fix(impressum): Findings aus 12 §5-TMG-Pattern-MCs statt verunreinigtem DB-Set

Der Agent lieferte "alles gruen": _load_controls gab auf macmini nur 3 von 75
doc_type='impressum'-MCs zurueck (Sidecar mc_classification.db hat nur 4/75 als
text-matchbar klassifiziert). Tiefere Ursache: die 75 doc_type='impressum'-MCs
sind fehl-klassifiziert (60/75 canonical_scope='other'; Prefixes TRD/SEC/GOV =
Geschaeftsbriefe/Marktplatz/Bestellung, NICHT §5 TMG Website-Impressum).

Fix: Der Impressum-Agent erzeugt Findings jetzt aus seinen 12 autoritativen
§5-TMG/DDG-Pattern-MCs (mcs.py) statt aus dem verunreinigten DB-Set —
deterministisch, scope-aware, field_id = semantisches Feld. Semantic-Validator-
Demote + Massnahmen + Rollup bleiben. Die 5-Impressum-GT-Tests laufen jetzt
echt durch: 0 Falsch-Positive.

DB-Master-Controls fuer Impressum deaktiviert bis zum MC-Re-Filtering (separate
Aufgabe: die doc_type-Klassifizierung der Vorgaenger-Session muss bereinigt
werden).

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

backend-compliance/compliance/services/specialist_agents/impressum/agent.py

@@ -39,8 +39,7 @@ from .._base import (
 from .._pattern_library import record as record_pattern
 from .._rollup import rollup
 from .._semantic_validator import build_rename_action, validate_present
-from .mcs import MC_IDS, MCS, detect_automotive
-from .v3_engine import run_v3_pipeline
+from .mcs import MC_IDS, MCS, detect_automotive, scope_matches
 
 logger = logging.getLogger(__name__)
 
@@ -107,73 +106,53 @@ class ImpressumAgent(BaseSpecialistAgent):
                 notes="Impressum-Text zu kurz oder leer.",
             )
 
-        # ── Layer 0 + 1 + 2 (Boost + Keyword + Embedding) ──────────
-        results, telemetry = await run_v3_pipeline(text, scope)
-        notes_parts.append(
-            f"v3-pipeline: {telemetry.get('total_mcs', 0)} DB-MCs · "
-            f"{telemetry.get('layer_0_field_hits', 0)} Pattern-Boosts · "
-            f"{telemetry.get('layer_0_boost_overrides', 0)} Boost-Overrides"
-        )
-        sec_drop = telemetry.get("sector_dropped", 0)
-        off_drop = telemetry.get("offtopic_dropped", 0)
-        if sec_drop or off_drop:
-            notes_parts.append(
-                f"Scope-Filter: {sec_drop} Branchen-MCs + "
-                f"{off_drop} themenfremde MCs entfernt"
-            )
-
-        # DB-MCs → Findings + Coverage
-        seen_db_mcs: set[str] = set()
-        for r in results:
-            mc_id = r.get("control_id") or ""
-            if not mc_id or mc_id in seen_db_mcs:
+        # ── Findings aus den 12 autoritativen §5-TMG/DDG-MCs (mcs.py) ──
+        # Die DB-Menge doc_type='impressum' ist verunreinigt (TRD/SEC/GOV-
+        # Controls statt §5 TMG, fehl-klassifiziert von einer Vorgaenger-
+        # Session), daher sind bis zum MC-Re-Filtering die 12 praezisen
+        # Pattern-MCs die Findings-Quelle. field_id = semantisches Feld
+        # (passt zum Semantic-Validator + den GT-Tests).
+        is_auto = "automotive" in scope
+        for mc in MCS:
+            if not scope_matches(mc, scope, is_auto):
+                coverage.append(McCoverage(
+                    mc_id=mc.mc_id, status="na",
+                    reason="nicht im Business-Scope",
+                ))
                 continue
-            seen_db_mcs.add(mc_id)
-            passed = bool(r.get("passed"))
-            sev = _SEV_TO_ENUM.get(
-                (r.get("severity") or "MEDIUM").upper(), Severity.MEDIUM,
-            )
-            coverage.append(McCoverage(
-                mc_id=mc_id,
-                status="ok" if passed else sev.value.lower(),
-                reason=str(r.get("matched_text") or r.get("hint") or "")[:120],
-            ))
-            if passed:
+            if any(p.search(text) for p in mc.patterns):
+                coverage.append(McCoverage(
+                    mc_id=mc.mc_id, status="ok", reason="Pattern-Treffer",
+                ))
                 continue
-            label = r.get("label") or r.get("hint") or ""
-            norm_str = str(r.get("regulation") or "")
-            if r.get("article"):
-                norm_str = (norm_str + f" Art. {r.get('article')}").strip()
+            sev = _SEV_TO_ENUM.get(mc.severity_if_missing, Severity.MEDIUM)
             findings.append(Finding(
-                check_id=f"DBMC-{mc_id}",
+                check_id=f"IMP-{mc.field_id}",
                 agent=self.agent_id,
                 agent_version=self.agent_version,
-                field_id=mc_id,
+                field_id=mc.field_id,
                 severity=sev,
-                severity_reason="db_mc_failed",
-                title=str(label)[:200] or f"DB-MC {mc_id} nicht erfüllt",
-                norm=norm_str,
+                severity_reason="pflichtangabe_missing",
+                title=f"Pflichtangabe fehlt: {mc.label}",
+                norm=mc.norm,
                 evidence="",
-                action=_build_measure(str(label), norm_str)[:400],
+                action=_build_measure(mc.label, mc.norm),
                 confidence=0.9,
                 sources=[EvidenceSource(
-                    source_type=SourceType.MC,
-                    source_id=mc_id,
-                    detail=str(r.get("source") or "keyword_match")[:120],
+                    source_type=SourceType.REGEX,
+                    source_id=mc.mc_id,
+                    detail="kein Pattern-Treffer im Text",
                     confidence=0.9,
                 )],
             ))
-
-        # Layer 0: eigene Pattern-IDs immer mit ins coverage (für UI)
-        boost_ids = set(telemetry.get("layer_0_field_ids") or [])
-        for mc in MCS:
-            cov_status = "ok" if mc.field_id in boost_ids else "na"
-            cov_reason = ("regex-boost hit"
-                           if mc.field_id in boost_ids
-                           else "kein Pattern-Treffer (kein Veto)")
             coverage.append(McCoverage(
-                mc_id=mc.mc_id, status=cov_status, reason=cov_reason,
+                mc_id=mc.mc_id, status=sev.value.lower(),
+                reason="kein Pattern-Treffer",
             ))
+        notes_parts.append(
+            f"{len(MCS)} §5-TMG-MCs geprüft · "
+            f"{len(findings)} Pflichtangabe(n) offen"
+        )
 
         # ── Layer 3: Semantic-Validator nur für HIGH/MEDIUM-Fails ──
         await self._semantic_demote(text, findings, coverage)
@@ -199,7 +178,7 @@ class ImpressumAgent(BaseSpecialistAgent):
         candidates = [
             f for f in findings
             if f.severity in (Severity.HIGH.value, Severity.MEDIUM.value)
-            and f.severity_reason == "db_mc_failed"
+            and f.severity_reason == "pflichtangabe_missing"
         ]
         if not candidates:
             return
@@ -232,9 +211,11 @@ class ImpressumAgent(BaseSpecialistAgent):
                 detail=f"LLM-confirmed: '{label_used}'",
                 confidence=conf,
             ))
-            # Coverage update + auto-learning
+            # Coverage update + auto-learning (mc_id steckt in der Quelle)
+            mc_id_for_cov = (finding.sources[0].source_id
+                             if finding.sources else "")
             for c in coverage:
-                if c.mc_id and c.mc_id == f"DBMC-{finding.field_id}":
+                if c.mc_id and c.mc_id == mc_id_for_cov:
                     c.status = "low"
                     c.reason = f"label_mismatch: '{label_used}'"
             try:

backend-compliance/tests/test_impressum_v3.py

@@ -18,6 +18,7 @@ from compliance.services.specialist_agents import (
 from compliance.services.specialist_agents.impressum.agent import (
     _build_measure,
 )
+from compliance.services.specialist_agents.impressum.mcs import MCS
 from compliance.services.specialist_agents.impressum.regex_boost import (
     BOOST_KEYWORDS,
     boost_matches_db_mc,
@@ -108,80 +109,49 @@ def test_boost_keywords_cover_all_field_ids():
 
 
 @pytest.fixture
-def mock_v3(monkeypatch):
-    """Mockt run_v3_pipeline mit deterministischen Fake-Results."""
-    async def _fake_pipeline(text, scope, db_url=""):
-        results = [
-            {"control_id": "AUTH-1954-A04",
-             "passed": True,
-             "label": "Anbieterkennzeichnung dokumentiert",
-             "severity": "HIGH",
-             "regulation": "TMG",
-             "article": "§ 5",
-             "hint": "",
-             "matched_text": "Tesla Germany GmbH",
-             "source": "keyword_match"},
-            {"control_id": "DATA-2786-A04",
-             "passed": False,
-             "label": "Freiwilligkeit der TDDDG-Einwilligungen",
-             "severity": "MEDIUM",
-             "regulation": "TDDDG",
-             "article": "§ 25",
-             "hint": "Bitte Freiwilligkeit dokumentieren",
-             "matched_text": "",
-             "source": ""},
-        ]
-        telemetry = {
-            "layer_0_field_hits": 5,
-            "layer_0_field_ids": ["kontakt_email", "kontakt_telefon",
-                                    "handelsregister", "ust_id",
-                                    "vertretungsberechtigte"],
-            "layer_1_pass": 1,
-            "layer_1_fail": 1,
-            "layer_0_boost_overrides": 0,
-            "total_mcs": 2,
-        }
-        return results, telemetry
-    monkeypatch.setattr(
-        "compliance.services.specialist_agents.impressum.agent.run_v3_pipeline",
-        _fake_pipeline,
-    )
-    async def _no_validator(*a, **kw): return {}
+def no_llm(monkeypatch):
+    """Deaktiviert den LLM-Semantic-Validator — der Agent prueft die 12
+    mcs.py-Pattern-MCs deterministisch direkt am Text."""
+    async def _no_validator(*a, **kw):
+        return {}
     monkeypatch.setattr(
         "compliance.services.specialist_agents.impressum.agent.validate_present",
         _no_validator,
     )
 
 
-def test_agent_uses_db_mcs(mock_v3):
+def test_agent_emits_pflichtangabe_findings(no_llm):
     agent = ImpressumAgent()
     out = _run(agent.evaluate(AgentInput(doc_type="impressum",
                                             text=TESLA_TEXT)))
-    db_mc_findings = [f for f in out.findings
-                      if f.check_id.startswith("DBMC-")]
-    assert len(db_mc_findings) == 1
-    assert db_mc_findings[0].check_id == "DBMC-DATA-2786-A04"
-    assert db_mc_findings[0].severity == Severity.MEDIUM.value
-    assert "TDDDG" in db_mc_findings[0].norm
+    fids = {f.field_id for f in out.findings}
+    # Tesla nennt 'Management' (englisch) → deutsches GF-Label fehlt
+    assert "vertretungsberechtigte_label_korrekt" in fids
+    f = next(f for f in out.findings
+             if f.field_id == "vertretungsberechtigte_label_korrekt")
+    assert f.severity == Severity.MEDIUM.value
+    assert f.check_id == "IMP-vertretungsberechtigte_label_korrekt"
+    assert f.severity_reason == "pflichtangabe_missing"
+    # Vorhandene Pflichtangaben erzeugen KEIN Finding
+    assert "kontakt_email" not in fids
+    assert "handelsregister" not in fids
 
 
-def test_agent_emits_boost_coverage(mock_v3):
+def test_agent_coverage_has_all_12(no_llm):
     agent = ImpressumAgent()
     out = _run(agent.evaluate(AgentInput(doc_type="impressum",
                                             text=TESLA_TEXT)))
-    # 2 DB-MCs + 12 Pattern-Boost-Slots = 14 coverage entries
-    assert out.mc_total >= 14
-    boost_ok = [c for c in out.mc_coverage
-                if c.mc_id.startswith("IMP-MC-") and c.status == "ok"]
-    assert len(boost_ok) == 5  # 5 boost_ids im fake
+    assert out.mc_total == len(MCS)  # je MC genau 1 Coverage-Eintrag
+    ok = [c for c in out.mc_coverage if c.status == "ok"]
+    # name, email, telefon, HR, USt, vertretungsberechtigte = 6 vorhanden
+    assert len(ok) == 6
 
 
-def test_agent_notes_telemetry(mock_v3):
+def test_agent_notes(no_llm):
     agent = ImpressumAgent()
     out = _run(agent.evaluate(AgentInput(doc_type="impressum",
                                             text=TESLA_TEXT)))
-    assert "v3-pipeline" in out.notes
-    assert "Pattern-Boosts" in out.notes
+    assert "§5-TMG-MCs geprüft" in out.notes
 
 
 def test_short_text_skipped():