+
+def extract_contacts(source_html: str, source_url: str, *, suppressions=None, max_results=MAX_RESULTS) -> list[dict]:
+    if not isinstance(source_html, str) or len(source_html.encode("utf-8")) > MAX_HTML_BYTES:
+        raise ValueError("html_too_large")
+    try: max_results = int(max_results)
+    except (ValueError, TypeError): raise ValueError("invalid_limits")
+    if max_results < 1 or max_results > MAX_RESULTS: raise ValueError("invalid_limits")
+    parser = _ContactParser(); parser.feed(source_html)
+    visible = _decode_obfuscation(_clean(" ".join(parser.text)))
+    suppression = {(str(x.get("kind", "")), str(x.get("value", "")).lower()) for x in (suppressions or [])}
+    out, seen = [], set()
+    def add(kind, value, label, confidence, classification="unknown", provenance="visible_text"):
+        value = value.strip().lower() if kind == "email" else value.strip()
+        if kind == "email":
+            if not valid_email(value): return
+            local, domain = value.rsplit("@", 1)
+            classification = "free_mail" if domain in FREE_MAIL else ("role" if local in ROLE_NAMES else "named")
+            key = (kind, value); blocked = ("email", value) in suppression or ("domain", domain) in suppression
+        else:
+            if kind in {"phone", "whatsapp"}: value = normalize_phone(value)
+            if len(re.sub(r"\D", "", value)) < 7: return
+            key = (kind, value); blocked = (kind, value.lower()) in suppression
+        if key in seen or len(out) >= max_results: return
+        seen.add(key); out.append(_record(kind, value, label or kind.title(), source_url, confidence, classification, suppressed=blocked, provenance=provenance))
+    for href, anchor_text in parser.links:
+        raw = _decode_obfuscation(href)
+        if raw.lower().startswith("mailto:"):
+            address = raw[7:].split("?", 1)[0]
+            add("email", address, _clean(anchor_text), 0.98, provenance="mailto")
+        elif raw.lower().startswith("tel:"):
+            add("phone", raw[4:].split("?", 1)[0], _clean(anchor_text), 0.98, provenance="tel")
+        else:
+            parsed = urlparse(raw if "://" in raw else "https://" + raw)
+            host = (parsed.hostname or "").lower().removeprefix("www.")
+            if host == "wa.me" or "whatsapp" in host:
+                number = re.sub(r"\D", "", parsed.path)
+                if number: add("whatsapp", "+" + number, _clean(anchor_text) or "WhatsApp", 0.98, provenance="whatsapp_link")
+            elif any(host == d or host.endswith("." + d) for d in SOCIAL_HOSTS):
+                add("social", raw, _clean(anchor_text) or host, 0.95, provenance="social_link")
+    for match in EMAIL_RE.finditer(visible):
+        context = visible[max(0, match.start() - 40):match.start()]
+        if re.search(r"(?:password|passwd|token|secret|credential|api[_ -]?key|authorization)\s*[:=]?\s*$", context, re.I):
+            continue
+        add("email", match.group(0), "Email", 0.88)
+    for value in PHONE_RE.findall(visible): add("phone", value, "Phone", 0.82)
+    # A form is provenance, not proof of a destination address.
+    if parser.form_fields:
+        for field in parser.form_fields[:5]:
+            if re.search(r"email|contact|phone|whatsapp", field, re.I):
+                add("form", field, "Contact form", 0.7, provenance="form_field")
+    return out