breakpilot-compliance/backend-compliance/compliance/services/impressum_multi_entity_check.py

"""B9 — Multi-Entity-Impressum-Check.

Findings, wenn ein Impressum mehrere Entitäten (mehrere GmbH/AG/UG)
nennt, aber Pflichtangaben nur bei einer davon vollständig sind.

Konkreter Elli-Pattern (GT IMPRESSUM-001):
  - Entity 1: "Elli Mobility GmbH ... USt-IdNr DE814424009 ..."
  - Entity 2: "VW Group Charging GmbH ... [keine USt-IdNr] ..."
  → USt-IdNr fehlt bei Entity 2.

Heuristik:
  1. Entitäten erkennen: jede Match auf "<Name> (GmbH|AG|UG|KG|SE)" als
     Entity-Boundary; Text-Slice von dort bis zur nächsten Entity.
  2. Pro Entity prüfen: USt-IdNr, Handelsregister, Vertretungsberechtigte.
  3. Wenn Entity N ein Feld nennt, das Entity M nicht hat → MEDIUM.
"""

from __future__ import annotations

import logging
import re

logger = logging.getLogger(__name__)

_ENTITY_PAT = re.compile(
    r"([A-ZÄÖÜ][\w\-\& ]{1,50}?\s+(?:GmbH|AG|UG|KG|SE|"
    r"e\.V\.|GbR|OHG|Limited|Ltd|LLC))",
    re.IGNORECASE,
)

_NAME_NOISE_PAT = re.compile(
    r"^(?:Impressum|Anbieter|Anbieterkennzeichnung|Diensteanbieter|"
    r"Verantwortlich(?:er)?|Kontakt|Adresse|@\S+|.+@.+)\s*[:|\-]?\s*",
    re.IGNORECASE,
)

# HRB-/HRA-Eintrag pro Entity. Dies ist der stärkste Anker: jede
# juristische Person muss in Deutschland einen eindeutigen
# HRB/HRA-Eintrag haben. Mehrere HRB-Vorkommen = mehrere Entities.
_HRB_PAT = re.compile(
    r"HR[BA]\s*(?:Nr\.?|Nummer)?\s*\d+(?:\s*[A-Z])?",
    re.IGNORECASE,
)

# Worte, die NIE ein gültiger Firmenname sind. Filtern False-Positives
# wie "Programmierung der Webseite Elli Mobility GmbH" oder
# "Umsatzsteueridentifikationsnummer der Elli Mobility GmbH".
_NAME_BLOCKLIST = (
    "programmierung", "webseite", "umsatzsteueridentifik",
    "schlichtungsstelle", "auftragsverarbeitung", "haftung",
    "verantwortlich", "diensteanbieter", "geschäftsführer",
    "geschaeftsfuehrer", "vorstand", "gesellschaftsregister",
    "registergericht", "registriert",
)

_USTID_PAT = re.compile(
    # Form A: Abkürzung mit Separator (USt-IdNr.: DE…)
    r"(?:"
    r"\b(?:USt-?Id(?:Nr)?\.?|VAT(?:-?Id)?)\s*[:.\s]\s*"
    r"(DE\d{8,10}|[A-Z]{2}\d{6,12})"
    r"|"
    # Form B: Vollform mit Bridge (Umsatzsteueridentifikationsnummer
    # der Elli Mobility GmbH ist DE…). Max 80 Zeichen Bridge,
    # \n erlaubt (Cap schützt vor Cross-Paragraph-Drift).
    r"\bUmsatzsteuer[\s-]?Id(?:entifikationsnummer)?\b[\s\S]{0,80}?"
    r"(DE\d{8,10}|[A-Z]{2}\d{6,12})"
    r")",
    re.IGNORECASE,
)
_HR_PAT = re.compile(r"\b(?:HR[BA]|Handelsregister|Registergericht)"
                     r"\s*[:.\s]*([\w\s\d\-/]{4,80})", re.IGNORECASE)
_GF_PAT = re.compile(r"(?:Geschäftsführer|Vertretungsberechtigt|"
                     r"vertreten\s+durch)\s*[:.\s]+", re.IGNORECASE)


_LEADING_NOISE_RE = re.compile(
    r"^(?:eco|com|de|net|"  # email/domain TLD-Artefakte
    r"Die|Der|Das|"          # deutsche Artikel
    r"www\.[\w\.-]+|"         # URL-Reste
    r"@[\w\.-]+)\s+",
    re.IGNORECASE,
)


def _clean_entity_name(raw: str) -> str:
    """Strip leading header noise, leading artifacts + collapse whitespace."""
    name = raw.strip()
    # If the match spans multiple lines, keep only the last line.
    if "\n" in name:
        name = name.rsplit("\n", 1)[-1].strip()
    name = _NAME_NOISE_PAT.sub("", name).strip()
    # Strip leading email-TLD-artifacts ("eco "), German articles
    # ("Die ", "Der "), URL-fragments etc. — iterate until stable.
    for _ in range(3):
        new = _LEADING_NOISE_RE.sub("", name)
        if new == name:
            break
        name = new
    # Strip leading lowercase prose words (English connectors leaking
    # into the match: "If ...", "by ...", "according to ...").
    # Drop tokens until we hit the first Capitalized one.
    tokens = name.split()
    while tokens and not tokens[0][:1].isupper():
        tokens = tokens[1:]
    name = " ".join(tokens)
    return re.sub(r"\s+", " ", name).strip()


def _dedup_substring(slices: list[tuple[str, str]]) -> list[tuple[str, str]]:
    """Collapse entities whose names are substrings of each other.

    'mfi Immobilien Marketing GmbH' and 'Marketing GmbH' both refer to
    the same legal person — keep only the longest unique name.
    """
    sorted_by_len = sorted(slices, key=lambda x: -len(x[0]))
    kept: list[tuple[str, str]] = []
    kept_names_lc: list[str] = []
    for name, slc in sorted_by_len:
        nl = name.lower()
        if any(nl in k or k in nl for k in kept_names_lc):
            continue
        kept.append((name, slc))
        kept_names_lc.append(nl)
    return kept


def _name_is_blocked(name: str) -> bool:
    nl = name.lower()
    if any(b in nl for b in _NAME_BLOCKLIST):
        return True
    # Minimum-name-quality: must have ≥ 2 words, at least one ≥ 4 chars
    # before the legal-form suffix. Filters out "Se", "As a se" frags.
    parts = name.strip().split()
    if len(parts) < 2:
        return True
    # Strip legal-form from the end if present
    legal_suffixes = {
        "gmbh", "ag", "ug", "kg", "se", "e.v.", "gbr", "ohg",
        "limited", "ltd", "llc",
    }
    if parts[-1].lower() in legal_suffixes:
        non_suffix = parts[:-1]
    else:
        non_suffix = parts
    if not non_suffix or len(non_suffix) < 1:
        return True
    # At least one company-name token ≥ 4 chars and capitalized
    if not any(p[0].isupper() and len(p) >= 4 for p in non_suffix):
        return True
    return False


def _slice_entities(text: str) -> list[tuple[str, str]]:
    """Return [(entity_name, text_slice)] for each detected entity.

    Anker-Strategie: jeder eigenständige HRB/HRA-Eintrag markiert eine
    Entity. Pro HRB-Block laufen wir RÜCKWÄRTS bis zum vorherigen
    Block-Ende und nehmen den letzten Legal-Form-Match (GmbH/AG/...)
    in diesem Fenster als Firmennamen.

    Falls keine HRBs vorhanden sind (z.B. e.V. ohne HR-Pflicht),
    fallen wir auf den alten Legal-Form-Anker zurück.
    """
    hrb_matches = list(_HRB_PAT.finditer(text))
    if len(hrb_matches) >= 2:
        slices: list[tuple[str, str]] = []
        seen_names: set[str] = set()
        for i, h in enumerate(hrb_matches):
            # Window: from end of previous HRB (or start) to this HRB.
            win_start = hrb_matches[i - 1].end() if i > 0 else 0
            win_end = h.start()
            window = text[win_start:win_end]
            # Find the LAST legal-form-match in the window — that's
            # the entity-name closest to this HRB-marker.
            legal_matches = list(_ENTITY_PAT.finditer(window))
            if not legal_matches:
                continue
            name_raw = legal_matches[-1].group(1)
            name = _clean_entity_name(name_raw)
            if not name or _name_is_blocked(name):
                continue
            if name in seen_names:
                continue
            seen_names.add(name)
            # Block-Slice: from the matched name to the END of this
            # HRB-block (next HRB-start or EOF).
            slice_start = win_start + legal_matches[-1].start()
            slice_end = (hrb_matches[i + 1].start()
                         if i + 1 < len(hrb_matches) else len(text))
            slices.append((name, text[slice_start:slice_end]))
        slices = _dedup_substring(slices)
        if len(slices) >= 2:
            return slices

    # Fallback (no HRBs / nur einer / parsing scheiterte):
    # alter Legal-Form-Anker mit Blocklist-Filter + dedup.
    matches = list(_ENTITY_PAT.finditer(text))
    if len(matches) < 2:
        return []
    slices = []
    seen_names = set()
    for i, m in enumerate(matches):
        name = _clean_entity_name(m.group(1))
        if not name or _name_is_blocked(name) or name in seen_names:
            continue
        seen_names.add(name)
        start = m.start()
        end = matches[i + 1].start() if i + 1 < len(matches) else len(text)
        slices.append((name, text[start:end]))
    slices = _dedup_substring(slices)
    return slices if len(slices) >= 2 else []


def check_multi_entity_impressum(state: dict) -> list[dict]:
    doc_texts = state.get("doc_texts") or {}
    imp = doc_texts.get("impressum") or ""
    if not imp:
        return []
    slices = _slice_entities(imp)
    if not slices:
        return []
    # Compute features per entity
    features = []
    for name, slc in slices:
        features.append({
            "name": name,
            "ust_id": bool(_USTID_PAT.search(slc)),
            "hr": bool(_HR_PAT.search(slc)),
            "gf": bool(_GF_PAT.search(slc)),
        })
    # If ALL share the same flags → no inconsistency
    findings: list[dict] = []
    for field, label in (
        ("ust_id", "USt-IdNr."),
        ("hr", "Handelsregister-Eintrag"),
        ("gf", "Vertretungsberechtigte"),
    ):
        present = [f for f in features if f[field]]
        missing = [f for f in features if not f[field]]
        if present and missing and len(present) >= 1:
            findings.append({
                "check_id": f"IMPRESSUM-MULTI-{field.upper()}",
                "severity": "MEDIUM",
                "severity_reason": "incomplete",
                "title": (
                    f"{label} fehlt bei "
                    f"{len(missing)} von {len(features)} Entitäten"
                ),
                "norm": "§ 5 Abs. 1 TMG (Pflichtangabe pro Diensteanbieter)",
                "entities_present": [f["name"] for f in present],
                "entities_missing": [f["name"] for f in missing],
                "action": (
                    f"{label} im Impressum für "
                    f"{', '.join(f['name'] for f in missing)} ergänzen. "
                    "Pflichtangabe ist pro Diensteanbieter zu erfüllen, "
                    "nicht 'eine reicht für alle'."
                ),
            })
    if findings:
        logger.info("B9 multi-entity impressum: %d findings", len(findings))
    return findings