fix(source_health): paywall-Strategie nicht ueber removepaywall fuer Feed-URL

removepaywall.com liefert HTML (Article-Renderer), nicht XML - der
Feed-Validity-Check schlug daher fehl mit "Kein gueltiger RSS/Atom-Feed".

Korrektur:
- paywall: Feed-URL direkt mit Browser-UA laden (kein URL-Rewrite).
- Bei paywall + 4xx: status=warning (erwartbar), Feed-Validity skippen.
- removepaywall.com bleibt im Researcher-Prompt fuer Article-Inhalte
  (das ist der korrekte Use-Case).

RELEASES.json

@@ -1,4 +1,13 @@
 [
+  {
+    "version": "2026-05-03T15:21Z",
+    "date": "2026-05-03",
+    "title": "Übersichtlichere Navigation in der Seitenleiste",
+    "items": [
+      "Schaltflächen in der Seitenleiste haben jetzt klarere Icons und kürzere Beschriftungen",
+      "Der Feedback-Button zeigt nun ein Brief-Symbol für bessere Erkennbarkeit"
+    ]
+  },
   {
     "version": "2026-04-30T23:12Z",
     "date": "2026-04-30",

scripts/migrate_sources_classification.py

@@ -0,0 +1,64 @@
+"""Einmalige LLM-Klassifikation aller noch unklassifizierten Quellen.
+
+Verwendung:
+    python3 scripts/migrate_sources_classification.py --limit 50
+    python3 scripts/migrate_sources_classification.py --limit 500     # Alle
+    python3 scripts/migrate_sources_classification.py --recheck-pending  # bereits Pending neu
+
+Schreibt Vorschlaege in proposed_*-Spalten. Approval erfolgt anschliessend
+ueber das Verwaltungs-UI / API (POST /api/sources/{id}/classification/approve).
+"""
+import argparse
+import asyncio
+import logging
+import sys
+from pathlib import Path
+
+# src/ in PYTHONPATH aufnehmen, wenn Skript direkt aufgerufen wird
+HERE = Path(__file__).resolve().parent
+SRC = HERE.parent / "src"
+if str(SRC) not in sys.path:
+    sys.path.insert(0, str(SRC))
+
+from database import get_db  # noqa: E402
+from services.source_classifier import bulk_classify  # noqa: E402
+
+logging.basicConfig(
+    level=logging.INFO,
+    format="%(asctime)s [%(name)s] %(levelname)s: %(message)s",
+)
+logger = logging.getLogger("migrate_sources")
+
+
+async def main():
+    parser = argparse.ArgumentParser(description="LLM-Klassifikation aller Quellen.")
+    parser.add_argument("--limit", type=int, default=50, help="Max. Quellen pro Lauf")
+    parser.add_argument(
+        "--recheck-pending",
+        action="store_true",
+        help="Auch Quellen mit classification_source='llm_pending' neu klassifizieren",
+    )
+    args = parser.parse_args()
+
+    db = await get_db()
+    try:
+        result = await bulk_classify(
+            db,
+            limit=args.limit,
+            only_unclassified=not args.recheck_pending,
+        )
+    finally:
+        await db.close()
+
+    print(f"Verarbeitet: {result['processed']}")
+    print(f"Erfolgreich: {result['success']}")
+    print(f"Fehler:      {len(result['errors'])}")
+    print(f"Kosten:      ${result['total_cost_usd']:.4f}")
+    if result["errors"]:
+        print("\nFehler-Details:")
+        for e in result["errors"][:10]:
+            print(f"  source_id={e['source_id']}: {e['error']}")
+
+
+if __name__ == "__main__":
+    asyncio.run(main())

src/agents/analyzer.py

@@ -16,7 +16,7 @@ WICHTIG: Verwende IMMER echte UTF-8-Umlaute (ä, ö, ü, ß) — NIEMALS Umschre
 VORFALL: {title}
 KONTEXT: {description}
 
-VORHANDENE MELDUNGEN:
+{fact_context_block}VORHANDENE MELDUNGEN:
 {articles_text}
 
 AUFTRAG:
@@ -47,7 +47,6 @@ Antworte AUSSCHLIESSLICH als JSON-Objekt mit diesen Feldern:
 - "summary": Zusammenfassung auf {output_language} mit Quellenverweisen [1], [2] etc. im Text (Markdown-Überschriften ## erlaubt wenn sinnvoll, aber KEINE "## ZUSAMMENFASSUNG"/"## ÜBERBLICK"-Sektion)
 - "sources": Array von Quellenobjekten, je: {{"nr": 1, "name": "Quellenname", "url": "https://..."}}
 - "key_facts": Array von bestätigten Kernfakten (Strings, in Ausgabesprache)
-- "translations": Array von Objekten mit "article_id", "headline_de", "content_de" (nur für fremdsprachige Artikel)
 
 Antworte NUR mit dem JSON-Objekt. Keine Einleitung, keine Erklärung."""
 
@@ -60,7 +59,7 @@ WICHTIG: Verwende IMMER echte UTF-8-Umlaute (ä, ö, ü, ß) — NIEMALS Umschre
 THEMA: {title}
 KONTEXT: {description}
 
-VORLIEGENDE QUELLEN:
+{fact_context_block}VORLIEGENDE QUELLEN:
 {articles_text}
 
 AUFTRAG:
@@ -102,7 +101,6 @@ Antworte AUSSCHLIESSLICH als JSON-Objekt mit diesen Feldern:
 - "summary": Das strukturierte Briefing als Markdown-Text mit Quellenverweisen [1], [2] etc.
 - "sources": Array von Quellenobjekten, je: {{"nr": 1, "name": "Quellenname", "url": "https://..."}}
 - "key_facts": Array von gesicherten Kernfakten (Strings, in Ausgabesprache)
-- "translations": Array von Objekten mit "article_id", "headline_de", "content_de" (nur für fremdsprachige Artikel)
 
 Antworte NUR mit dem JSON-Objekt. Keine Einleitung, keine Erklärung."""
 
@@ -120,7 +118,7 @@ BISHERIGES LAGEBILD:
 BISHERIGE QUELLEN:
 {previous_sources_text}
 
-NEUE MELDUNGEN SEIT DEM LETZTEN UPDATE:
+{fact_context_block}NEUE MELDUNGEN SEIT DEM LETZTEN UPDATE:
 {new_articles_text}
 
 AUFTRAG:
@@ -149,7 +147,6 @@ Antworte AUSSCHLIESSLICH als JSON-Objekt mit diesen Feldern:
 - "summary": Aktualisierte Zusammenfassung mit Quellenverweisen [1], [2] etc.
 - "sources": Array mit NUR den NEUEN Quellen aus den neuen Meldungen, je: {{"nr": <fortlaufende ganze Zahl, KEINE Buchstaben-Suffixe>, "name": "Quellenname", "url": "https://..."}}. Alte Quellen werden automatisch gemerged.
 - "key_facts": Array aller aktuellen Kernfakten (in Ausgabesprache)
-- "translations": Array von Objekten mit "article_id", "headline_de", "content_de" (nur für neue fremdsprachige Artikel)
 
 Antworte NUR mit dem JSON-Objekt. Keine Einleitung, keine Erklärung."""
 
@@ -168,7 +165,7 @@ BISHERIGES BRIEFING:
 BISHERIGE QUELLEN:
 {previous_sources_text}
 
-NEUE QUELLEN SEIT DEM LETZTEN UPDATE:
+{fact_context_block}NEUE QUELLEN SEIT DEM LETZTEN UPDATE:
 {new_articles_text}
 
 AUFTRAG:
@@ -201,7 +198,6 @@ Antworte AUSSCHLIESSLICH als JSON-Objekt mit diesen Feldern:
 - "summary": Das aktualisierte Briefing als Markdown-Text mit Quellenverweisen
 - "sources": Array mit NUR den NEUEN Quellen aus den neuen Meldungen, je: {{"nr": <fortlaufende ganze Zahl, KEINE Buchstaben-Suffixe>, "name": "Quellenname", "url": "https://..."}}. Alte Quellen werden automatisch gemerged.
 - "key_facts": Array aller gesicherten Kernfakten (in Ausgabesprache)
-- "translations": Array von Objekten mit "article_id", "headline_de", "content_de" (nur für neue fremdsprachige Artikel)
 
 Antworte NUR mit dem JSON-Objekt. Keine Einleitung, keine Erklärung."""
 
@@ -268,6 +264,112 @@ Antworte AUSSCHLIESSLICH als JSON-Objekt — KEINE Erklärung, KEINE Einleitung:
 {{"relevant_ids": [1, 3, 7]}}"""
 
 
+
+
+# Status-Gruppen fuer den Fakten-Kontext im Analyse-Prompt.
+# adhoc nutzt confirmed/unconfirmed/contradicted/developing,
+# research nutzt established/unverified/disputed/developing — beide Domaenen
+# werden in dieselben vier Anzeige-Gruppen abgebildet.
+_FACT_STATUS_GROUPS = [
+    ("Bestätigt (mehrere unabhängige Quellen oder durch Faktencheck als gesichert eingestuft):",
+     {"confirmed", "established"}),
+    ("Umstritten (Quellen widersprechen sich oder Faktencheck hat Widersprüche dokumentiert):",
+     {"contradicted", "disputed"}),
+    ("Unbestätigt (nur eine einzelne Quelle, eine unabhängige Bestätigung steht aus):",
+     {"unconfirmed", "unverified"}),
+    ("In Entwicklung (laufender Sachverhalt, Stand offen):",
+     {"developing"}),
+]
+
+_FACT_STATUS_PRIORITY = {
+    "confirmed": 5, "established": 5,
+    "contradicted": 4, "disputed": 4,
+    "unconfirmed": 3, "unverified": 3,
+    "developing": 1,
+}
+
+
+def build_fact_context_block(
+    existing_facts: list[dict] | None,
+    new_or_updated_facts: list[dict] | None,
+    incident_type: str,
+    max_total: int = 20,
+) -> str:
+    """Baut den 'GEPRUEFTE FAKTEN'-Block fuer den Analyse-Prompt.
+
+    Wird vom Orchestrator zwischen Faktencheck und Lagebild aufgerufen, damit
+    das Lagebild auf gepruefter Faktenbasis schreibt und Unklarheiten explizit
+    benennt. Bei leerer Faktenliste wird ein leerer String zurueckgegeben — der
+    Prompt laeuft dann ohne Fakten-Kontext (Fallback bei Faktencheck-Fail oder
+    bei Lagen ohne bisherige Fakten).
+    """
+    existing_facts = existing_facts or []
+    new_or_updated_facts = new_or_updated_facts or []
+    if not existing_facts and not new_or_updated_facts:
+        return ""
+
+    seen_claims: set[str] = set()
+    merged: list[dict] = []
+    # Neue/aktualisierte Fakten zuerst (Status ist aktueller Stand).
+    for f in new_or_updated_facts:
+        c = (f.get("claim") or "").strip().lower()
+        if not c or c in seen_claims:
+            continue
+        seen_claims.add(c)
+        merged.append(f)
+    # Dann alte unveraenderte Fakten.
+    for f in existing_facts:
+        c = (f.get("claim") or "").strip().lower()
+        if not c or c in seen_claims:
+            continue
+        seen_claims.add(c)
+        merged.append(f)
+
+    if not merged:
+        return ""
+
+    merged.sort(key=lambda f: (
+        -_FACT_STATUS_PRIORITY.get((f.get("status") or "").lower(), 0),
+        -(f.get("sources_count") or 0),
+    ))
+    merged = merged[:max_total]
+
+    grouped: dict[str, list[dict]] = {label: [] for label, _ in _FACT_STATUS_GROUPS}
+    for f in merged:
+        s = (f.get("status") or "").lower()
+        for label, codes in _FACT_STATUS_GROUPS:
+            if s in codes:
+                grouped[label].append(f)
+                break
+
+    if not any(grouped.values()):
+        return ""
+
+    lines: list[str] = []
+    lines.append("GEPRÜFTE FAKTEN (Stand nach dem Faktencheck dieses Refresh, max. {n} priorisiert):".format(n=max_total))
+    for label, _codes in _FACT_STATUS_GROUPS:
+        items = grouped[label]
+        if not items:
+            continue
+        lines.append("")
+        lines.append(label)
+        for f in items:
+            claim = (f.get("claim") or "").strip()
+            sc = f.get("sources_count") or 0
+            sc_text = f" ({sc} {'Quellen' if sc != 1 else 'Quelle'})" if sc else ""
+            lines.append(f"- {claim}{sc_text}")
+
+    lines.append("")
+    lines.append("AUSSAGE-DISZIPLIN für das Lagebild:")
+    lines.append("- Bestätigte Fakten als Grundgerüst nehmen, ohne Hedging.")
+    lines.append("- Umstrittene Punkte explizit als umstritten kennzeichnen, beide Seiten knapp benennen.")
+    lines.append("- Unbestätigtes klar einordnen ('Eine einzelne Quelle berichtet ...', 'Eine unabhängige Bestätigung steht aus.').")
+    lines.append("- Bei Aussagen, die durch keinen geprüften Fakt gedeckt sind und auch nicht direkt aus einer der vorliegenden Meldungen hervorgehen: NICHT spekulieren — entweder weglassen oder als unklar kennzeichnen.")
+    lines.append("- Triff KEINE Aussagen, die mit den oben gelisteten geprüften Fakten in Widerspruch stehen.")
+    lines.append("")
+    return "\n".join(lines)
+
+
 class AnalyzerAgent:
     """Analysiert und übersetzt Meldungen über Claude CLI."""
 
@@ -294,7 +396,7 @@ class AnalyzerAgent:
                 articles_text += f"Inhalt: {content[:800]}\n"
         return articles_text
 
-    async def analyze(self, title: str, description: str, articles: list[dict], incident_type: str = "adhoc") -> tuple[dict | None, ClaudeUsage | None]:
+    async def analyze(self, title: str, description: str, articles: list[dict], incident_type: str = "adhoc", fact_context_block: str = "") -> tuple[dict | None, ClaudeUsage | None]:
         """Erstanalyse: Analysiert alle Meldungen zu einem Vorfall (erster Refresh)."""
         if not articles:
             return None, None
@@ -310,6 +412,7 @@ class AnalyzerAgent:
             articles_text=articles_text,
             today=today,
             output_language=OUTPUT_LANGUAGE,
+            fact_context_block=fact_context_block,
         )
 
         try:
@@ -331,6 +434,7 @@ class AnalyzerAgent:
         previous_summary: str,
         previous_sources_json: str | None,
         incident_type: str = "adhoc",
+        fact_context_block: str = "",
     ) -> tuple[dict | None, ClaudeUsage | None]:
         """Inkrementelle Analyse: Aktualisiert das Lagebild mit nur den neuen Artikeln.
 
@@ -373,6 +477,7 @@ class AnalyzerAgent:
             new_articles_text=new_articles_text,
             today=today,
             output_language=OUTPUT_LANGUAGE,
+            fact_context_block=fact_context_block,
         )
 
         try:
@@ -796,5 +901,5 @@ class AnalyzerAgent:
             except json.JSONDecodeError:
                 pass
 
-        return {"summary": summary, "sources": sources, "key_facts": [], "translations": []}
+        return {"summary": summary, "sources": sources, "key_facts": []}
 

src/agents/orchestrator.py

@@ -489,6 +489,9 @@ class AgentOrchestrator:
 
             logger.info(f"Lage {incident_id} aus Warteschlange entfernt (removed={removed})")
 
+            # refresh_log-Eintrag schreiben, damit Auto-Refresh nicht im naechsten Tick erneut einreiht
+            await self._log_queued_cancellation(incident_id)
+
             # Send cancelled event
             if self._ws_manager:
                 try:
@@ -624,18 +627,56 @@ class AgentOrchestrator:
                 self._queue.task_done()
 
     async def _mark_refresh_cancelled(self, incident_id: int):
-        """Markiert den laufenden Refresh-Log-Eintrag als cancelled."""
+        """Markiert den laufenden Refresh-Log-Eintrag als cancelled und schliesst
+        alle noch aktiven Pipeline-Schritte. Ohne den zweiten Schritt blieb der
+        zuletzt aktive Step-Eintrag verwaist und das Frontend zeigte dauerhaft
+        'Schritt X laeuft', weil /api/incidents/<id>/pipeline aus
+        refresh_pipeline_steps liest."""
         from database import get_db
+        from services.pipeline_tracker import cancel_active_steps
         db = await get_db()
         try:
+            now_str = datetime.now(TIMEZONE).strftime('%Y-%m-%d %H:%M:%S')
+            cur = await db.execute(
+                "SELECT id FROM refresh_log WHERE incident_id = ? AND status = 'running'",
+                (incident_id,),
+            )
+            row = await cur.fetchone()
+            refresh_log_id = row["id"] if row else None
+
             await db.execute(
                 """UPDATE refresh_log SET status = 'cancelled', error_message = 'Vom Nutzer abgebrochen',
                    completed_at = ? WHERE incident_id = ? AND status = 'running'""",
-                (datetime.now(TIMEZONE).strftime('%Y-%m-%d %H:%M:%S'), incident_id),
+                (now_str, incident_id),
+            )
+            await db.commit()
+
+            if refresh_log_id is not None:
+                await cancel_active_steps(db, refresh_log_id=refresh_log_id)
+        except Exception as e:
+            logger.warning(f"Konnte Refresh-Log nicht als abgebrochen markieren: {e}")
+        finally:
+            await db.close()
+
+    async def _log_queued_cancellation(self, incident_id: int):
+        """Schreibt einen cancelled-Eintrag fuer einen Queue-Abbruch (Lage war noch nicht laufend).
+        Verhindert, dass der Auto-Refresh-Scheduler im naechsten Tick sofort wieder einreiht."""
+        from database import get_db
+        db = await get_db()
+        try:
+            cur = await db.execute("SELECT tenant_id FROM incidents WHERE id = ?", (incident_id,))
+            row = await cur.fetchone()
+            tid = row["tenant_id"] if row else None
+            now_str = datetime.now(TIMEZONE).strftime("%Y-%m-%d %H:%M:%S")
+            await db.execute(
+                """INSERT INTO refresh_log (incident_id, started_at, completed_at, status,
+                   trigger_type, error_message, tenant_id)
+                   VALUES (?, ?, ?, 'cancelled', 'manual', 'Aus Warteschlange entfernt', ?)""",
+                (incident_id, now_str, now_str, tid),
             )
             await db.commit()
         except Exception as e:
-            logger.warning(f"Konnte Refresh-Log nicht als abgebrochen markieren: {e}")
+            logger.warning(f"Konnte Queue-Cancel nicht in refresh_log loggen: {e}")
         finally:
             await db.close()
 
@@ -1258,18 +1299,22 @@ class AgentOrchestrator:
                 except Exception as e:
                     logger.warning("Bias-Anreicherung fehlgeschlagen (Pipeline laeuft weiter): %s", e)
 
-                # --- Analyse-Task ---
-                async def _do_analysis():
+                # --- Analyse-Task (wird nach _do_factcheck mit fact_context_block aufgerufen) ---
+                async def _do_analysis(fact_context_block: str = ""):
                     analyzer = AnalyzerAgent()
                     if previous_summary and new_count > 0:
                         logger.info(f"Inkrementelle Analyse: {new_count} neue Artikel zum bestehenden Lagebild")
                         return await analyzer.analyze_incremental(
                             title, description, new_articles_for_analysis,
                             previous_summary, previous_sources_json, incident_type,
+                            fact_context_block=fact_context_block,
                         )
                     else:
                         logger.info("Erstanalyse: Alle Artikel werden analysiert")
-                        return await analyzer.analyze(title, description, all_articles_preloaded, incident_type)
+                        return await analyzer.analyze(
+                            title, description, all_articles_preloaded, incident_type,
+                            fact_context_block=fact_context_block,
+                        )
 
                 # --- Faktencheck-Task ---
                 async def _do_factcheck():
@@ -1303,20 +1348,61 @@ class AgentOrchestrator:
                             articles_for_check = [dict(row) for row in await cursor.fetchall()]
                         return await factchecker.check(title, articles_for_check, incident_type)
 
-                # Pipeline-Schritte 6+7: Lagebild verfassen + Fakten prüfen (Start, parallel)
-                await _pipe_start("summary")
+                # Pipeline-Schritt 6: Faktencheck zuerst (sequenziell). Liefert den
+                # Faktenkontext fuer das Lagebild, damit dieses auf geprueftem Stand
+                # schreibt und Unklarheiten explizit benennt. Variante 1: bei
+                # Faktencheck-Fehler faellt das Lagebild auf den alten Pfad ohne
+                # Faktenkontext zurueck (Refresh bricht NICHT ab).
                 await _pipe_start("factcheck")
-
-                # Beide Tasks PARALLEL starten
-                logger.info("Starte Analyse und Faktencheck parallel...")
-                analysis_result, factcheck_result = await asyncio.gather(
-                    _do_analysis(),
-                    _do_factcheck(),
+                factcheck_result: tuple = ([], None)
+                fact_context_block = ""
+                factcheck_failed_reason: str | None = None
+                try:
+                    factcheck_result = await _do_factcheck()
+                except Exception as fc_err:
+                    factcheck_failed_reason = str(fc_err)
+                    logger.warning(
+                        "Faktencheck fehlgeschlagen, Lagebild laeuft ohne Faktenkontext: %s",
+                        fc_err, exc_info=True,
                     )
 
+                fact_checks, fc_usage = factcheck_result if factcheck_result else ([], None)
+
+                # Pipeline-Schritt 6 done direkt nach dem Aufruf — die finale
+                # DB-Persistierung passiert weiter unten, aber fuer die UI ist
+                # der Faktencheck-Aufruf hier abgeschlossen. Der count_value
+                # ist eine Schaetzung (echte Zahl steht spaeter in der DB).
+                _fc_estimated_new = max(0, len(fact_checks or []) - len(existing_facts or []))
+                await _pipe_done(
+                    "factcheck",
+                    count_value=_fc_estimated_new,
+                    count_secondary=len(fact_checks) if fact_checks else 0,
+                )
+
+                # Faktenkontext fuer das Lagebild bauen.
+                try:
+                    from agents.analyzer import build_fact_context_block as _build_fc_ctx
+                    fact_context_block = _build_fc_ctx(
+                        existing_facts or [], fact_checks or [], incident_type,
+                    )
+                    if fact_context_block:
+                        logger.info(
+                            "Faktenkontext fuer Lagebild: %d Zeichen, basierend auf %d alten + %d neuen Fakten",
+                            len(fact_context_block), len(existing_facts or []), len(fact_checks or []),
+                        )
+                except Exception as ctx_err:
+                    logger.warning("build_fact_context_block fehlgeschlagen: %s", ctx_err, exc_info=True)
+                    fact_context_block = ""
+
+                # Pipeline-Schritt 7: Lagebild verfassen (jetzt mit Faktenkontext)
+                await _pipe_start("summary")
+                logger.info(
+                    "Starte Lagebild (sequenziell nach Faktencheck%s)",
+                    " — OHNE Faktenkontext (Fallback)" if factcheck_failed_reason else "",
+                )
+                analysis_result = await _do_analysis(fact_context_block)
+
                 analysis, analysis_usage = analysis_result
-                fact_checks, fc_usage = factcheck_result
-                # Pipeline-Schritt 6: Lagebild verfassen (fertig, keine Zahl, nur Status)
                 await _pipe_done("summary", count_value=None, count_secondary=None)
 
                 # --- Analyse-Ergebnisse verarbeiten ---
@@ -1410,20 +1496,64 @@ class AgentOrchestrator:
                              snap_articles, snap_fcs, log_id, now, tenant_id),
                         )
 
-                    # Übersetzungen aktualisieren (nur für gültige DB-IDs)
-                    for translation in analysis.get("translations", []):
-                        article_id = translation.get("article_id")
-                        if isinstance(article_id, int):
-                            await db.execute(
-                                "UPDATE articles SET headline_de = ?, content_de = ? WHERE id = ? AND incident_id = ?",
-                                (translation.get("headline_de"), translation.get("content_de"), article_id, incident_id),
-                            )
+                    # Translations werden vom dedizierten Translator-Agent unten
+                    # erzeugt (frueher inline im Analyzer-Output, das war token-
+                    # instabil und schaetzte regelmaessig content_de aus).
 
                     await db.commit()
 
                 # Cancel-Check nach paralleler Verarbeitung
                 self._check_cancelled(incident_id)
 
+                # --- Translator (Haiku) fuer fremdsprachige Artikel ohne DE-Texte ---
+                # Idempotent: nur Artikel ohne headline_de/content_de werden geholt.
+                # Lauft nach der Analyse (Lagebild ist schon committed) und vor QC
+                # (damit normalize_umlaut_articles auch die frischen DE-Texte fasst).
+                try:
+                    tr_cursor = await db.execute(
+                        """SELECT id, headline, content_original, language
+                           FROM articles
+                           WHERE incident_id = ?
+                             AND language IS NOT NULL AND LOWER(language) != 'de'
+                             AND (headline_de IS NULL OR headline_de = ''
+                                  OR content_de IS NULL OR content_de = '')""",
+                        (incident_id,),
+                    )
+                    pending_translations = [dict(r) for r in await tr_cursor.fetchall()]
+                    if pending_translations:
+                        logger.info(
+                            "Translator fuer Incident %d: %d Artikel ohne DE-Uebersetzung",
+                            incident_id, len(pending_translations),
+                        )
+                        from agents.translator import translate_articles
+                        from services.post_refresh_qc import normalize_german_umlauts as _norm_de2
+                        translations = await translate_articles(
+                            pending_translations,
+                            output_lang="de",
+                            usage_accumulator=usage_acc,
+                        )
+                        for t in translations:
+                            hd = t.get("headline_de")
+                            cd = t.get("content_de")
+                            if hd:
+                                hd, _ = _norm_de2(hd)
+                            if cd:
+                                cd, _ = _norm_de2(cd)
+                            if hd or cd:
+                                await db.execute(
+                                    "UPDATE articles SET headline_de = COALESCE(?, headline_de), "
+                                    "content_de = COALESCE(?, content_de) WHERE id = ? AND incident_id = ?",
+                                    (hd, cd, t["id"], incident_id),
+                                )
+                        await db.commit()
+                        logger.info(
+                            "Translator fuer Incident %d: %d/%d Artikel uebersetzt",
+                            incident_id, len(translations), len(pending_translations),
+                        )
+                except Exception as e:
+                    logger.error("Translator-Fehler fuer Incident %d: %s", incident_id, e, exc_info=True)
+                    # Refresh trotz Translator-Fehler weiterlaufen lassen
+
                 # --- Neueste Entwicklungen (nur Live-Monitoring / adhoc) ---
                 # Basis ist jetzt das frisch generierte Lagebild (autoritativ, thematisch sauber).
                 # Zeitstempel und Quellen kommen aus den jüngsten belegenden Artikeln.
@@ -1571,9 +1701,10 @@ class AgentOrchestrator:
 
                 await db.commit()
 
-                # Pipeline-Schritt 7: Fakten prüfen (fertig)
-                _new_facts_count = max(0, len(fact_checks) - len(existing_facts))
-                await _pipe_done("factcheck", count_value=_new_facts_count, count_secondary=len(fact_checks) if fact_checks else 0)
+                # Pipeline-Schritt 7 (Fakten pruefen) wurde bereits frueher als done
+                # markiert (siehe weiter oben — direkt nach dem _do_factcheck-Aufruf,
+                # bevor das Lagebild generiert wurde). Hier nur noch die DB-
+                # Persistierung der Fakten, ohne den Step erneut zu schliessen.
 
                 # Pipeline-Schritt 8: Qualitätscheck (Start, ohne Zahlen)
                 await _pipe_start("qc")

src/agents/researcher.py

@@ -77,7 +77,7 @@ REGELN:
 {language_instruction}
 - Faktenbasiert und neutral - keine Spekulationen
 - KRITISCH für source_url: Kopiere die EXAKTE URL aus den WebSearch-Ergebnissen. Erfinde oder konstruiere NIEMALS URLs aus Mustern oder Erinnerung. Wenn du die exakte URL eines Artikels nicht aus den Suchergebnissen hast, lass diesen Artikel komplett weg.
-- Nutze removepaywalls.com für Paywall-geschützte Artikel (z.B. Spiegel+, Zeit+, SZ+): https://www.removepaywalls.com/search?url=ARTIKEL_URL
+- Nutze removepaywall.com für Paywall-geschützte Artikel (z.B. Spiegel+, Zeit+, SZ+): https://www.removepaywall.com/search?url=ARTIKEL_URL
 - Nutze WebFetch um die 3-5 wichtigsten Artikel vollständig abzurufen und zusammenzufassen
 
 Gib die Ergebnisse AUSSCHLIESSLICH als JSON-Array zurück, ohne Erklärungen davor oder danach.
@@ -124,7 +124,7 @@ Nutze spezifische Suchbegriffe für institutionelle Quellen. Ziel: 6-10 weitere
 PHASE 4 — VERIFIKATION UND VERTIEFUNG:
 Nutze WebFetch um die 6-10 wichtigsten Artikel vollständig abzurufen und ausführlich zusammenzufassen.
 Priorisiere dabei Primärquellen und investigative Berichte.
-Nutze removepaywalls.com für Paywall-geschützte Artikel (z.B. https://www.removepaywalls.com/search?url=ARTIKEL_URL)
+Nutze removepaywall.com für Paywall-geschützte Artikel (z.B. https://www.removepaywall.com/search?url=ARTIKEL_URL)
 
 {language_instruction}
 
@@ -199,14 +199,22 @@ AKTUELLE HEADLINES (die letzten Meldungen zu diesem Thema):
 
 AUFGABE:
 Generiere 5 Begriffspaare (DE + EN), mit denen neue RSS-Artikel zu diesem Thema gefunden werden.
-Ein Artikel gilt als relevant, wenn mindestens 2 dieser Begriffe im Titel oder der Beschreibung vorkommen.
+Ein Artikel gilt als relevant, wenn mindestens 2 dieser Begriffe im Titel oder der Beschreibung vorkommen
+- bei spezifischen Begriffen (Eigennamen, lange Begriffe ab 7 Zeichen) reicht 1 Treffer.
 
 REGELN:
-- Die ersten 2 Begriffspaare MUESSEN die zentralen Akteure/Laender/Themen sein (z.B. iran, israel, usa) — also die Begriffe, die in fast JEDEM Artikel zum Thema vorkommen
-- Die letzten 3 Begriffspaare sind aktuelle Entwicklungen aus den Headlines (Orte, Akteure, Schluesselwoerter der aktuellen Phase)
-- Begriffe muessen so gewaehlt sein, dass sie in kurzen RSS-Titeln matchen (einzelne Woerter, keine Phrasen)
-- Alle Begriffe in Kleinbuchstaben
-- Exakt 5 Begriffspaare
+- ZWINGEND: Eigennamen oder spezifische Begriffe aus dem THEMA (z.B. Personennamen, Tiernamen,
+  Ortsnamen wie "timmy", "buckelwal", "merz", "dobrindt") MUESSEN als eigene Begriffspaare
+  enthalten sein. Solche Begriffe sind oft das einzige, was in kurzen Headlines vorkommt.
+- Die ersten 2 Begriffspaare sind die zentralen Akteure/Laender/Themen (z.B. iran, israel,
+  buckelwal, timmy) — also die Begriffe, die in fast JEDEM Artikel zum Thema vorkommen.
+- Die uebrigen 3 Begriffspaare sind aktuelle Entwicklungen aus den Headlines (Orte, Akteure,
+  Schluesselwoerter der aktuellen Phase).
+- Wenn DE und EN identisch sind (Eigennamen), trotzdem das Paar einreichen.
+- Begriffe muessen so gewaehlt sein, dass sie in kurzen RSS-Titeln matchen (einzelne Woerter,
+  keine Phrasen, keine Konjunktionen).
+- Alle Begriffe in Kleinbuchstaben.
+- Exakt 5 Begriffspaare.
 
 Antwort NUR als JSON-Array:
 [{{"de": "iran", "en": "iran"}}, {{"de": "israel", "en": "israel"}}, {{"de": "teheran", "en": "tehran"}}, {{"de": "luftangriff", "en": "airstrike"}}, {{"de": "trump", "en": "trump"}}]"""
@@ -365,6 +373,17 @@ class ResearcherAgent:
                 if en and en != de:
                     keywords.append(en)
 
+            # Bug-2-Fallback: Lagentitel-Wörter (>=4 Zeichen) zwingend in Keyword-Liste,
+            # falls Haiku sie weggelassen hat. Verhindert "Buckelwal timmy"-Bug, bei dem
+            # der Eigenname "timmy" fehlte und damit Headlines mit nur "Buckelwal" durchfielen.
+            STOPWORDS = {"der", "die", "das", "und", "oder", "von", "vom", "zum", "zur",
+                         "the", "and", "for", "with", "ueber", "über", "von", "for"}
+            for word in (title or "").lower().split():
+                w = word.strip(".,;:!?\"\'()[]{}")
+                if len(w) >= 4 and w not in STOPWORDS and w not in keywords:
+                    keywords.append(w)
+                    logger.info(f"Lagentitel-Keyword '{w}' nachträglich injiziert")
+
             if keywords:
                 logger.info(f"Dynamische Keywords ({len(keywords)}): {keywords}")
             return keywords if keywords else None, usage

src/agents/translator.py

@@ -0,0 +1,254 @@
+"""Translator-Agent: uebersetzt fremdsprachige Artikel ins Deutsche.
+
+Eigener Agent (separat vom Analyzer), damit Token-Limits nicht zwischen
+Lagebild und Uebersetzung konkurrieren. Nutzt CLAUDE_MODEL_FAST (Haiku) in
+Batches.
+
+Aufgerufen vom Orchestrator nach analyzer.analyze() und vor post_refresh_qc.
+Backfill-Skript nutzt dieselbe Funktion fuer rueckwirkendes Auffuellen.
+"""
+import json
+import logging
+import re
+
+from agents.claude_client import call_claude, ClaudeUsage, UsageAccumulator
+from config import CLAUDE_MODEL_FAST, TRANSLATOR_ENABLED
+
+logger = logging.getLogger("osint.translator")
+
+# Pro Batch nicht mehr als so viele Artikel an Claude geben.
+# Bei Haiku ist das Output-Limit ca. 8k Tokens. Pro Artikel kommen leicht
+# 400-600 Tokens raus (headline_de + content_de bis 1000 Zeichen). Bei 15
+# wurde regelmaessig getrunkt (mid-JSON broken). 5 ist sicher mit Reserve.
+DEFAULT_BATCH_SIZE = 5
+
+# content_original wird ohnehin auf 1000 Zeichen gecappt (rss_parser).
+# Fuer den Translator nochmal verkuerzen, falls vorhanden mehr.
+CONTENT_INPUT_MAX = 1200
+
+# content_de soll wie content_original auf 1000 Zeichen begrenzt sein.
+CONTENT_OUTPUT_MAX = 1000
+
+
+def _extract_complete_objects(text: str) -> list[dict]:
+    """Extrahiert vollstaendige JSON-Objekte aus moeglicherweise abgeschnittenem Text.
+
+    Klammer-Counter-Ansatz: jedes balancierte {...} wird probiert.
+    """
+    results = []
+    depth = 0
+    start = -1
+    in_string = False
+    escape = False
+    for i, ch in enumerate(text):
+        if escape:
+            escape = False
+            continue
+        if ch == "\\":
+            escape = True
+            continue
+        if ch == '"' and not escape:
+            in_string = not in_string
+            continue
+        if in_string:
+            continue
+        if ch == "{":
+            if depth == 0:
+                start = i
+            depth += 1
+        elif ch == "}":
+            depth -= 1
+            if depth == 0 and start >= 0:
+                obj_text = text[start:i + 1]
+                try:
+                    obj = json.loads(obj_text)
+                    if isinstance(obj, dict):
+                        results.append(obj)
+                except json.JSONDecodeError:
+                    pass
+                start = -1
+    return results
+
+
+def _build_prompt(articles: list[dict], output_lang: str = "de") -> str:
+    """Bauen den Translation-Prompt fuer eine Batch."""
+    lang_label = {"de": "Deutsch", "en": "Englisch"}.get(output_lang, output_lang)
+
+    items = []
+    for a in articles:
+        items.append({
+            "id": a["id"],
+            "headline": a.get("headline", "") or "",
+            "content": (a.get("content_original") or "")[:CONTENT_INPUT_MAX],
+            "source_lang": a.get("language", "en"),
+        })
+
+    return f"""Du bist ein praeziser Uebersetzer fuer Nachrichten-Artikel.
+Uebersetze die folgenden Artikel nach {lang_label}.
+
+WICHTIG:
+- Verwende IMMER echte UTF-8-Umlaute (ä, ö, ü, ß) - NIEMALS Umschreibungen wie ae, oe, ue, ss.
+  Beispiele: "Gespraeche" -> "Gespräche", "Fuehrer" -> "Führer", "grosse" -> "große".
+- Behalte Eigennamen (Personen, Orte, Organisationen) im Original.
+- Headline kurz und buendig wie im Original.
+- Content auf MAX {CONTENT_OUTPUT_MAX} Zeichen kuerzen, kein HTML, kein Markdown.
+- Wenn der Artikel schon auf {lang_label} ist (z.B. source_lang="{output_lang}"),
+  kopiere headline und content unveraendert.
+
+Antworte AUSSCHLIESSLICH mit einem flachen JSON-Array (kein Wrapper-Objekt!).
+Format genau so:
+[
+  {{"id": 1, "headline_de": "Titel auf Deutsch", "content_de": "Inhalt auf Deutsch"}},
+  {{"id": 2, "headline_de": "...", "content_de": "..."}}
+]
+
+NICHT erlaubt: {{"translations": [...]}} oder {{"items": [...]}} oder Markdown-Codefences.
+Nur das Array, ohne Einleitung, ohne Erklaerung.
+
+ARTIKEL:
+{json.dumps(items, ensure_ascii=False, indent=2)}
+"""
+
+
+def _parse_response(text: str) -> list[dict]:
+    """Robustes JSON-Array-Parsing.
+
+    Handhabt:
+    - reines JSON
+    - JSON in Markdown-Codefence ```json ... ```
+    - abgeschnittene Antworten (extrahiert vollstaendige Top-Level-Objekte)
+    """
+    text = text.strip()
+    # Markdown-Codefence entfernen
+    if text.startswith("```"):
+        text = re.sub(r"^```(?:json)?\s*", "", text)
+        text = re.sub(r"\s*```\s*$", "", text)
+        text = text.strip()
+
+    try:
+        data = json.loads(text)
+    except json.JSONDecodeError:
+        # Erst Array versuchen
+        match = re.search(r"\[.*\]", text, re.DOTALL)
+        if match:
+            try:
+                data = json.loads(match.group(0))
+            except json.JSONDecodeError:
+                # Truncate-Fallback: einzelne Top-Level-Objekte extrahieren
+                data = _extract_complete_objects(text)
+        else:
+            data = _extract_complete_objects(text)
+
+    # Claude wraps das Array gelegentlich in {"translations": [...]} oder {"items": [...]}
+    if isinstance(data, dict):
+        for key in ("translations", "items", "results", "data"):
+            if isinstance(data.get(key), list):
+                data = data[key]
+                break
+        else:
+            # Einzelnes Objekt? Dann als Liste mit einem Element behandeln
+            if "id" in data:
+                data = [data]
+            else:
+                raise ValueError(f"Translator-Antwort: Dict ohne erwarteten Array-Key (keys={list(data.keys())[:5]})")
+
+    if not isinstance(data, list):
+        raise ValueError(f"Translator-Antwort ist kein Array: {type(data).__name__}")
+
+    cleaned = []
+    for item in data:
+        if not isinstance(item, dict):
+            continue
+        aid = item.get("id")
+        if not isinstance(aid, int):
+            try:
+                aid = int(aid)
+            except (TypeError, ValueError):
+                continue
+        cleaned.append({
+            "id": aid,
+            "headline_de": (item.get("headline_de") or "").strip() or None,
+            "content_de": (item.get("content_de") or "").strip() or None,
+        })
+    return cleaned
+
+
+async def translate_articles_batch(
+    articles: list[dict],
+    output_lang: str = "de",
+) -> tuple[list[dict], ClaudeUsage]:
+    """Uebersetzt eine Batch von Artikeln.
+
+    Erwartet articles als Liste von Dicts mit den Feldern id, headline,
+    content_original, language.
+
+    Rueckgabe: (uebersetzte_artikel, usage)
+    Wenn der Call fehlschlaegt, wird ([], leere_usage) zurueckgegeben - der
+    Caller kann entscheiden, ob retry oder skip.
+    """
+    if not articles:
+        return [], ClaudeUsage()
+
+    prompt = _build_prompt(articles, output_lang)
+
+    try:
+        result_text, usage = await call_claude(prompt, tools=None, model=CLAUDE_MODEL_FAST)
+    except Exception as e:
+        logger.error(f"Translator Claude-Call fehlgeschlagen: {e}")
+        return [], ClaudeUsage()
+
+    try:
+        translations = _parse_response(result_text)
+    except Exception as e:
+        logger.error(f"Translator JSON-Parsing fehlgeschlagen: {e}; raw: {result_text[:300]!r}")
+        return [], usage
+
+    # Validierung: nur Translations zurueckgeben, deren id wirklich
+    # in der angefragten Batch war
+    requested_ids = {a["id"] for a in articles}
+    valid = [t for t in translations if t["id"] in requested_ids]
+    if len(valid) != len(translations):
+        logger.warning(
+            "Translator: %d von %d Translations referenzieren unbekannte IDs",
+            len(translations) - len(valid), len(translations),
+        )
+    return valid, usage
+
+
+async def translate_articles(
+    articles: list[dict],
+    output_lang: str = "de",
+    batch_size: int = DEFAULT_BATCH_SIZE,
+    usage_accumulator: UsageAccumulator | None = None,
+) -> list[dict]:
+    """Uebersetzt eine beliebige Anzahl Artikel in Batches.
+
+    Bringt die Batches durch Logik in `translate_articles_batch` und gibt
+    EINE flache Liste der Translations zurueck. Wenn ein Batch fehlschlaegt,
+    wird er uebersprungen (anderer Batches laufen weiter).
+    """
+    if not articles:
+        return []
+
+    if not TRANSLATOR_ENABLED:
+        logger.info(
+            "Translator deaktiviert (TRANSLATOR_ENABLED=false), %d Artikel uebersprungen",
+            len(articles),
+        )
+        return []
+
+    all_translations = []
+    for i in range(0, len(articles), batch_size):
+        batch = articles[i : i + batch_size]
+        translations, usage = await translate_articles_batch(batch, output_lang)
+        if usage_accumulator is not None:
+            usage_accumulator.add(usage)
+        all_translations.extend(translations)
+        logger.info(
+            "Translator-Batch %d/%d: %d/%d uebersetzt (cost=$%.4f)",
+            (i // batch_size) + 1,
+            (len(articles) + batch_size - 1) // batch_size,
+            len(translations), len(batch),
+            usage.cost_usd,
+        )
+    return all_translations

src/config.py

@@ -41,6 +41,10 @@ OUTPUT_LANGUAGE = "Deutsch"
 # In Kundenversion auf False setzen oder Env-Variable entfernen
 DEV_MODE = os.environ.get("DEV_MODE", "true").lower() == "true"
 
+# Feature-Flag: Translator-Agent (Haiku) komplett deaktivieren.
+# False = keine Uebersetzungen mehr, fremdsprachige Artikel bleiben unuebersetzt.
+TRANSLATOR_ENABLED = os.environ.get("TRANSLATOR_ENABLED", "true").lower() == "true"
+
 # RSS-Feeds (Fallback, primär aus DB geladen)
 RSS_FEEDS = {
     "deutsch": [
@@ -91,3 +95,9 @@ TELEGRAM_API_ID = int(os.environ.get("TELEGRAM_API_ID", "0"))
 TELEGRAM_API_HASH = os.environ.get("TELEGRAM_API_HASH", "")
 TELEGRAM_SESSION_PATH = os.environ.get("TELEGRAM_SESSION_PATH", "/home/claude-dev/.telegram/telegram_session")
 
+# Health-Check (genutzt von services/source_health.py)
+HEALTH_CHECK_USER_AGENT = os.environ.get(
+    "HEALTH_CHECK_USER_AGENT",
+    "Mozilla/5.0 (compatible; AegisSight-HealthCheck/1.0)",
+)
+HEALTH_CHECK_TIMEOUT_S = float(os.environ.get("HEALTH_CHECK_TIMEOUT_S", "15.0"))

src/database.py

@@ -158,7 +158,36 @@ CREATE TABLE IF NOT EXISTS sources (
     article_count INTEGER DEFAULT 0,
     last_seen_at TIMESTAMP,
     created_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP,
-    tenant_id INTEGER REFERENCES organizations(id)
+    tenant_id INTEGER REFERENCES organizations(id),
+    language TEXT,
+    bias TEXT,
+    political_orientation TEXT DEFAULT 'na',
+    media_type TEXT DEFAULT 'sonstige',
+    reliability TEXT DEFAULT 'na',
+    state_affiliated INTEGER DEFAULT 0,
+    country_code TEXT,
+    classification_source TEXT DEFAULT 'legacy',
+    classified_at TIMESTAMP,
+    proposed_political_orientation TEXT,
+    proposed_media_type TEXT,
+    proposed_reliability TEXT,
+    proposed_state_affiliated INTEGER,
+    proposed_country_code TEXT,
+    proposed_alignments_json TEXT,
+    proposed_confidence REAL,
+    proposed_reasoning TEXT,
+    proposed_at TIMESTAMP,
+    eu_disinfo_listed INTEGER DEFAULT 0,
+    eu_disinfo_case_count INTEGER DEFAULT 0,
+    eu_disinfo_last_seen TIMESTAMP,
+    ifcn_signatory INTEGER DEFAULT 0,
+    external_data_synced_at TIMESTAMP
+);
+
+CREATE TABLE IF NOT EXISTS source_alignments (
+    source_id INTEGER NOT NULL REFERENCES sources(id) ON DELETE CASCADE,
+    alignment TEXT NOT NULL,
+    PRIMARY KEY (source_id, alignment)
 );
 
 CREATE TABLE IF NOT EXISTS notifications (
@@ -611,6 +640,71 @@ async def init_db():
             await db.execute("ALTER TABLE sources ADD COLUMN tenant_id INTEGER REFERENCES organizations(id)")
             await db.commit()
 
+        # Migration: language + bias (Freitext, schon laenger im Einsatz, Schema-Lueck schliessen)
+        if "language" not in src_columns:
+            await db.execute("ALTER TABLE sources ADD COLUMN language TEXT")
+            await db.commit()
+        if "bias" not in src_columns:
+            await db.execute("ALTER TABLE sources ADD COLUMN bias TEXT")
+            await db.commit()
+
+        # Migration: strukturierte Klassifikations-Spalten fuer sources
+        for col, ddl in [
+            ("political_orientation", "ALTER TABLE sources ADD COLUMN political_orientation TEXT DEFAULT 'na'"),
+            ("media_type", "ALTER TABLE sources ADD COLUMN media_type TEXT DEFAULT 'sonstige'"),
+            ("reliability", "ALTER TABLE sources ADD COLUMN reliability TEXT DEFAULT 'na'"),
+            ("state_affiliated", "ALTER TABLE sources ADD COLUMN state_affiliated INTEGER DEFAULT 0"),
+            ("country_code", "ALTER TABLE sources ADD COLUMN country_code TEXT"),
+            ("classification_source", "ALTER TABLE sources ADD COLUMN classification_source TEXT DEFAULT 'legacy'"),
+            ("classified_at", "ALTER TABLE sources ADD COLUMN classified_at TIMESTAMP"),
+            ("proposed_political_orientation", "ALTER TABLE sources ADD COLUMN proposed_political_orientation TEXT"),
+            ("proposed_media_type", "ALTER TABLE sources ADD COLUMN proposed_media_type TEXT"),
+            ("proposed_reliability", "ALTER TABLE sources ADD COLUMN proposed_reliability TEXT"),
+            ("proposed_state_affiliated", "ALTER TABLE sources ADD COLUMN proposed_state_affiliated INTEGER"),
+            ("proposed_country_code", "ALTER TABLE sources ADD COLUMN proposed_country_code TEXT"),
+            ("proposed_alignments_json", "ALTER TABLE sources ADD COLUMN proposed_alignments_json TEXT"),
+            ("proposed_confidence", "ALTER TABLE sources ADD COLUMN proposed_confidence REAL"),
+            ("proposed_reasoning", "ALTER TABLE sources ADD COLUMN proposed_reasoning TEXT"),
+            ("proposed_at", "ALTER TABLE sources ADD COLUMN proposed_at TIMESTAMP"),
+        ]:
+            if col not in src_columns:
+                await db.execute(ddl)
+                await db.commit()
+        if any(c not in src_columns for c in ("political_orientation", "media_type", "reliability")):
+            logger.info("Migration: Klassifikations-Spalten zu sources hinzugefuegt")
+
+        # Migration: externe Reputations-Daten (EUvsDisinfo + IFCN)
+        for col, ddl in [
+            ("eu_disinfo_listed", "ALTER TABLE sources ADD COLUMN eu_disinfo_listed INTEGER DEFAULT 0"),
+            ("eu_disinfo_case_count", "ALTER TABLE sources ADD COLUMN eu_disinfo_case_count INTEGER DEFAULT 0"),
+            ("eu_disinfo_last_seen", "ALTER TABLE sources ADD COLUMN eu_disinfo_last_seen TIMESTAMP"),
+            ("ifcn_signatory", "ALTER TABLE sources ADD COLUMN ifcn_signatory INTEGER DEFAULT 0"),
+            ("external_data_synced_at", "ALTER TABLE sources ADD COLUMN external_data_synced_at TIMESTAMP"),
+        ]:
+            if col not in src_columns:
+                await db.execute(ddl)
+                await db.commit()
+        if any(c not in src_columns for c in ("eu_disinfo_listed", "ifcn_signatory")):
+            logger.info("Migration: externe Reputations-Spalten zu sources hinzugefuegt")
+
+        # Migration: source_alignments-Tabelle (Mehrfach-Tags fuer geopolitische Naehe)
+        cursor = await db.execute(
+            "SELECT name FROM sqlite_master WHERE type='table' AND name='source_alignments'"
+        )
+        if not await cursor.fetchone():
+            await db.executescript(
+                """
+                CREATE TABLE source_alignments (
+                    source_id INTEGER NOT NULL REFERENCES sources(id) ON DELETE CASCADE,
+                    alignment TEXT NOT NULL,
+                    PRIMARY KEY (source_id, alignment)
+                );
+                CREATE INDEX IF NOT EXISTS idx_source_alignments_alignment ON source_alignments(alignment);
+                """
+            )
+            await db.commit()
+            logger.info("Migration: source_alignments-Tabelle erstellt")
+
         # Migration: tenant_id fuer notifications
         cursor = await db.execute("PRAGMA table_info(notifications)")
         notif_columns = [row[1] for row in await cursor.fetchall()]

src/feeds/rss_parser.py

@@ -6,6 +6,8 @@ import httpx
 from datetime import datetime, timezone
 from config import TIMEZONE, MAX_ARTICLES_PER_DOMAIN_RSS
 from source_rules import _extract_domain
+from feeds.transcript_extractors._common import html_to_text
+from services.post_refresh_qc import normalize_german_umlauts
 
 logger = logging.getLogger("osint.rss")
 
@@ -152,7 +154,15 @@ class RSSParser:
 
             for entry in feed.entries[:50]:
                 title = entry.get("title", "")
-                summary = entry.get("summary", "")
+                # RSS-summary ist bei vielen Quellen HTML (Guardian, AP, SZ, ...).
+                # Vor weiterer Verwendung strippen, sonst landet HTML in DB
+                # und KI-Agenten und Sprach-Heuristik werden gestoert.
+                summary_raw = entry.get("summary", "")
+                summary = html_to_text(summary_raw) if summary_raw else ""
+                # ASCII-Umlaut-Normalisierung (z.B. dpa-AFX schreibt "Gespraeche").
+                # Dictionary-basiert, sicher gegen englische Woerter wie "Boeing".
+                title, _ = normalize_german_umlauts(title)
+                summary, _ = normalize_german_umlauts(summary)
                 text = f"{title} {summary}".lower()
 
                 # Adaptive Match-Schwelle:

src/main.py

@@ -124,7 +124,7 @@ async def check_auto_refresh():
 
             # Letzten abgeschlossenen oder laufenden Refresh pruefen
             cursor = await db.execute(
-                "SELECT started_at, status FROM refresh_log WHERE incident_id = ? AND status IN ('completed', 'running') ORDER BY id DESC LIMIT 1",
+                "SELECT started_at, status FROM refresh_log WHERE incident_id = ? AND status IN ('completed', 'running', 'cancelled', 'error') ORDER BY id DESC LIMIT 1",
                 (incident_id,),
             )
             last_refresh = await cursor.fetchone()

src/middleware/license_check.py

@@ -47,7 +47,7 @@ async def require_writable_license(
     if lic.get("read_only"):
         reason = lic.get("read_only_reason") or "expired"
         if reason == "budget_exceeded":
-            detail = "Token-Budget aufgebraucht. Bitte Verwaltung kontaktieren."
+            detail = "Token-Budget aufgebraucht. Für Aufstockung oder Upgrade bitte info@aegis-sight.de kontaktieren."
         elif reason == "expired":
             detail = "Lizenz abgelaufen. Nur Lesezugriff moeglich."
         elif reason == "no_license":

src/models.py

@@ -54,7 +54,7 @@ class IncidentCreate(BaseModel):
     refresh_interval: int = Field(default=15, ge=10, le=10080)
     refresh_start_time: Optional[str] = Field(default=None, pattern=r"^([01]\d|2[0-3]):[0-5]\d$")
     retention_days: int = Field(default=0, ge=0, le=999)
-    international_sources: bool = True
+    international_sources: bool = False
     include_telegram: bool = False
     visibility: str = Field(default="public", pattern="^(public|private)$")
 
@@ -139,24 +139,51 @@ class IncidentListItem(BaseModel):
 
 
 # Sources (Quellenverwaltung)
+SOURCE_TYPE_PATTERN = "^(rss_feed|web_source|excluded|telegram_channel|podcast_feed)$"
+SOURCE_CATEGORY_PATTERN = "^(nachrichtenagentur|oeffentlich-rechtlich|qualitaetszeitung|behoerde|fachmedien|think-tank|international|regional|boulevard|sonstige)$"
+SOURCE_STATUS_PATTERN = "^(active|inactive)$"
+POLITICAL_ORIENTATION_PATTERN = "^(links_extrem|links|mitte_links|liberal|mitte|konservativ|mitte_rechts|rechts|rechts_extrem|na)$"
+MEDIA_TYPE_PATTERN = "^(tageszeitung|wochenzeitung|magazin|tv_sender|radio|oeffentlich_rechtlich|nachrichtenagentur|online_only|blog|telegram_kanal|telegram_bot|podcast|social_media|imageboard|think_tank|ngo|behoerde|staatsmedium|fachmedium|sonstige)$"
+RELIABILITY_PATTERN = "^(sehr_hoch|hoch|gemischt|niedrig|sehr_niedrig|na)$"
+ALIGNMENT_PATTERN = "^(prorussisch|proiranisch|prowestlich|proukrainisch|prochinesisch|projapanisch|proisraelisch|propalaestinensisch|protuerkisch|panarabisch|neutral|sonstige)$"
+COUNTRY_CODE_PATTERN = "^[A-Z]{2}$"
+CLASSIFICATION_SOURCE_PATTERN = "^(manual|llm_approved|llm_pending|legacy)$"
+
+
 class SourceCreate(BaseModel):
     name: str = Field(min_length=1, max_length=200)
     url: Optional[str] = None
     domain: Optional[str] = None
-    source_type: str = Field(default="rss_feed", pattern="^(rss_feed|web_source|excluded|telegram_channel|podcast_feed)$")
-    category: str = Field(default="sonstige", pattern="^(nachrichtenagentur|oeffentlich-rechtlich|qualitaetszeitung|behoerde|fachmedien|think-tank|international|regional|boulevard|sonstige)$")
-    status: str = Field(default="active", pattern="^(active|inactive)$")
+    source_type: str = Field(default="rss_feed", pattern=SOURCE_TYPE_PATTERN)
+    category: str = Field(default="sonstige", pattern=SOURCE_CATEGORY_PATTERN)
+    status: str = Field(default="active", pattern=SOURCE_STATUS_PATTERN)
     notes: Optional[str] = None
+    language: Optional[str] = None
+    bias: Optional[str] = None
+    political_orientation: Optional[str] = Field(default=None, pattern=POLITICAL_ORIENTATION_PATTERN)
+    media_type: Optional[str] = Field(default=None, pattern=MEDIA_TYPE_PATTERN)
+    reliability: Optional[str] = Field(default=None, pattern=RELIABILITY_PATTERN)
+    state_affiliated: Optional[bool] = None
+    country_code: Optional[str] = Field(default=None, pattern=COUNTRY_CODE_PATTERN)
+    alignments: Optional[list[str]] = None
 
 
 class SourceUpdate(BaseModel):
     name: Optional[str] = Field(default=None, max_length=200)
     url: Optional[str] = None
     domain: Optional[str] = None
-    source_type: Optional[str] = Field(default=None, pattern="^(rss_feed|web_source|excluded|telegram_channel|podcast_feed)$")
-    category: Optional[str] = Field(default=None, pattern="^(nachrichtenagentur|oeffentlich-rechtlich|qualitaetszeitung|behoerde|fachmedien|think-tank|international|regional|boulevard|sonstige)$")
-    status: Optional[str] = Field(default=None, pattern="^(active|inactive)$")
+    source_type: Optional[str] = Field(default=None, pattern=SOURCE_TYPE_PATTERN)
+    category: Optional[str] = Field(default=None, pattern=SOURCE_CATEGORY_PATTERN)
+    status: Optional[str] = Field(default=None, pattern=SOURCE_STATUS_PATTERN)
     notes: Optional[str] = None
+    language: Optional[str] = None
+    bias: Optional[str] = None
+    political_orientation: Optional[str] = Field(default=None, pattern=POLITICAL_ORIENTATION_PATTERN)
+    media_type: Optional[str] = Field(default=None, pattern=MEDIA_TYPE_PATTERN)
+    reliability: Optional[str] = Field(default=None, pattern=RELIABILITY_PATTERN)
+    state_affiliated: Optional[bool] = None
+    country_code: Optional[str] = Field(default=None, pattern=COUNTRY_CODE_PATTERN)
+    alignments: Optional[list[str]] = None
 
 
 class SourceResponse(BaseModel):
@@ -174,7 +201,20 @@ class SourceResponse(BaseModel):
     created_at: str
     language: Optional[str] = None
     bias: Optional[str] = None
+    political_orientation: Optional[str] = None
+    media_type: Optional[str] = None
+    reliability: Optional[str] = None
+    state_affiliated: bool = False
+    country_code: Optional[str] = None
+    classification_source: Optional[str] = None
+    classified_at: Optional[str] = None
+    alignments: list[str] = []
     is_global: bool = False
+    ifcn_signatory: bool = False
+    eu_disinfo_listed: bool = False
+    eu_disinfo_case_count: int = 0
+    eu_disinfo_last_seen: Optional[str] = None
+    external_data_synced_at: Optional[str] = None
 
 
 # Source Discovery

src/routers/sources.py

@@ -1,10 +1,13 @@
 """Sources-Router: Quellenverwaltung (Multi-Tenant)."""
+import json
 import logging
 from collections import defaultdict
-from fastapi import APIRouter, Depends, HTTPException, status
+from fastapi import APIRouter, BackgroundTasks, Depends, HTTPException, status
 from models import SourceCreate, SourceUpdate, SourceResponse, DiscoverRequest, DiscoverResponse, DiscoverMultiResponse, DomainActionRequest
 from auth import get_current_user
-from database import db_dependency, refresh_source_counts
+from database import db_dependency, get_db, refresh_source_counts
+from services.external_reputation import apply_reputation_overrides, sync_all as sync_external_reputation
+from services.source_classifier import bulk_classify, classify_source
 from source_rules import discover_source, discover_all_feeds, evaluate_feeds_with_claude, _extract_domain, _detect_category, domain_to_display_name, _DOMAIN_ALIASES
 import aiosqlite
 
@@ -12,7 +15,56 @@ logger = logging.getLogger("osint.sources")
 
 router = APIRouter(prefix="/api/sources", tags=["sources"])
 
-SOURCE_UPDATE_COLUMNS = {"name", "url", "domain", "source_type", "category", "status", "notes"}
+SOURCE_UPDATE_COLUMNS = {
+    "name", "url", "domain", "source_type", "category", "status", "notes",
+    "language", "bias",
+    "political_orientation", "media_type", "reliability",
+    "state_affiliated", "country_code",
+}
+SOURCE_CLASSIFICATION_FIELDS = {
+    "political_orientation", "media_type", "reliability",
+    "state_affiliated", "country_code",
+}
+ALLOWED_ALIGNMENTS = {
+    "prorussisch", "proiranisch", "prowestlich", "proukrainisch",
+    "prochinesisch", "projapanisch", "proisraelisch", "propalaestinensisch",
+    "protuerkisch", "panarabisch", "neutral", "sonstige",
+}
+
+
+async def _load_alignments_for(db: aiosqlite.Connection, source_ids: list[int]) -> dict[int, list[str]]:
+    """Lädt alignments fuer mehrere Quellen in einer Query und gibt {source_id: [alignment, ...]} zurück."""
+    if not source_ids:
+        return {}
+    placeholders = ",".join("?" for _ in source_ids)
+    cursor = await db.execute(
+        f"SELECT source_id, alignment FROM source_alignments WHERE source_id IN ({placeholders}) ORDER BY alignment",
+        source_ids,
+    )
+    out: dict[int, list[str]] = {sid: [] for sid in source_ids}
+    for row in await cursor.fetchall():
+        out.setdefault(row["source_id"], []).append(row["alignment"])
+    return out
+
+
+async def _replace_alignments(db: aiosqlite.Connection, source_id: int, alignments: list[str]):
+    """Ersetzt die alignments-Liste einer Quelle (DELETE + INSERT) — Aufrufer muss commit() machen."""
+    await db.execute("DELETE FROM source_alignments WHERE source_id = ?", (source_id,))
+    seen: set[str] = set()
+    for raw in alignments:
+        a = (raw or "").strip().lower()
+        if not a or a in seen:
+            continue
+        if a not in ALLOWED_ALIGNMENTS:
+            raise HTTPException(
+                status_code=status.HTTP_422_UNPROCESSABLE_ENTITY,
+                detail=f"Ungueltiger alignment-Wert: '{a}'",
+            )
+        seen.add(a)
+        await db.execute(
+            "INSERT INTO source_alignments (source_id, alignment) VALUES (?, ?)",
+            (source_id, a),
+        )
 
 
 def _check_source_ownership(source: dict, username: str):
@@ -34,6 +86,13 @@ async def list_sources(
     source_type: str = None,
     category: str = None,
     source_status: str = None,
+    political_orientation: str = None,
+    media_type: str = None,
+    reliability: str = None,
+    state_affiliated: bool = None,
+    alignment: str = None,
+    ifcn_signatory: bool = None,
+    eu_disinfo_listed: bool = None,
     current_user: dict = Depends(get_current_user),
     db: aiosqlite.Connection = Depends(db_dependency),
 ):
@@ -41,27 +100,51 @@ async def list_sources(
     tenant_id = current_user.get("tenant_id")
 
     # Global (tenant_id=NULL) + eigene Org
-    query = "SELECT * FROM sources WHERE (tenant_id IS NULL OR tenant_id = ?)"
-    params = [tenant_id]
+    query = "SELECT s.* FROM sources s WHERE (s.tenant_id IS NULL OR s.tenant_id = ?)"
+    params: list = [tenant_id]
 
     if source_type:
-        query += " AND source_type = ?"
+        query += " AND s.source_type = ?"
         params.append(source_type)
     if category:
-        query += " AND category = ?"
+        query += " AND s.category = ?"
         params.append(category)
     if source_status:
-        query += " AND status = ?"
+        query += " AND s.status = ?"
         params.append(source_status)
+    if political_orientation:
+        query += " AND s.political_orientation = ?"
+        params.append(political_orientation)
+    if media_type:
+        query += " AND s.media_type = ?"
+        params.append(media_type)
+    if reliability:
+        query += " AND s.reliability = ?"
+        params.append(reliability)
+    if state_affiliated is not None:
+        query += " AND s.state_affiliated = ?"
+        params.append(1 if state_affiliated else 0)
+    if alignment:
+        query += " AND EXISTS (SELECT 1 FROM source_alignments sa WHERE sa.source_id = s.id AND sa.alignment = ?)"
+        params.append(alignment.lower())
+    if ifcn_signatory is not None:
+        query += " AND s.ifcn_signatory = ?"
+        params.append(1 if ifcn_signatory else 0)
+    if eu_disinfo_listed is not None:
+        query += " AND s.eu_disinfo_listed = ?"
+        params.append(1 if eu_disinfo_listed else 0)
 
-    query += " ORDER BY source_type, category, name"
+    query += " ORDER BY s.source_type, s.category, s.name"
     cursor = await db.execute(query, params)
     rows = await cursor.fetchall()
-    results = []
-    for row in rows:
-        d = dict(row)
+    results = [dict(row) for row in rows]
+    alignments_map = await _load_alignments_for(db, [r["id"] for r in results])
+    for d in results:
         d["is_global"] = d.get("tenant_id") is None
-        results.append(d)
+        d["state_affiliated"] = bool(d.get("state_affiliated"))
+        d["ifcn_signatory"] = bool(d.get("ifcn_signatory"))
+        d["eu_disinfo_listed"] = bool(d.get("eu_disinfo_listed"))
+        d["alignments"] = alignments_map.get(d["id"], [])
     return results
 
 
@@ -454,10 +537,16 @@ async def create_source(
                     detail=f"Domain '{domain}' bereits als Quelle vorhanden: {domain_existing['name']}. Für einen neuen RSS-Feed bitte die Feed-URL angeben.",
                 )
 
-    cursor = await db.execute(
-        """INSERT INTO sources (name, url, domain, source_type, category, status, notes, added_by, tenant_id)
-           VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?)""",
-        (
+    payload = data.model_dump(exclude_unset=True)
+    alignments = payload.pop("alignments", None)
+    classification_touched = bool(SOURCE_CLASSIFICATION_FIELDS & payload.keys()) or alignments is not None
+
+    cols = ["name", "url", "domain", "source_type", "category", "status", "notes",
+            "language", "bias",
+            "political_orientation", "media_type", "reliability",
+            "state_affiliated", "country_code",
+            "added_by", "tenant_id"]
+    vals = [
         data.name,
         data.url,
         domain,
@@ -465,15 +554,43 @@ async def create_source(
         data.category,
         data.status,
         data.notes,
+        payload.get("language"),
+        payload.get("bias"),
+        payload.get("political_orientation"),
+        payload.get("media_type"),
+        payload.get("reliability"),
+        1 if payload.get("state_affiliated") else 0,
+        payload.get("country_code"),
         current_user["username"],
         tenant_id,
-        ),
+    ]
+    if classification_touched:
+        cols += ["classification_source", "classified_at"]
+        vals += ["manual"]
+        ts_marker = True
+    else:
+        ts_marker = False
+
+    placeholders = ", ".join(["?"] * len(vals) + (["CURRENT_TIMESTAMP"] if ts_marker else []))
+    cursor = await db.execute(
+        f"INSERT INTO sources ({', '.join(cols)}) VALUES ({placeholders})",
+        vals,
     )
+    new_id = cursor.lastrowid
+
+    if alignments:
+        await _replace_alignments(db, new_id, alignments)
+
     await db.commit()
 
-    cursor = await db.execute("SELECT * FROM sources WHERE id = ?", (cursor.lastrowid,))
+    cursor = await db.execute("SELECT * FROM sources WHERE id = ?", (new_id,))
     row = await cursor.fetchone()
-    return dict(row)
+    result = dict(row)
+    result["is_global"] = result.get("tenant_id") is None
+    result["state_affiliated"] = bool(result.get("state_affiliated"))
+    alignments_map = await _load_alignments_for(db, [new_id])
+    result["alignments"] = alignments_map.get(new_id, [])
+    return result
 
 
 @router.put("/{source_id}", response_model=SourceResponse)
@@ -494,27 +611,51 @@ async def update_source(
 
     _check_source_ownership(dict(row), current_user["username"])
 
+    payload = data.model_dump(exclude_unset=True)
+    alignments = payload.pop("alignments", None)
+
     updates = {}
-    for field, value in data.model_dump(exclude_none=True).items():
+    for field, value in payload.items():
         if field not in SOURCE_UPDATE_COLUMNS:
             continue
         # Domain normalisieren
         if field == "domain" and value:
             value = _DOMAIN_ALIASES.get(value.lower(), value.lower())
+        if field == "state_affiliated":
+            value = 1 if value else 0
         updates[field] = value
 
-    if not updates:
-        return dict(row)
+    classification_touched = bool(SOURCE_CLASSIFICATION_FIELDS & updates.keys()) or alignments is not None
+    if classification_touched:
+        updates["classification_source"] = "manual"
+        updates["classified_at"] = "CURRENT_TIMESTAMP_MARKER"
 
-    set_clause = ", ".join(f"{k} = ?" for k in updates)
-    values = list(updates.values()) + [source_id]
+    if updates:
+        set_parts = []
+        values = []
+        for k, v in updates.items():
+            if v == "CURRENT_TIMESTAMP_MARKER":
+                set_parts.append(f"{k} = CURRENT_TIMESTAMP")
+            else:
+                set_parts.append(f"{k} = ?")
+                values.append(v)
+        values.append(source_id)
+        await db.execute(f"UPDATE sources SET {', '.join(set_parts)} WHERE id = ?", values)
 
-    await db.execute(f"UPDATE sources SET {set_clause} WHERE id = ?", values)
+    if alignments is not None:
+        await _replace_alignments(db, source_id, alignments)
+
+    if updates or alignments is not None:
         await db.commit()
 
     cursor = await db.execute("SELECT * FROM sources WHERE id = ?", (source_id,))
     row = await cursor.fetchone()
-    return dict(row)
+    result = dict(row)
+    result["is_global"] = result.get("tenant_id") is None
+    result["state_affiliated"] = bool(result.get("state_affiliated"))
+    alignments_map = await _load_alignments_for(db, [source_id])
+    result["alignments"] = alignments_map.get(source_id, [])
+    return result
 
 
 @router.delete("/{source_id}", status_code=status.HTTP_204_NO_CONTENT)
@@ -572,3 +713,328 @@ async def trigger_refresh_counts(
     """Artikelzaehler fuer alle Quellen neu berechnen."""
     await refresh_source_counts(db)
     return {"status": "ok"}
+
+
+# === Klassifikations-Review (LLM-Vorschlaege approve/reject/reclassify) ===
+
+def _require_admin_for_global(row: dict, current_user: dict):
+    """Globale Quellen (tenant_id IS NULL) duerfen nur org_admins approve-en/reclassify-en."""
+    if row.get("tenant_id") is None and current_user.get("role") != "org_admin":
+        raise HTTPException(
+            status_code=status.HTTP_403_FORBIDDEN,
+            detail="Globale Quellen koennen nur von Admins klassifiziert werden",
+        )
+
+
+@router.get("/classification/stats")
+async def classification_stats(
+    current_user: dict = Depends(get_current_user),
+    db: aiosqlite.Connection = Depends(db_dependency),
+):
+    """Counts pro classification_source-Wert (global + eigene Org)."""
+    tenant_id = current_user.get("tenant_id")
+    cursor = await db.execute(
+        """SELECT classification_source, COUNT(*) as cnt
+           FROM sources
+           WHERE (tenant_id IS NULL OR tenant_id = ?) AND status = 'active'
+           GROUP BY classification_source""",
+        (tenant_id,),
+    )
+    by_source = {row["classification_source"] or "legacy": row["cnt"] for row in await cursor.fetchall()}
+    cursor = await db.execute(
+        """SELECT COUNT(*) as cnt FROM sources
+           WHERE (tenant_id IS NULL OR tenant_id = ?) AND status = 'active'
+           AND proposed_political_orientation IS NOT NULL""",
+        (tenant_id,),
+    )
+    pending = (await cursor.fetchone())["cnt"]
+    return {
+        "by_classification_source": by_source,
+        "pending_review": pending,
+        "total": sum(by_source.values()),
+    }
+
+
+@router.get("/classification/queue")
+async def classification_queue(
+    limit: int = 50,
+    min_confidence: float = 0.0,
+    current_user: dict = Depends(get_current_user),
+    db: aiosqlite.Connection = Depends(db_dependency),
+):
+    """Liefert Quellen mit nicht-leeren proposed_*-Spalten (Review-Queue)."""
+    tenant_id = current_user.get("tenant_id")
+    cursor = await db.execute(
+        """SELECT s.* FROM sources s
+           WHERE (s.tenant_id IS NULL OR s.tenant_id = ?)
+             AND s.proposed_political_orientation IS NOT NULL
+             AND COALESCE(s.proposed_confidence, 0) >= ?
+           ORDER BY s.proposed_confidence DESC, s.proposed_at DESC
+           LIMIT ?""",
+        (tenant_id, min_confidence, limit),
+    )
+    rows = [dict(r) for r in await cursor.fetchall()]
+    alignments_map = await _load_alignments_for(db, [r["id"] for r in rows])
+    out = []
+    for d in rows:
+        try:
+            proposed_aligns = json.loads(d.get("proposed_alignments_json") or "[]")
+        except (json.JSONDecodeError, TypeError):
+            proposed_aligns = []
+        out.append({
+            "id": d["id"],
+            "name": d["name"],
+            "url": d.get("url"),
+            "domain": d.get("domain"),
+            "source_type": d.get("source_type"),
+            "category": d.get("category"),
+            "is_global": d.get("tenant_id") is None,
+            "current": {
+                "political_orientation": d.get("political_orientation"),
+                "media_type": d.get("media_type"),
+                "reliability": d.get("reliability"),
+                "state_affiliated": bool(d.get("state_affiliated")),
+                "country_code": d.get("country_code"),
+                "alignments": alignments_map.get(d["id"], []),
+                "classification_source": d.get("classification_source"),
+            },
+            "proposed": {
+                "political_orientation": d.get("proposed_political_orientation"),
+                "media_type": d.get("proposed_media_type"),
+                "reliability": d.get("proposed_reliability"),
+                "state_affiliated": bool(d.get("proposed_state_affiliated")),
+                "country_code": d.get("proposed_country_code"),
+                "alignments": proposed_aligns,
+                "confidence": d.get("proposed_confidence"),
+                "reasoning": d.get("proposed_reasoning"),
+                "proposed_at": d.get("proposed_at"),
+            },
+        })
+    return out
+
+
+async def _clear_proposed(db: aiosqlite.Connection, source_id: int):
+    """Loescht die proposed_*-Felder einer Quelle (ohne commit)."""
+    await db.execute(
+        """UPDATE sources SET
+            proposed_political_orientation = NULL,
+            proposed_media_type = NULL,
+            proposed_reliability = NULL,
+            proposed_state_affiliated = NULL,
+            proposed_country_code = NULL,
+            proposed_alignments_json = NULL,
+            proposed_confidence = NULL,
+            proposed_reasoning = NULL,
+            proposed_at = NULL
+        WHERE id = ?""",
+        (source_id,),
+    )
+
+
+@router.post("/{source_id}/classification/approve")
+async def approve_classification(
+    source_id: int,
+    current_user: dict = Depends(get_current_user),
+    db: aiosqlite.Connection = Depends(db_dependency),
+):
+    """Uebernimmt proposed_* in echte Felder, setzt classification_source='llm_approved'."""
+    cursor = await db.execute("SELECT * FROM sources WHERE id = ?", (source_id,))
+    row = await cursor.fetchone()
+    if not row:
+        raise HTTPException(status_code=404, detail="Quelle nicht gefunden")
+    src = dict(row)
+    _require_admin_for_global(src, current_user)
+
+    if src.get("proposed_political_orientation") is None:
+        raise HTTPException(status_code=400, detail="Keine LLM-Vorschlaege fuer diese Quelle vorhanden")
+
+    try:
+        proposed_aligns = json.loads(src.get("proposed_alignments_json") or "[]")
+    except (json.JSONDecodeError, TypeError):
+        proposed_aligns = []
+
+    await db.execute(
+        """UPDATE sources SET
+            political_orientation = ?,
+            media_type = ?,
+            reliability = ?,
+            state_affiliated = ?,
+            country_code = ?,
+            classification_source = 'llm_approved',
+            classified_at = CURRENT_TIMESTAMP
+        WHERE id = ?""",
+        (
+            src["proposed_political_orientation"],
+            src["proposed_media_type"],
+            src["proposed_reliability"],
+            1 if src.get("proposed_state_affiliated") else 0,
+            src.get("proposed_country_code"),
+            source_id,
+        ),
+    )
+    await _replace_alignments(db, source_id, [a for a in proposed_aligns if a in ALLOWED_ALIGNMENTS])
+    await _clear_proposed(db, source_id)
+    await db.commit()
+    # Reliability-Override anwenden (IFCN/EUvsDisinfo)
+    try:
+        await apply_reputation_overrides(db, source_id)
+    except Exception as e:
+        logger.warning("Reputation-Override fuer source_id=%s fehlgeschlagen: %s", source_id, e)
+    return {"source_id": source_id, "status": "approved"}
+
+
+@router.post("/{source_id}/classification/reject")
+async def reject_classification(
+    source_id: int,
+    current_user: dict = Depends(get_current_user),
+    db: aiosqlite.Connection = Depends(db_dependency),
+):
+    """Verwirft die LLM-Vorschlaege ohne Uebernahme. classification_source bleibt unveraendert."""
+    cursor = await db.execute("SELECT * FROM sources WHERE id = ?", (source_id,))
+    row = await cursor.fetchone()
+    if not row:
+        raise HTTPException(status_code=404, detail="Quelle nicht gefunden")
+    src = dict(row)
+    _require_admin_for_global(src, current_user)
+
+    await _clear_proposed(db, source_id)
+    # Wenn classification_source noch 'llm_pending' war, zurueck auf 'legacy'
+    if src.get("classification_source") == "llm_pending":
+        await db.execute(
+            "UPDATE sources SET classification_source = 'legacy' WHERE id = ?",
+            (source_id,),
+        )
+    await db.commit()
+    return {"source_id": source_id, "status": "rejected"}
+
+
+@router.post("/{source_id}/classification/reclassify")
+async def reclassify_source(
+    source_id: int,
+    current_user: dict = Depends(get_current_user),
+    db: aiosqlite.Connection = Depends(db_dependency),
+):
+    """Triggert eine LLM-Klassifikation einer einzelnen Quelle (synchron, ~3-5s)."""
+    cursor = await db.execute("SELECT * FROM sources WHERE id = ?", (source_id,))
+    row = await cursor.fetchone()
+    if not row:
+        raise HTTPException(status_code=404, detail="Quelle nicht gefunden")
+    src = dict(row)
+    _require_admin_for_global(src, current_user)
+
+    try:
+        result = await classify_source(db, source_id)
+    except Exception as e:
+        logger.error("Reclassify source_id=%s fehlgeschlagen: %s", source_id, e, exc_info=True)
+        raise HTTPException(status_code=500, detail=f"Klassifikation fehlgeschlagen: {e}")
+    return result
+
+
+async def _bulk_classify_background(limit: int, only_unclassified: bool):
+    """Hintergrund-Task: oeffnet eigene DB-Connection."""
+    db = await get_db()
+    try:
+        await bulk_classify(db, limit=limit, only_unclassified=only_unclassified)
+    finally:
+        await db.close()
+
+
+@router.post("/classification/bulk-classify")
+async def trigger_bulk_classify(
+    background_tasks: BackgroundTasks,
+    limit: int = 50,
+    only_unclassified: bool = True,
+    current_user: dict = Depends(get_current_user),
+):
+    """Startet eine Bulk-Klassifikation im Hintergrund (nur Admins)."""
+    if current_user.get("role") != "org_admin":
+        raise HTTPException(status_code=403, detail="Nur Admins koennen Bulk-Klassifikation starten")
+    if limit < 1 or limit > 500:
+        raise HTTPException(status_code=400, detail="limit muss zwischen 1 und 500 liegen")
+    background_tasks.add_task(_bulk_classify_background, limit, only_unclassified)
+    return {"status": "started", "limit": limit, "only_unclassified": only_unclassified}
+
+
+@router.post("/external-reputation/sync")
+async def trigger_external_reputation_sync(
+    background_tasks: BackgroundTasks,
+    current_user: dict = Depends(get_current_user),
+):
+    """Startet Sync von IFCN- und EUvsDisinfo-Daten (Admin, Hintergrund)."""
+    if current_user.get("role") != "org_admin":
+        raise HTTPException(status_code=403, detail="Nur Admins koennen den externen Sync starten")
+
+    async def _bg():
+        db = await get_db()
+        try:
+            await sync_external_reputation(db)
+        finally:
+            await db.close()
+
+    background_tasks.add_task(_bg)
+    return {"status": "started"}
+
+
+@router.post("/classification/bulk-approve")
+async def bulk_approve_classifications(
+    min_confidence: float = 0.85,
+    current_user: dict = Depends(get_current_user),
+    db: aiosqlite.Connection = Depends(db_dependency),
+):
+    """Genehmigt alle Pending-Vorschlaege ueber dem confidence-Schwellwert (nur Admins).
+
+    Globale Quellen werden nur bearbeitet, wenn der Aufrufer org_admin ist;
+    Tenant-eigene Quellen sowieso.
+    """
+    if current_user.get("role") != "org_admin":
+        raise HTTPException(status_code=403, detail="Nur Admins koennen Bulk-Approve nutzen")
+    tenant_id = current_user.get("tenant_id")
+    cursor = await db.execute(
+        """SELECT id, proposed_political_orientation, proposed_media_type,
+                  proposed_reliability, proposed_state_affiliated,
+                  proposed_country_code, proposed_alignments_json, tenant_id
+           FROM sources
+           WHERE proposed_political_orientation IS NOT NULL
+             AND COALESCE(proposed_confidence, 0) >= ?
+             AND (tenant_id IS NULL OR tenant_id = ?)""",
+        (min_confidence, tenant_id),
+    )
+    rows = [dict(r) for r in await cursor.fetchall()]
+    approved_ids: list[int] = []
+    for src in rows:
+        try:
+            proposed_aligns = json.loads(src.get("proposed_alignments_json") or "[]")
+        except (json.JSONDecodeError, TypeError):
+            proposed_aligns = []
+        await db.execute(
+            """UPDATE sources SET
+                political_orientation = ?,
+                media_type = ?,
+                reliability = ?,
+                state_affiliated = ?,
+                country_code = ?,
+                classification_source = 'llm_approved',
+                classified_at = CURRENT_TIMESTAMP
+            WHERE id = ?""",
+            (
+                src["proposed_political_orientation"],
+                src["proposed_media_type"],
+                src["proposed_reliability"],
+                1 if src.get("proposed_state_affiliated") else 0,
+                src.get("proposed_country_code"),
+                src["id"],
+            ),
+        )
+        await _replace_alignments(
+            db, src["id"], [a for a in proposed_aligns if a in ALLOWED_ALIGNMENTS]
+        )
+        await _clear_proposed(db, src["id"])
+        approved_ids.append(src["id"])
+    await db.commit()
+    # Reliability-Override fuer alle gerade Approved
+    try:
+        for sid in approved_ids:
+            await apply_reputation_overrides(db, sid)
+    except Exception as e:
+        logger.warning("Bulk Reputation-Override fehlgeschlagen: %s", e)
+    return {"approved_count": len(approved_ids), "min_confidence": min_confidence}

src/services/external_reputation.py

@@ -0,0 +1,282 @@
+"""Externe Reputations-Daten fuer Quellen.
+
+Synchronisiert Domain-Listen von oeffentlichen Reputations-/Faktencheck-Datenbanken
+und schreibt die Treffer in die sources-Spalten:
+
+- IFCN-Signatories (anerkannte Faktenchecker) -> ifcn_signatory
+- EUvsDisinfo (pro-Kreml-Desinformation, Zenodo-CSV) -> eu_disinfo_listed,
+  eu_disinfo_case_count, eu_disinfo_last_seen
+
+Anschliessend wendet apply_reputation_overrides() Override-Regeln auf die
+reliability-Spalte an:
+- ifcn_signatory=1 -> reliability='sehr_hoch'
+- eu_disinfo_case_count >= 5 -> reliability='sehr_niedrig'
+- eu_disinfo_case_count >= 1 -> reliability eine Stufe runter (max bis 'niedrig')
+"""
+import csv
+import io
+import logging
+from collections import defaultdict
+from urllib.parse import urlparse
+
+import aiosqlite
+import httpx
+
+logger = logging.getLogger("osint.external_reputation")
+
+IFCN_LIST_URL = "https://raw.githubusercontent.com/IFCN/verified-signatories/main/list"
+EU_DISINFO_CSV_URL = "https://zenodo.org/records/10514307/files/euvsdisinfo_base.csv?download=1"
+
+HTTP_TIMEOUT = httpx.Timeout(60.0, connect=10.0)
+
+# Generische Plattform-Domains, die NICHT als Quelle markiert werden duerfen
+# (EUvsDisinfo aggregiert anonyme Telegram-/Twitter-Posts unter Plattform-Domains).
+PLATFORM_DOMAINS = {
+    "t.me", "telegram.me", "telegram.org",
+    "twitter.com", "x.com", "mobile.twitter.com",
+    "youtube.com", "youtu.be", "m.youtube.com",
+    "facebook.com", "fb.com", "m.facebook.com",
+    "instagram.com", "tiktok.com", "vk.com", "ok.ru",
+    "rumble.com", "bitchute.com", "odysee.com",
+    "reddit.com", "old.reddit.com",
+    "wordpress.com", "blogspot.com", "medium.com",
+    "substack.com", "wixsite.com",
+}
+
+# Reliability-Skala in Stufenfolge (schlecht -> gut)
+RELIABILITY_ORDER = ["sehr_niedrig", "niedrig", "gemischt", "hoch", "sehr_hoch"]
+
+
+def _normalize_domain(raw: str | None) -> str | None:
+    """Normalisiert eine Domain: lowercase, ohne www., ohne Schema/Pfad."""
+    if not raw:
+        return None
+    raw = raw.strip().lower()
+    if not raw:
+        return None
+    # Falls eine vollstaendige URL uebergeben wurde
+    if "://" in raw:
+        try:
+            raw = urlparse(raw).netloc or raw
+        except ValueError:
+            pass
+    # Pfad/Query strippen
+    raw = raw.split("/")[0].split("?")[0].split("#")[0]
+    if raw.startswith("www."):
+        raw = raw[4:]
+    return raw or None
+
+
+async def _fetch_text(url: str) -> str:
+    """Laedt Text von einer URL. Wirft HTTPException bei Fehler."""
+    async with httpx.AsyncClient(timeout=HTTP_TIMEOUT, follow_redirects=True) as client:
+        resp = await client.get(url)
+        resp.raise_for_status()
+        return resp.text
+
+
+async def sync_ifcn_signatories(db: aiosqlite.Connection) -> dict:
+    """Laedt IFCN-Domain-Liste und matcht gegen sources.domain.
+
+    Setzt ifcn_signatory=1 wo die Domain in der Liste vorkommt, sonst 0.
+    """
+    text = await _fetch_text(IFCN_LIST_URL)
+    domains: set[str] = set()
+    for line in text.splitlines():
+        d = _normalize_domain(line)
+        if d:
+            domains.add(d)
+    logger.info("IFCN-Liste geladen: %d Domains", len(domains))
+
+    # Aktuelle Quellen mit Domain laden
+    cursor = await db.execute(
+        "SELECT id, domain FROM sources WHERE domain IS NOT NULL AND domain != ''"
+    )
+    sources = [dict(r) for r in await cursor.fetchall()]
+
+    matched_ids: list[int] = []
+    unmatched_ids: list[int] = []
+    for s in sources:
+        nd = _normalize_domain(s["domain"])
+        if nd and nd not in PLATFORM_DOMAINS and nd in domains:
+            matched_ids.append(s["id"])
+        else:
+            unmatched_ids.append(s["id"])
+
+    # Bulk-Update in zwei Statements
+    if matched_ids:
+        placeholders = ",".join("?" for _ in matched_ids)
+        await db.execute(
+            f"UPDATE sources SET ifcn_signatory = 1 WHERE id IN ({placeholders})",
+            matched_ids,
+        )
+    if unmatched_ids:
+        placeholders = ",".join("?" for _ in unmatched_ids)
+        await db.execute(
+            f"UPDATE sources SET ifcn_signatory = 0 WHERE id IN ({placeholders})",
+            unmatched_ids,
+        )
+    await db.commit()
+    logger.info("IFCN-Sync: %d Quellen als Faktenchecker markiert (von %d)",
+                len(matched_ids), len(sources))
+    return {
+        "list_size": len(domains),
+        "sources_checked": len(sources),
+        "matched": len(matched_ids),
+    }
+
+
+async def sync_eu_disinfo(db: aiosqlite.Connection) -> dict:
+    """Laedt EUvsDisinfo-CSV von Zenodo, aggregiert pro Domain, schreibt sources.
+
+    - eu_disinfo_listed: 1 wenn Domain mindestens 1x als 'disinformation' debunkt
+    - eu_disinfo_case_count: Anzahl Disinformation-Faelle
+    - eu_disinfo_last_seen: spaetestes debunk_date
+    """
+    text = await _fetch_text(EU_DISINFO_CSV_URL)
+    reader = csv.DictReader(io.StringIO(text))
+
+    # Per-Domain aggregieren (nur class='disinformation')
+    counts: dict[str, int] = defaultdict(int)
+    last_seen: dict[str, str] = {}
+    total_rows = 0
+    for row in reader:
+        total_rows += 1
+        if (row.get("class") or "").strip().lower() != "disinformation":
+            continue
+        d = _normalize_domain(row.get("article_domain"))
+        if not d:
+            continue
+        counts[d] += 1
+        debunk_date = (row.get("debunk_date") or "").strip()
+        if debunk_date:
+            prev = last_seen.get(d)
+            if not prev or debunk_date > prev:
+                last_seen[d] = debunk_date
+    logger.info("EUvsDisinfo-CSV: %d Zeilen, %d Domains mit Desinformation",
+                total_rows, len(counts))
+
+    # Quellen laden + matchen
+    cursor = await db.execute(
+        "SELECT id, domain FROM sources WHERE domain IS NOT NULL AND domain != ''"
+    )
+    sources = [dict(r) for r in await cursor.fetchall()]
+
+    matched = 0
+    for s in sources:
+        nd = _normalize_domain(s["domain"])
+        if nd and nd not in PLATFORM_DOMAINS and nd in counts:
+            await db.execute(
+                """UPDATE sources SET
+                    eu_disinfo_listed = 1,
+                    eu_disinfo_case_count = ?,
+                    eu_disinfo_last_seen = ?
+                WHERE id = ?""",
+                (counts[nd], last_seen.get(nd), s["id"]),
+            )
+            matched += 1
+        else:
+            await db.execute(
+                """UPDATE sources SET
+                    eu_disinfo_listed = 0,
+                    eu_disinfo_case_count = 0,
+                    eu_disinfo_last_seen = NULL
+                WHERE id = ?""",
+                (s["id"],),
+            )
+    await db.commit()
+    logger.info("EUvsDisinfo-Sync: %d Quellen als Desinformations-Quelle markiert (von %d)",
+                matched, len(sources))
+    return {
+        "rows_in_csv": total_rows,
+        "domains_with_disinfo_in_csv": len(counts),
+        "sources_checked": len(sources),
+        "matched": matched,
+    }
+
+
+def _override_reliability(current: str | None, ifcn: bool, eu_count: int) -> str | None:
+    """Wendet Override-Regeln auf eine reliability-Stufe an.
+
+    Rueckgabe: neue Stufe (oder None, wenn unveraendert).
+    """
+    cur = current or "na"
+
+    # IFCN gewinnt: zertifizierter Faktenchecker -> sehr_hoch (immer)
+    if ifcn:
+        return "sehr_hoch" if cur != "sehr_hoch" else None
+
+    # EUvsDisinfo: Downgrade
+    if eu_count >= 5:
+        return "sehr_niedrig" if cur != "sehr_niedrig" else None
+    if eu_count >= 1:
+        # Eine Stufe runter, mindestens bis 'niedrig'
+        if cur == "na":
+            return "niedrig"
+        if cur in RELIABILITY_ORDER:
+            idx = RELIABILITY_ORDER.index(cur)
+            new_idx = max(0, idx - 1)
+            new = RELIABILITY_ORDER[new_idx]
+            # Mindeststufe 'niedrig' bei eu_count >= 1
+            if RELIABILITY_ORDER.index(new) > RELIABILITY_ORDER.index("niedrig"):
+                new = "niedrig"
+            return new if new != cur else None
+    return None
+
+
+async def apply_reputation_overrides(db: aiosqlite.Connection, source_id: int | None = None) -> dict:
+    """Wendet Reliability-Override-Regeln an.
+
+    Wenn source_id angegeben ist, nur fuer diese Quelle. Sonst fuer alle Quellen.
+    """
+    if source_id is not None:
+        cursor = await db.execute(
+            "SELECT id, reliability, ifcn_signatory, eu_disinfo_case_count "
+            "FROM sources WHERE id = ?",
+            (source_id,),
+        )
+    else:
+        cursor = await db.execute(
+            "SELECT id, reliability, ifcn_signatory, eu_disinfo_case_count FROM sources"
+        )
+    sources = [dict(r) for r in await cursor.fetchall()]
+
+    changed = 0
+    for s in sources:
+        new = _override_reliability(
+            s.get("reliability"),
+            bool(s.get("ifcn_signatory")),
+            int(s.get("eu_disinfo_case_count") or 0),
+        )
+        if new is not None:
+            await db.execute(
+                "UPDATE sources SET reliability = ? WHERE id = ?",
+                (new, s["id"]),
+            )
+            changed += 1
+    await db.commit()
+    logger.info("Reliability-Override: %d Quellen angepasst (von %d gepruefte)",
+                changed, len(sources))
+    return {"checked": len(sources), "changed": changed}
+
+
+async def sync_all(db: aiosqlite.Connection) -> dict:
+    """Vollstaendiger Sync: IFCN + EUvsDisinfo + Reliability-Override.
+
+    Setzt external_data_synced_at fuer alle Quellen.
+    """
+    ifcn_result = await sync_ifcn_signatories(db)
+    eu_result = await sync_eu_disinfo(db)
+    override_result = await apply_reputation_overrides(db)
+
+    await db.execute(
+        "UPDATE sources SET external_data_synced_at = CURRENT_TIMESTAMP "
+        "WHERE domain IS NOT NULL AND domain != ''"
+    )
+    await db.commit()
+
+    return {
+        "ifcn": ifcn_result,
+        "eu_disinfo": eu_result,
+        "override": override_result,
+    }

src/services/pipeline_tracker.py

@@ -228,3 +228,25 @@ async def error_step(db, ws_manager, *, step_id: Optional[int], refresh_log_id:
         "status": "error",
         "pass_number": pass_number,
     }, visibility, created_by, tenant_id)
+
+
+async def cancel_active_steps(db, *, refresh_log_id: int) -> int:
+    """Schliesst alle noch aktiven Pipeline-Schritte eines Refreshs als 'cancelled' ab.
+
+    Wird vom Orchestrator nach einem User-Cancel aufgerufen. Ohne diesen Schritt
+    bleibt der zuletzt aktive Step-Eintrag verwaist und der Pipeline-Endpoint
+    liefert dauerhaft 'Schritt X laeuft' an die UI.
+    """
+    try:
+        cur = await db.execute(
+            """UPDATE refresh_pipeline_steps
+               SET status = 'cancelled', completed_at = ?
+               WHERE refresh_log_id = ? AND status = 'active'""",
+            (_now_db(), refresh_log_id),
+        )
+        await db.commit()
+        return cur.rowcount or 0
+    except Exception as e:
+        logger.warning(f"Pipeline cancel_active_steps DB-Fehler: {e}")
+        return 0
+

src/services/post_refresh_qc.py

@@ -400,18 +400,20 @@ async def run_post_refresh_qc(db, incident_id: int) -> dict:
             db, incident_id, incident_title, incident_desc
         )
         umlauts_fixed = await normalize_umlaut_fields(db, incident_id)
+        article_umlauts_fixed = await normalize_umlaut_articles(db, incident_id)
 
-        if facts_removed > 0 or locations_fixed > 0 or umlauts_fixed > 0:
+        total_umlaut_changes = umlauts_fixed + article_umlauts_fixed
+        if facts_removed > 0 or locations_fixed > 0 or total_umlaut_changes > 0:
             await db.commit()
             logger.info(
-                "Post-Refresh QC fuer Incident %d: %d Duplikate entfernt, %d Locations korrigiert, %d Umlaute normalisiert",
-                incident_id, facts_removed, locations_fixed, umlauts_fixed,
+                "Post-Refresh QC fuer Incident %d: %d Duplikate entfernt, %d Locations korrigiert, %d Umlaute normalisiert (davon %d in Articles)",
+                incident_id, facts_removed, locations_fixed, total_umlaut_changes, article_umlauts_fixed,
             )
 
         return {
             "facts_removed": facts_removed,
             "locations_fixed": locations_fixed,
-            "umlauts_fixed": umlauts_fixed,
+            "umlauts_fixed": total_umlaut_changes,
         }
 
     except Exception as e:
@@ -568,3 +570,64 @@ async def normalize_umlaut_fields(db, incident_id: int) -> int:
         incident_id, count_summary, count_dev,
     )
     return total
+
+
+async def normalize_umlaut_articles(db, incident_id: int) -> int:
+    """Normalisiert Umlaute in allen Artikel-Texten des Incidents.
+
+    Felder die behandelt werden:
+    - headline_de und content_de bei allen Artikeln (LLM-Uebersetzung kann
+      ASCII-Umlaute liefern trotz Prompt-Anweisung)
+    - headline und content_original bei language='de' (manche Quellen wie
+      dpa-AFX, Telegram-Kanaele liefern selbst schon ASCII-Umlaute)
+
+    Idempotent: Wenn der Text schon korrekt ist, macht das Dict-Lookup
+    keine Aenderung und wir schreiben nicht zurueck.
+
+    Rueckgabe: Gesamtzahl der Wort-Ersetzungen ueber alle Artikel.
+    """
+    cursor = await db.execute(
+        """SELECT id, language, headline, headline_de, content_original, content_de
+           FROM articles WHERE incident_id = ?""",
+        (incident_id,),
+    )
+    rows = await cursor.fetchall()
+    if not rows:
+        return 0
+
+    total = 0
+    for row in rows:
+        is_de = (row["language"] or "").lower() == "de"
+        updates = {}
+
+        # Felder die immer behandelt werden (LLM-Uebersetzungen)
+        if row["headline_de"]:
+            new, n = normalize_german_umlauts(row["headline_de"])
+            if n > 0:
+                updates["headline_de"] = new
+                total += n
+        if row["content_de"]:
+            new, n = normalize_german_umlauts(row["content_de"])
+            if n > 0:
+                updates["content_de"] = new
+                total += n
+
+        # Originalfelder nur bei deutschen Quellen
+        if is_de:
+            if row["headline"]:
+                new, n = normalize_german_umlauts(row["headline"])
+                if n > 0:
+                    updates["headline"] = new
+                    total += n
+            if row["content_original"]:
+                new, n = normalize_german_umlauts(row["content_original"])
+                if n > 0:
+                    updates["content_original"] = new
+                    total += n
+
+        if updates:
+            set_clause = ", ".join(f"{k} = ?" for k in updates)
+            values = list(updates.values()) + [row["id"]]
+            await db.execute(f"UPDATE articles SET {set_clause} WHERE id = ?", values)
+
+    return total

src/services/source_classifier.py

@@ -0,0 +1,295 @@
+"""Klassifiziert Quellen via Claude (Haiku) nach 4 Achsen + state_affiliated + country.
+
+Schreibt Vorschlaege in die proposed_*-Spalten von sources und setzt
+classification_source='llm_pending'. Approval erfolgt ueber separate Endpoints,
+die proposed_* in die echten Spalten kopieren.
+"""
+import asyncio
+import json
+import logging
+import re
+
+import aiosqlite
+
+from agents.claude_client import call_claude
+from config import CLAUDE_MODEL_FAST
+
+logger = logging.getLogger("osint.source_classifier")
+
+POLITICAL_VALUES = {
+    "links_extrem", "links", "mitte_links", "liberal", "mitte",
+    "konservativ", "mitte_rechts", "rechts", "rechts_extrem", "na",
+}
+MEDIA_TYPE_VALUES = {
+    "tageszeitung", "wochenzeitung", "magazin", "tv_sender", "radio",
+    "oeffentlich_rechtlich", "nachrichtenagentur", "online_only", "blog",
+    "telegram_kanal", "telegram_bot", "podcast", "social_media", "imageboard",
+    "think_tank", "ngo", "behoerde", "staatsmedium", "fachmedium", "sonstige",
+}
+RELIABILITY_VALUES = {"sehr_hoch", "hoch", "gemischt", "niedrig", "sehr_niedrig", "na"}
+ALIGNMENT_VALUES = {
+    "prorussisch", "proiranisch", "prowestlich", "proukrainisch",
+    "prochinesisch", "projapanisch", "proisraelisch", "propalaestinensisch",
+    "protuerkisch", "panarabisch", "neutral", "sonstige",
+}
+
+
+def _build_prompt(src: dict, sample_articles: list[dict]) -> str:
+    sample_text = ""
+    if sample_articles:
+        lines = []
+        for i, art in enumerate(sample_articles[:5], 1):
+            headline = (art.get("headline") or art.get("headline_de") or "").strip()
+            if headline:
+                lines.append(f"{i}. {headline[:200]}")
+        if lines:
+            sample_text = "\nLetzte Artikel/Headlines:\n" + "\n".join(lines)
+
+    return f"""Du bist ein OSINT-Analyst und klassifizierst Nachrichten- und Medienquellen fuer ein Lagebild-Monitoring-System (DACH-Raum).
+
+QUELLE:
+Name: {src.get('name')}
+URL: {src.get('url') or '-'}
+Domain: {src.get('domain') or '-'}
+Quellentyp: {src.get('source_type')}
+Bisherige Kategorie: {src.get('category')}
+Sprache: {src.get('language') or 'unbekannt'}
+Bisherige Notiz (Freitext): {src.get('bias') or '-'}{sample_text}
+
+AUFGABE: Klassifiziere die Quelle nach folgenden Achsen.
+
+1. political_orientation:
+   - links_extrem (z.B. linksunten.indymedia)
+   - links (klar links, z.B. junge Welt, taz)
+   - mitte_links (linksliberal/sozialdemokratisch, z.B. SZ, Spiegel)
+   - liberal (wirtschafts-/grünliberal, z.B. NZZ, Zeit)
+   - mitte (politisch neutral, Agentur, z.B. dpa, Reuters, tagesschau)
+   - konservativ (buergerlich-konservativ, z.B. FAZ, Welt)
+   - mitte_rechts (rechts-buergerlich, z.B. Tichys Einblick, Achgut)
+   - rechts (klar rechts, z.B. Junge Freiheit, EpochTimes)
+   - rechts_extrem (z.B. Compact, PI-News)
+   - na (nicht klassifizierbar: Behoerde, Fachmedium, Think Tank ohne klare politische Linie)
+
+2. media_type (genau einer):
+   tageszeitung, wochenzeitung, magazin, tv_sender, radio, oeffentlich_rechtlich,
+   nachrichtenagentur, online_only, blog, telegram_kanal, telegram_bot, podcast,
+   social_media, imageboard, think_tank, ngo, behoerde, staatsmedium, fachmedium, sonstige
+
+3. reliability:
+   - sehr_hoch (etablierte Qualitaet, Faktencheck: tagesschau, dpa, FAZ, Reuters)
+   - hoch (serioes mit gelegentlichen Schwaechen: taz, Welt, BILD bei harten News)
+   - gemischt (Mix Meinung/Einseitigkeit: Tichys Einblick, Achgut, Boulevard)
+   - niedrig (haeufig irrefuehrend, schwache Quellenarbeit: Junge Freiheit, EpochTimes)
+   - sehr_niedrig (bekannt fuer Desinformation/Verschwoerung: Compact, RT, Sputnik, PI-News)
+   - na (nicht bewertbar)
+
+4. alignments (Mehrfach, leeres Array wenn keine ausgepraegte Naehe):
+   prorussisch, proiranisch, prowestlich, proukrainisch, prochinesisch, projapanisch,
+   proisraelisch, propalaestinensisch, protuerkisch, panarabisch, neutral, sonstige
+
+5. state_affiliated (true/false): true wenn vom Staat finanziert/kontrolliert
+   (RT, Sputnik, CGTN, PressTV, Xinhua, TRT). Public Service Broadcaster
+   wie ARD/ZDF/BBC sind NICHT state_affiliated.
+
+6. country_code (ISO 3166-1 alpha-2): Heimatland (DE, AT, CH, RU, US, ...). null wenn unklar.
+
+7. confidence (0.0-1.0): 0.85+ fuer bekannte Outlets, 0.5-0.85 fuer mittelbekannt, <0.5 fuer unsicher.
+
+8. reasoning (1-2 Saetze): Kurze Begruendung der Hauptklassifikationen.
+
+WICHTIG:
+- Antworte AUSSCHLIESSLICH mit einem JSON-Objekt, kein Text drumherum.
+- Nutze ausschliesslich die genannten enum-Werte (snake_case).
+- Bei Unklarheit lieber `na` und niedrige confidence.
+
+JSON-Schema:
+{{
+  "political_orientation": "...",
+  "media_type": "...",
+  "reliability": "...",
+  "alignments": ["..."],
+  "state_affiliated": false,
+  "country_code": "DE",
+  "confidence": 0.9,
+  "reasoning": "..."
+}}"""
+
+
+async def _load_sample_articles(db: aiosqlite.Connection, name: str, domain: str | None, limit: int = 5) -> list[dict]:
+    """Laedt die letzten Headlines einer Quelle (per name oder Domain-Match)."""
+    rows: list = []
+    if name:
+        cursor = await db.execute(
+            "SELECT headline, headline_de FROM articles WHERE source = ? ORDER BY collected_at DESC LIMIT ?",
+            (name, limit),
+        )
+        rows = await cursor.fetchall()
+    if not rows and domain:
+        cursor = await db.execute(
+            "SELECT headline, headline_de FROM articles WHERE source_url LIKE ? ORDER BY collected_at DESC LIMIT ?",
+            (f"%{domain}%", limit),
+        )
+        rows = await cursor.fetchall()
+    return [dict(r) for r in rows]
+
+
+def _validate(parsed: dict) -> dict:
+    """Validiert + normalisiert eine LLM-Antwort gegen die Enums."""
+    pol = parsed.get("political_orientation", "na")
+    if pol not in POLITICAL_VALUES:
+        pol = "na"
+    mt = parsed.get("media_type", "sonstige")
+    if mt not in MEDIA_TYPE_VALUES:
+        mt = "sonstige"
+    rel = parsed.get("reliability", "na")
+    if rel not in RELIABILITY_VALUES:
+        rel = "na"
+    aligns_raw = parsed.get("alignments") or []
+    if not isinstance(aligns_raw, list):
+        aligns_raw = []
+    aligns = sorted({a for a in aligns_raw if isinstance(a, str) and a in ALIGNMENT_VALUES})
+    sa = bool(parsed.get("state_affiliated", False))
+    cc = parsed.get("country_code")
+    if isinstance(cc, str) and len(cc) == 2 and cc.isalpha():
+        cc = cc.upper()
+    else:
+        cc = None
+    try:
+        confidence = float(parsed.get("confidence", 0.5))
+        confidence = max(0.0, min(1.0, confidence))
+    except (TypeError, ValueError):
+        confidence = 0.5
+    reasoning = str(parsed.get("reasoning", ""))[:1000]
+    return {
+        "political_orientation": pol,
+        "media_type": mt,
+        "reliability": rel,
+        "alignments": aligns,
+        "state_affiliated": sa,
+        "country_code": cc,
+        "confidence": confidence,
+        "reasoning": reasoning,
+    }
+
+
+async def classify_source(
+    db: aiosqlite.Connection,
+    source_id: int,
+    sample_limit: int = 5,
+    model: str = CLAUDE_MODEL_FAST,
+) -> dict:
+    """Klassifiziert eine einzelne Quelle und schreibt die Vorschlaege in proposed_*-Spalten."""
+    cursor = await db.execute(
+        "SELECT id, name, url, domain, source_type, category, language, bias, "
+        "classification_source FROM sources WHERE id = ?",
+        (source_id,),
+    )
+    row = await cursor.fetchone()
+    if not row:
+        raise ValueError(f"Quelle {source_id} nicht gefunden")
+    src = dict(row)
+
+    sample = await _load_sample_articles(db, src["name"], src.get("domain"), sample_limit)
+    prompt = _build_prompt(src, sample)
+    response, usage = await call_claude(prompt, tools=None, model=model)
+
+    json_match = re.search(r"\{.*\}", response, re.DOTALL)
+    if not json_match:
+        raise ValueError(f"Keine JSON-Antwort von Claude fuer source_id={source_id}: {response[:200]}")
+    parsed = json.loads(json_match.group(0))
+    result = _validate(parsed)
+
+    # Nur classification_source auf 'llm_pending' setzen, wenn nicht bereits manuell/approved
+    new_src = "CASE WHEN classification_source IN ('manual','llm_approved') THEN classification_source ELSE 'llm_pending' END"
+    await db.execute(
+        f"""UPDATE sources SET
+            proposed_political_orientation = ?,
+            proposed_media_type = ?,
+            proposed_reliability = ?,
+            proposed_state_affiliated = ?,
+            proposed_country_code = ?,
+            proposed_alignments_json = ?,
+            proposed_confidence = ?,
+            proposed_reasoning = ?,
+            proposed_at = CURRENT_TIMESTAMP,
+            classification_source = {new_src}
+        WHERE id = ?""",
+        (
+            result["political_orientation"],
+            result["media_type"],
+            result["reliability"],
+            1 if result["state_affiliated"] else 0,
+            result["country_code"],
+            json.dumps(result["alignments"], ensure_ascii=False),
+            result["confidence"],
+            result["reasoning"],
+            source_id,
+        ),
+    )
+    await db.commit()
+
+    logger.info(
+        "Klassifiziert source_id=%s '%s' -> %s/%s/%s conf=%.2f ($%.4f)",
+        source_id, src["name"], result["political_orientation"],
+        result["media_type"], result["reliability"], result["confidence"],
+        usage.cost_usd,
+    )
+
+    result["source_id"] = source_id
+    result["usage"] = {
+        "cost_usd": usage.cost_usd,
+        "input_tokens": usage.input_tokens,
+        "output_tokens": usage.output_tokens,
+    }
+    return result
+
+
+async def bulk_classify(
+    db: aiosqlite.Connection,
+    limit: int = 50,
+    only_unclassified: bool = True,
+    model: str = CLAUDE_MODEL_FAST,
+) -> dict:
+    """Klassifiziert noch unklassifizierte Quellen (sequenziell).
+
+    Args:
+        limit: Maximale Anzahl Quellen pro Aufruf
+        only_unclassified: Wenn True, nur classification_source='legacy'.
+                           Wenn False, auch 'llm_pending' neu klassifizieren.
+    """
+    if only_unclassified:
+        where = "classification_source = 'legacy'"
+    else:
+        where = "classification_source IN ('legacy', 'llm_pending')"
+    cursor = await db.execute(
+        f"SELECT id FROM sources WHERE {where} AND status = 'active' "
+        f"AND source_type != 'excluded' ORDER BY id LIMIT ?",
+        (limit,),
+    )
+    ids = [row["id"] for row in await cursor.fetchall()]
+
+    total_cost = 0.0
+    success = 0
+    errors: list[dict] = []
+
+    for sid in ids:
+        try:
+            r = await classify_source(db, sid, model=model)
+            total_cost += r["usage"]["cost_usd"]
+            success += 1
+        except asyncio.CancelledError:
+            raise
+        except Exception as e:
+            logger.error("Klassifikation source_id=%s fehlgeschlagen: %s", sid, e, exc_info=True)
+            errors.append({"source_id": sid, "error": str(e)})
+
+    logger.info(
+        "Bulk-Klassifikation fertig: %d/%d erfolgreich, $%.4f Kosten, %d Fehler",
+        success, len(ids), total_cost, len(errors),
+    )
+    return {
+        "processed": len(ids),
+        "success": success,
+        "errors": errors,
+        "total_cost_usd": total_cost,
+    }

src/services/source_health.py

@@ -1,41 +1,69 @@
-"""Quellen-Health-Check Engine - prüft Erreichbarkeit, Feed-Validität, Duplikate."""
+"""Quellen-Health-Check Engine - prüft Erreichbarkeit, Feed-Validität, Duplikate."""
 import asyncio
 import logging
 import json
+import uuid
 from urllib.parse import urlparse
 
 import httpx
 import feedparser
 import aiosqlite
 
+try:
+    from config import HEALTH_CHECK_USER_AGENT, HEALTH_CHECK_TIMEOUT_S
+except ImportError:
+    HEALTH_CHECK_USER_AGENT = "Mozilla/5.0 (compatible; AegisSight-HealthCheck/1.0)"
+    HEALTH_CHECK_TIMEOUT_S = 15.0
+
+# Phase 18: alternative User-Agents fuer Bot-Block-Bypass
+USER_AGENT_GOOGLEBOT = "Mozilla/5.0 (compatible; Googlebot/2.1; +http://www.google.com/bot.html)"
+USER_AGENT_BROWSER = (
+    "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 "
+    "(KHTML, like Gecko) Chrome/120.0 Safari/537.36"
+)
+REMOVEPAYWALLS_PREFIX = "https://www.removepaywall.com/search?url="
+
+# HTTP-Codes, die einen Retry mit anderem UA rechtfertigen
+RETRY_ON_STATUS = {403, 406, 429}
+
 logger = logging.getLogger("osint.source_health")
 
 
 async def run_health_checks(db: aiosqlite.Connection) -> dict:
-    """Führt alle Health-Checks für aktive Grundquellen durch."""
+    """Führt Health-Checks für alle aktiven Quellen durch (global + Tenant)."""
     logger.info("Starte Quellen-Health-Check...")
 
-    # Alle aktiven Grundquellen laden
+    # Alle aktiven Quellen laden (global UND Tenant-spezifisch)
     cursor = await db.execute(
-        "SELECT id, name, url, domain, source_type, article_count, last_seen_at "
-        "FROM sources WHERE status = 'active' AND tenant_id IS NULL"
+        "SELECT id, name, url, domain, source_type, article_count, last_seen_at, "
+        "COALESCE(fetch_strategy, 'default') AS fetch_strategy "
+        "FROM sources WHERE status = 'active' "
     )
     sources = [dict(row) for row in await cursor.fetchall()]
 
-    # Aktuelle Health-Check-Ergebnisse löschen (werden neu geschrieben)
+    # Bisherigen Stand in History archivieren, dann frisch starten
+    run_id = uuid.uuid4().hex[:12]
+    await db.execute(
+        "INSERT INTO source_health_history "
+        "(run_id, source_id, check_type, status, message, details, checked_at) "
+        "SELECT ?, source_id, check_type, status, message, details, checked_at "
+        "FROM source_health_checks",
+        (run_id,),
+    )
     await db.execute("DELETE FROM source_health_checks")
     await db.commit()
+    logger.info(f"Health-Check Run {run_id}: vorigen Stand archiviert")
 
     checks_done = 0
     issues_found = 0
 
-    # 1. Erreichbarkeit + Feed-Validität (nur Quellen mit URL)
+    # 1. Erreichbarkeit + Feed-Validität (nur Quellen mit URL)
     sources_with_url = [s for s in sources if s["url"]]
 
     async with httpx.AsyncClient(
-        timeout=15.0,
+        timeout=HEALTH_CHECK_TIMEOUT_S,
         follow_redirects=True,
-        headers={"User-Agent": "Mozilla/5.0 (compatible; OSINT-Monitor/1.0)"},
+        headers={"User-Agent": HEALTH_CHECK_USER_AGENT},
     ) as client:
         for i in range(0, len(sources_with_url), 5):
             batch = sources_with_url[i:i + 5]
@@ -46,7 +74,7 @@ async def run_health_checks(db: aiosqlite.Connection) -> dict:
                 if isinstance(result, Exception):
                     await _save_check(
                         db, source["id"], "reachability", "error",
-                        f"Prüfung fehlgeschlagen: {result}",
+                        f"Prüfung fehlgeschlagen: {result}",
                     )
                     issues_found += 1
                 else:
@@ -83,7 +111,7 @@ async def run_health_checks(db: aiosqlite.Connection) -> dict:
 
     await db.commit()
     logger.info(
-        f"Health-Check abgeschlossen: {checks_done} Quellen geprüft, "
+        f"Health-Check abgeschlossen: {checks_done} Quellen geprüft, "
         f"{issues_found} Probleme gefunden"
     )
     return {"checked": checks_done, "issues": issues_found}
@@ -92,12 +120,63 @@ async def run_health_checks(db: aiosqlite.Connection) -> dict:
 async def _check_source_reachability(
     client: httpx.AsyncClient, source: dict,
 ) -> list[dict]:
-    """Prüft Erreichbarkeit und Feed-Validität einer Quelle."""
+    """Prüft Erreichbarkeit und Feed-Validität einer Quelle.
+
+    Phase 18: pro Quelle eine fetch_strategy ('default' | 'googlebot' | 'paywall' | 'skip').
+    Bei 'default' wird im Fehlerfall (403/406/429) ein Retry mit Googlebot-UA gemacht.
+    Bei 'paywall' wird auf removepaywall.com umgeleitet.
+    Bei 'skip' wird kein Check ausgeführt.
+    """
     checks = []
     url = source["url"]
+    strategy = source.get("fetch_strategy") or "default"
+
+    # 'skip' -> kein Check (bekannte unerreichbare Quellen, z.B. Login-only)
+    if strategy == "skip":
+        checks.append({
+            "type": "reachability", "status": "ok",
+            "message": "Health-Check uebersprungen (fetch_strategy=skip)",
+        })
+        return checks
+
+    # URL-Schema sicherstellen
+    if url and not url.startswith(("http://", "https://")):
+        url = "https://" + url.lstrip("/")
+
+    # Initialen UA waehlen
+    initial_ua = HEALTH_CHECK_USER_AGENT
+    initial_url = url
+    if strategy == "googlebot":
+        initial_ua = USER_AGENT_GOOGLEBOT
+    elif strategy == "paywall":
+        # Paywall-Quellen: Feed-URL direkt laden, aber mit Browser-UA (versucht Bot-Detection zu umgehen).
+        # removepaywall.com ist fuer Article-URLs, NICHT fuer RSS-Feed-Validity-Checks
+        # (gibt HTML statt XML zurueck). Researcher-Pipeline nutzt removepaywall fuer Inhalte.
+        initial_ua = USER_AGENT_BROWSER
 
     try:
-        resp = await client.get(url)
+        resp = await client.get(initial_url, headers={"User-Agent": initial_ua})
+
+        # Paywall-Quellen: 4xx ist erwartbar (Bot-Detection), als warning markieren statt error
+        if strategy == "paywall" and resp.status_code in RETRY_ON_STATUS:
+            checks.append({
+                "type": "reachability", "status": "warning",
+                "message": f"Paywall-Quelle, Direkt-Zugang HTTP {resp.status_code} (Researcher-Pipeline nutzt removepaywall.com fuer Inhalte)",
+            })
+            return checks  # Feed-Validity-Check skippen (Paywall liefert kein RSS)
+
+        # Bot-Block-Retry nur bei strategy='default'
+        if (
+            strategy == "default"
+            and resp.status_code in RETRY_ON_STATUS
+        ):
+            retry = await client.get(url, headers={"User-Agent": USER_AGENT_GOOGLEBOT})
+            if retry.status_code < 400:
+                resp = retry  # Retry hat geholfen
+                checks.append({
+                    "type": "reachability", "status": "warning",
+                    "message": f"Erreichbar nur mit Googlebot-UA (Standard-UA bekam HTTP {initial_url and 'unknown' or 'XXX'})",
+                })
 
         if resp.status_code >= 400:
             checks.append({
@@ -125,14 +204,14 @@ async def _check_source_reachability(
                 "message": "Erreichbar",
             })
 
-        # Feed-Validität nur für RSS-Feeds
+        # Feed-Validität nur für RSS-Feeds
         if source["source_type"] == "rss_feed":
             text = resp.text[:20000]
             if "<rss" not in text and "<feed" not in text and "<channel" not in text:
                 checks.append({
                     "type": "feed_validity",
                     "status": "error",
-                    "message": "Kein gültiger RSS/Atom-Feed",
+                    "message": "Kein gültiger RSS/Atom-Feed",
                 })
             else:
                 feed = await asyncio.to_thread(feedparser.parse, text)
@@ -155,7 +234,7 @@ async def _check_source_reachability(
                     checks.append({
                         "type": "feed_validity",
                         "status": "ok",
-                        "message": f"Feed gültig ({len(feed.entries)} Einträge)",
+                        "message": f"Feed gültig ({len(feed.entries)} Einträge)",
                     })
 
     except httpx.TimeoutException:
@@ -181,7 +260,7 @@ async def _check_source_reachability(
 
 
 def _check_stale(source: dict) -> dict | None:
-    """Prüft ob eine Quelle veraltet ist (keine Artikel seit >30 Tagen)."""
+    """Prüft ob eine Quelle veraltet ist (keine Artikel seit >30 Tagen)."""
     if source["source_type"] == "excluded":
         return None
 
@@ -249,7 +328,7 @@ async def _save_check(
 
 
 async def get_health_summary(db: aiosqlite.Connection) -> dict:
-    """Gibt eine Zusammenfassung der letzten Health-Check-Ergebnisse zurück."""
+    """Gibt eine Zusammenfassung der letzten Health-Check-Ergebnisse zurück."""
     cursor = await db.execute("""
         SELECT
             h.id, h.source_id, s.name, s.domain, s.url, s.source_type,

src/services/source_suggester.py

@@ -1,4 +1,4 @@
-"""KI-gestützte Quellen-Vorschläge via Haiku."""
+"""KI-gestützte Quellen-Vorschläge via Haiku."""
 import json
 import logging
 import re
@@ -12,8 +12,8 @@ logger = logging.getLogger("osint.source_suggester")
 
 
 async def generate_suggestions(db: aiosqlite.Connection) -> int:
-    """Generiert Quellen-Vorschläge basierend auf Health-Checks und Lückenanalyse."""
-    logger.info("Starte Quellen-Vorschläge via Haiku...")
+    """Generiert Quellen-Vorschläge basierend auf Health-Checks und Lückenanalyse."""
+    logger.info("Starte Quellen-Vorschläge via Haiku...")
 
     # 1. Aktuelle Quellen laden
     cursor = await db.execute(
@@ -33,13 +33,13 @@ async def generate_suggestions(db: aiosqlite.Connection) -> int:
     """)
     issues = [dict(row) for row in await cursor.fetchall()]
 
-    # 3. Alte pending-Vorschläge entfernen (älter als 30 Tage)
+    # 3. Alte pending-Vorschläge entfernen (älter als 30 Tage)
     await db.execute(
         "DELETE FROM source_suggestions "
         "WHERE status = 'pending' AND created_at < datetime('now', '-30 days')"
     )
 
-    # 4. Quellen-Zusammenfassung für Haiku
+    # 4. Quellen-Zusammenfassung für Haiku
     categories = {}
     for s in sources:
         cat = s["category"]
@@ -67,7 +67,7 @@ async def generate_suggestions(db: aiosqlite.Connection) -> int:
                 f"{issue['check_type']} = {issue['status']} - {issue['message']}\n"
             )
 
-    prompt = f"""Du bist ein OSINT-Analyst und verwaltest die Quellensammlung eines Lagebildmonitors für Sicherheitsbehörden.
+    prompt = f"""Du bist ein OSINT-Analyst und verwaltest die Quellensammlung eines Lagebildmonitors für Sicherheitsbehörden.
 
 Aktuelle Quellensammlung:{source_summary}{issues_summary}
 
@@ -78,13 +78,13 @@ Beachte:
 2. Fehlende wichtige OSINT-Quellen: Schlage "add_source" mit konkreter RSS-Feed-URL vor
 3. Fokus auf deutschsprachige + wichtige internationale Nachrichtenquellen
 4. Nur Quellen vorschlagen, die NICHT bereits vorhanden sind
-5. Maximal 5 Vorschläge
+5. Maximal 5 Vorschläge
 
 Antworte NUR mit einem JSON-Array. Jedes Element:
 {{
   "type": "add_source|deactivate_source|fix_url|remove_source",
   "title": "Kurzer Titel",
-  "description": "Begründung",
+  "description": "Begründung",
   "priority": "low|medium|high",
   "source_id": null,
   "data": {{
@@ -104,7 +104,7 @@ Nur das JSON-Array, kein anderer Text."""
 
         json_match = re.search(r'\[.*\]', response, re.DOTALL)
         if not json_match:
-            logger.warning("Keine Vorschläge von Haiku erhalten (kein JSON)")
+            logger.warning("Keine Vorschläge von Haiku erhalten (kein JSON)")
             return 0
 
         suggestions = json.loads(json_match.group(0))
@@ -164,14 +164,14 @@ Nur das JSON-Array, kein anderer Text."""
 
         await db.commit()
         logger.info(
-            f"Quellen-Vorschläge: {count} neue Vorschläge generiert "
+            f"Quellen-Vorschläge: {count} neue Vorschläge generiert "
             f"(Haiku: {usage.input_tokens} in / {usage.output_tokens} out / "
             f"${usage.cost_usd:.4f})"
         )
         return count
 
     except Exception as e:
-        logger.error(f"Fehler bei Quellen-Vorschlägen: {e}", exc_info=True)
+        logger.error(f"Fehler bei Quellen-Vorschlägen: {e}", exc_info=True)
         return 0
 
 
@@ -218,7 +218,7 @@ async def apply_suggestion(
                     (url,),
                 )
                 if await cursor.fetchone():
-                    result["action"] = "übersprungen (URL bereits vorhanden)"
+                    result["action"] = "übersprungen (URL bereits vorhanden)"
                     new_status = "rejected"
                 else:
                     await db.execute(
@@ -230,7 +230,7 @@ async def apply_suggestion(
                     )
                     result["action"] = f"Quelle '{name}' angelegt"
             else:
-                result["action"] = "übersprungen (keine URL)"
+                result["action"] = "übersprungen (keine URL)"
                 new_status = "rejected"
 
         elif stype == "deactivate_source":
@@ -242,7 +242,7 @@ async def apply_suggestion(
                 )
                 result["action"] = "Quelle deaktiviert"
             else:
-                result["action"] = "übersprungen (keine source_id)"
+                result["action"] = "übersprungen (keine source_id)"
 
         elif stype == "remove_source":
             source_id = suggestion["source_id"]
@@ -250,9 +250,9 @@ async def apply_suggestion(
                 await db.execute(
                     "DELETE FROM sources WHERE id = ?", (source_id,),
                 )
-                result["action"] = "Quelle gelöscht"
+                result["action"] = "Quelle gelöscht"
             else:
-                result["action"] = "übersprungen (keine source_id)"
+                result["action"] = "übersprungen (keine source_id)"
 
         elif stype == "fix_url":
             source_id = suggestion["source_id"]
@@ -264,7 +264,7 @@ async def apply_suggestion(
                 )
                 result["action"] = f"URL aktualisiert auf {new_url}"
             else:
-                result["action"] = "übersprungen (keine source_id oder URL)"
+                result["action"] = "übersprungen (keine source_id oder URL)"
 
     await db.execute(
         "UPDATE source_suggestions SET status = ?, reviewed_at = CURRENT_TIMESTAMP "

src/static/css/style.css

@@ -549,6 +549,31 @@ a:hover {
     font-weight: 500;
 }
 
+.header-dropdown-action {
+    display: flex;
+    align-items: center;
+    gap: 8px;
+    width: 100%;
+    background: transparent;
+    border: 0;
+    padding: 8px 12px;
+    color: var(--text-secondary);
+    font-size: 12px;
+    font-family: inherit;
+    cursor: pointer;
+    border-radius: 6px;
+    text-align: left;
+    transition: background 0.15s ease, color 0.15s ease;
+}
+.header-dropdown-action:hover {
+    background: var(--bg-hover, rgba(255, 255, 255, 0.04));
+    color: var(--text-primary);
+}
+.header-dropdown-action svg {
+    flex-shrink: 0;
+    color: var(--accent);
+}
+
 .header-license-badge {
     display: inline-block;
     font-size: 10px;
@@ -3478,6 +3503,341 @@ a.dev-source-pill:hover {
     color: var(--info);
 }
 
+/* Sources-Modal: Tabs */
+.sources-tabs {
+    display: flex;
+    gap: 2px;
+    border-bottom: 1px solid var(--border-color, rgba(0,0,0,0.1));
+    margin-bottom: 12px;
+}
+.sources-tab {
+    background: transparent;
+    border: none;
+    padding: 8px 16px;
+    font-size: 13px;
+    font-weight: 500;
+    color: var(--text-secondary, #555);
+    cursor: pointer;
+    border-bottom: 2px solid transparent;
+    margin-bottom: -1px;
+    display: inline-flex;
+    align-items: center;
+    gap: 8px;
+}
+.sources-tab:hover {
+    color: var(--text-primary, #222);
+}
+.sources-tab.active {
+    color: var(--primary, #2a81cb);
+    border-bottom-color: var(--primary, #2a81cb);
+}
+.sources-tab-badge {
+    display: inline-flex;
+    align-items: center;
+    justify-content: center;
+    min-width: 20px;
+    padding: 0 6px;
+    height: 18px;
+    border-radius: 9px;
+    background: var(--primary, #2a81cb);
+    color: #fff;
+    font-size: 10px;
+    font-weight: 700;
+}
+
+/* Review-Queue */
+.review-toolbar {
+    display: flex;
+    align-items: center;
+    justify-content: space-between;
+    padding: 8px 12px;
+    background: var(--cat-sonstige-bg, #f6f6fa);
+    border-radius: var(--radius);
+    margin-bottom: 12px;
+    flex-wrap: wrap;
+    gap: 12px;
+}
+.review-toolbar-info {
+    display: flex;
+    align-items: center;
+    gap: 16px;
+    font-size: 13px;
+}
+.review-conf-filter {
+    display: inline-flex;
+    align-items: center;
+    gap: 6px;
+    font-size: 12px;
+    color: var(--text-secondary, #555);
+}
+.review-conf-filter select {
+    padding: 2px 6px;
+    font-size: 12px;
+    border-radius: var(--radius);
+    border: 1px solid var(--border-color, rgba(0,0,0,0.15));
+}
+.review-toolbar-actions {
+    display: flex;
+    gap: 6px;
+}
+
+.review-list {
+    display: flex;
+    flex-direction: column;
+    gap: 8px;
+}
+.review-card {
+    background: var(--surface, #fff);
+    border: 1px solid var(--border-color, rgba(0,0,0,0.08));
+    border-radius: var(--radius);
+    padding: 12px 14px;
+}
+.review-card-header {
+    display: flex;
+    justify-content: space-between;
+    align-items: flex-start;
+    gap: 12px;
+    margin-bottom: 10px;
+}
+.review-card-title {
+    display: flex;
+    flex-wrap: wrap;
+    align-items: center;
+    gap: 8px;
+}
+.review-card-name {
+    font-weight: 600;
+    font-size: 14px;
+}
+.review-card-domain {
+    font-size: 11px;
+    color: var(--text-disabled, #888);
+}
+.review-global-badge {
+    display: inline-flex;
+    align-items: center;
+    padding: 1px 6px;
+    border-radius: var(--radius);
+    background: #5e35b1;
+    color: #fff;
+    font-size: 9px;
+    font-weight: 600;
+    letter-spacing: 0.3px;
+    text-transform: uppercase;
+}
+.review-card-confidence {
+    display: inline-flex;
+    flex-direction: column;
+    align-items: center;
+    padding: 4px 10px;
+    border-radius: var(--radius);
+    min-width: 60px;
+}
+.review-card-confidence .conf-value {
+    font-size: 14px;
+    font-weight: 700;
+}
+.review-card-confidence .conf-label {
+    font-size: 9px;
+    text-transform: uppercase;
+    letter-spacing: 0.3px;
+    opacity: 0.8;
+}
+.review-card-confidence.conf-high { background: #e8f5e9; color: #2e7d32; }
+.review-card-confidence.conf-medium { background: #fff8e1; color: #ef6c00; }
+.review-card-confidence.conf-low { background: #ffebee; color: #c62828; }
+
+.review-card-diff {
+    display: grid;
+    grid-template-columns: 1fr;
+    gap: 4px;
+    font-size: 12px;
+    margin-bottom: 10px;
+}
+.review-diff-row {
+    display: grid;
+    grid-template-columns: 110px 1fr 24px 1fr;
+    align-items: center;
+    gap: 8px;
+    padding: 3px 6px;
+    border-radius: 3px;
+}
+.review-diff-row.changed {
+    background: #fff8e1;
+}
+.review-diff-label {
+    color: var(--text-secondary, #555);
+    font-weight: 500;
+}
+.review-diff-current {
+    color: var(--text-disabled, #888);
+}
+.review-diff-arrow {
+    text-align: center;
+    color: var(--text-disabled, #888);
+    font-weight: 600;
+}
+.review-diff-proposed {
+    color: var(--text-primary, #222);
+    font-weight: 500;
+}
+.review-diff-row.changed .review-diff-proposed {
+    color: #ef6c00;
+    font-weight: 600;
+}
+
+.review-card-reasoning {
+    font-size: 12px;
+    color: var(--text-secondary, #555);
+    background: var(--cat-sonstige-bg, #f6f6fa);
+    padding: 8px 10px;
+    border-radius: var(--radius);
+    margin-bottom: 10px;
+    line-height: 1.5;
+}
+.review-card-actions {
+    display: flex;
+    gap: 6px;
+    flex-wrap: wrap;
+}
+
+/* Klassifikations-Badges (politisch / reliability / alignments / state) */
+.source-classification-badges {
+    display: inline-flex;
+    align-items: center;
+    gap: 4px;
+    flex-wrap: wrap;
+}
+
+.source-political-badge {
+    display: inline-flex;
+    align-items: center;
+    justify-content: center;
+    min-width: 22px;
+    padding: 2px 6px;
+    border-radius: var(--radius);
+    font-size: 10px;
+    font-weight: 700;
+    letter-spacing: 0.4px;
+    color: #fff;
+    background: #9e9e9e;
+}
+.source-political-badge.pol-links_extrem { background: #b71c1c; }
+.source-political-badge.pol-links { background: #e53935; }
+.source-political-badge.pol-mitte_links { background: #ef9a9a; color: #4a0d0d; }
+.source-political-badge.pol-liberal { background: #fdd835; color: #4a3700; }
+.source-political-badge.pol-mitte { background: #9e9e9e; }
+.source-political-badge.pol-konservativ { background: #90caf9; color: #0d2740; }
+.source-political-badge.pol-mitte_rechts { background: #5c6bc0; }
+.source-political-badge.pol-rechts { background: #1976d2; }
+.source-political-badge.pol-rechts_extrem { background: #0d47a1; }
+
+.source-reliability-dot {
+    display: inline-block;
+    width: 10px;
+    height: 10px;
+    border-radius: 50%;
+    background: #9e9e9e;
+    border: 1px solid rgba(0, 0, 0, 0.15);
+}
+.source-reliability-dot.rel-sehr_hoch { background: #2e7d32; }
+.source-reliability-dot.rel-hoch { background: #66bb6a; }
+.source-reliability-dot.rel-gemischt { background: #fbc02d; }
+.source-reliability-dot.rel-niedrig { background: #ef6c00; }
+.source-reliability-dot.rel-sehr_niedrig { background: #c62828; }
+
+.source-state-badge {
+    display: inline-flex;
+    align-items: center;
+    justify-content: center;
+    width: 18px;
+    height: 18px;
+    border-radius: 50%;
+    background: #4a148c;
+    color: #fff;
+    font-size: 11px;
+    line-height: 1;
+}
+
+.source-ifcn-badge {
+    display: inline-flex;
+    align-items: center;
+    padding: 1px 6px;
+    border-radius: var(--radius);
+    background: #e8f5e9;
+    color: #1b5e20;
+    border: 1px solid #66bb6a;
+    font-size: 10px;
+    font-weight: 600;
+    letter-spacing: 0.3px;
+}
+
+.source-eu-disinfo-badge {
+    display: inline-flex;
+    align-items: center;
+    padding: 1px 6px;
+    border-radius: var(--radius);
+    background: #ffebee;
+    color: #b71c1c;
+    border: 1px solid #c62828;
+    font-size: 10px;
+    font-weight: 600;
+    letter-spacing: 0.3px;
+}
+
+.source-alignment-chip-badge {
+    display: inline-flex;
+    align-items: center;
+    padding: 1px 6px;
+    border-radius: 999px;
+    font-size: 10px;
+    font-weight: 500;
+    background: var(--cat-sonstige-bg, #eef);
+    color: var(--text-secondary, #555);
+    border: 1px solid rgba(0, 0, 0, 0.08);
+}
+
+/* Edit-Form: Klassifikations-Sektion */
+.sources-classification-section {
+    margin-top: 12px;
+    padding-top: 12px;
+    border-top: 1px solid var(--border-color, rgba(0,0,0,0.08));
+}
+.sources-classification-header {
+    font-size: 12px;
+    font-weight: 600;
+    color: var(--text-secondary, #555);
+    margin-bottom: 8px;
+    letter-spacing: 0.3px;
+    text-transform: uppercase;
+}
+.alignment-chips {
+    display: flex;
+    flex-wrap: wrap;
+    gap: 6px;
+}
+.alignment-chip {
+    display: inline-flex;
+    align-items: center;
+    padding: 4px 10px;
+    border-radius: 999px;
+    font-size: 11px;
+    font-weight: 500;
+    background: transparent;
+    color: var(--text-secondary, #555);
+    border: 1px solid var(--border-color, rgba(0,0,0,0.15));
+    cursor: pointer;
+    transition: all 0.12s ease;
+}
+.alignment-chip:hover {
+    background: var(--cat-sonstige-bg, #eef);
+}
+.alignment-chip.active {
+    background: var(--primary, #2a81cb);
+    color: #fff;
+    border-color: var(--primary, #2a81cb);
+}
+
 /* Typ-Badges */
 .source-type-badge {
     display: inline-flex;

src/static/dashboard.html

@@ -72,6 +72,11 @@
                                 <span class="credits-percent" id="credits-percent"></span>
                             </div>
                         </div>
+                        <div class="credits-divider"></div>
+                        <button class="header-dropdown-action" type="button" onclick="AIDisclaimer && AIDisclaimer.show()">
+                            <svg xmlns="http://www.w3.org/2000/svg" width="14" height="14" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" aria-hidden="true"><circle cx="12" cy="12" r="10"/><path d="M12 16v-4"/><path d="M12 8h.01"/></svg>
+                            <span>Über KI-Inhalte</span>
+                        </button>
                     </div>
                 </div>
                 <div class="header-license-warning" id="header-license-warning"></div>
@@ -118,8 +123,14 @@
                 <div id="archived-incidents" aria-live="polite" style="display:none;"></div>
             </div>
             <div class="sidebar-sources-link">
-                <button class="btn btn-secondary btn-full btn-small" onclick="App.openSourceManagement()">Quellen verwalten</button>
-                <button class="btn btn-secondary btn-full btn-small sidebar-feedback-btn" onclick="App.openFeedback()">Feedback senden</button>
+                <button class="btn btn-secondary btn-full btn-small" onclick="App.openSourceManagement()" title="Quellen verwalten">
+                    <svg xmlns="http://www.w3.org/2000/svg" width="14" height="14" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" aria-hidden="true"><ellipse cx="12" cy="5" rx="9" ry="3"/><path d="M3 5v14c0 1.66 4.03 3 9 3s9-1.34 9-3V5"/><path d="M3 12c0 1.66 4.03 3 9 3s9-1.34 9-3"/></svg>
+                    <span>Quellen</span>
+                </button>
+                <button class="btn btn-secondary btn-full btn-small sidebar-feedback-btn" onclick="App.openFeedback()" title="Feedback senden">
+                    <svg xmlns="http://www.w3.org/2000/svg" width="14" height="14" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" aria-hidden="true"><rect width="20" height="16" x="2" y="4" rx="2"/><path d="m22 7-10 5L2 7"/></svg>
+                    <span>Feedback</span>
+                </button>
                 <!-- Tutorial-Einstieg temporaer deaktiviert (Ueberarbeitung) - reaktivieren durch Entfernen der Kommentarzeichen:
                 <button class="btn btn-secondary btn-full btn-small" onclick="Tutorial.start()" title="Interaktiven Rundgang starten">Rundgang starten</button>
                 -->
@@ -351,9 +362,9 @@
                         <label>Quellen</label>
                         <div class="toggle-group">
                             <label class="toggle-label">
-                                <input type="checkbox" id="inc-international" checked>
+                                <input type="checkbox" id="inc-international">
                                 <span class="toggle-switch"></span>
-                                <span class="toggle-text">Internationale Quellen einbeziehen <span class="info-icon tooltip-below" data-tooltip="Aktiviert: Sucht auch in englischsprachigen und internationalen Medien.&#10;&#10;Deaktiviert: Nur deutschsprachige Quellen."><svg xmlns="http://www.w3.org/2000/svg" width="14" height="14" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round"><circle cx="12" cy="12" r="10"/><path d="M12 16v-4"/><path d="M12 8h.01"/></svg></span></span>
+                                <span class="toggle-text">Internationale Quellen einbeziehen <span class="info-icon tooltip-below" data-tooltip="Aktiviert: Sucht auch in englischsprachigen und internationalen Medien.&#10;&#10;Deaktiviert (Standard): Nur deutschsprachige Quellen - empfohlen für DACH-Lagen."><svg xmlns="http://www.w3.org/2000/svg" width="14" height="14" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round"><circle cx="12" cy="12" r="10"/><path d="M12 16v-4"/><path d="M12 8h.01"/></svg></span></span>
                             </label>
                         </div>
                         <div class="toggle-group" style="margin-top: 8px;">
@@ -445,6 +456,15 @@
                 <!-- Stats-Leiste -->
                 <div class="sources-stats-bar" id="sources-stats-bar"></div>
 
+                <!-- Tabs: Liste vs. Klassifikations-Review -->
+                <div class="sources-tabs" role="tablist">
+                    <button type="button" class="sources-tab active" id="sources-tab-list" role="tab" aria-selected="true" onclick="App.switchSourcesTab('list')">Quellenliste</button>
+                    <button type="button" class="sources-tab" id="sources-tab-review" role="tab" aria-selected="false" onclick="App.switchSourcesTab('review')" style="display:none;">Klassifikations-Review <span id="sources-review-count" class="sources-tab-badge">0</span></button>
+                </div>
+
+                <!-- View: Quellenliste -->
+                <div id="sources-list-view">
+
                 <!-- Toolbar -->
                 <div class="sources-toolbar">
                     <div class="sources-filters">
@@ -470,6 +490,76 @@
                             <option value="boulevard">Boulevard</option>
                             <option value="sonstige">Sonstige</option>
                         </select>
+                        <label for="sources-filter-political" class="sr-only">Politische Ausrichtung filtern</label>
+                        <select id="sources-filter-political" class="timeline-filter-select" onchange="App.filterSources()">
+                            <option value="">Alle Ausrichtungen</option>
+                            <option value="links_extrem">Links (extrem)</option>
+                            <option value="links">Links</option>
+                            <option value="mitte_links">Mitte-Links</option>
+                            <option value="liberal">Liberal</option>
+                            <option value="mitte">Mitte</option>
+                            <option value="konservativ">Konservativ</option>
+                            <option value="mitte_rechts">Mitte-Rechts</option>
+                            <option value="rechts">Rechts</option>
+                            <option value="rechts_extrem">Rechts (extrem)</option>
+                            <option value="na">Nicht eingeordnet</option>
+                        </select>
+                        <label for="sources-filter-mediatype" class="sr-only">Medientyp filtern</label>
+                        <select id="sources-filter-mediatype" class="timeline-filter-select" onchange="App.filterSources()">
+                            <option value="">Alle Medientypen</option>
+                            <option value="tageszeitung">Tageszeitung</option>
+                            <option value="wochenzeitung">Wochenzeitung</option>
+                            <option value="magazin">Magazin</option>
+                            <option value="tv_sender">TV-Sender</option>
+                            <option value="radio">Radio</option>
+                            <option value="oeffentlich_rechtlich">Öffentlich-Rechtlich</option>
+                            <option value="nachrichtenagentur">Nachrichtenagentur</option>
+                            <option value="online_only">Online-only</option>
+                            <option value="blog">Blog</option>
+                            <option value="telegram_kanal">Telegram-Kanal</option>
+                            <option value="telegram_bot">Telegram-Bot</option>
+                            <option value="podcast">Podcast</option>
+                            <option value="social_media">Social Media</option>
+                            <option value="imageboard">Imageboard</option>
+                            <option value="think_tank">Think Tank</option>
+                            <option value="ngo">NGO</option>
+                            <option value="behoerde">Behörde</option>
+                            <option value="staatsmedium">Staatsmedium</option>
+                            <option value="fachmedium">Fachmedium</option>
+                            <option value="sonstige">Sonstige</option>
+                        </select>
+                        <label for="sources-filter-reliability" class="sr-only">Glaubwürdigkeit filtern</label>
+                        <select id="sources-filter-reliability" class="timeline-filter-select" onchange="App.filterSources()">
+                            <option value="">Alle Glaubwürdigkeiten</option>
+                            <option value="sehr_hoch">Sehr hoch</option>
+                            <option value="hoch">Hoch</option>
+                            <option value="gemischt">Gemischt</option>
+                            <option value="niedrig">Niedrig</option>
+                            <option value="sehr_niedrig">Sehr niedrig</option>
+                            <option value="na">Nicht eingeordnet</option>
+                        </select>
+                        <label for="sources-filter-extern" class="sr-only">Externe Reputation filtern</label>
+                        <select id="sources-filter-extern" class="timeline-filter-select" onchange="App.filterSources()">
+                            <option value="">Externe Reputation: alle</option>
+                            <option value="ifcn">IFCN-Faktenchecker</option>
+                            <option value="eu_disinfo">EU-Desinfo gelistet</option>
+                        </select>
+                        <label for="sources-filter-alignment" class="sr-only">Geopolitische Nähe filtern</label>
+                        <select id="sources-filter-alignment" class="timeline-filter-select" onchange="App.filterSources()">
+                            <option value="">Alle Nähen</option>
+                            <option value="prorussisch">Prorussisch</option>
+                            <option value="proiranisch">Proiranisch</option>
+                            <option value="prowestlich">Prowestlich</option>
+                            <option value="proukrainisch">Proukrainisch</option>
+                            <option value="prochinesisch">Prochinesisch</option>
+                            <option value="projapanisch">Projapanisch</option>
+                            <option value="proisraelisch">Proisraelisch</option>
+                            <option value="propalaestinensisch">Propalästinensisch</option>
+                            <option value="protuerkisch">Protürkisch</option>
+                            <option value="panarabisch">Panarabisch</option>
+                            <option value="neutral">Neutral</option>
+                            <option value="sonstige">Sonstige</option>
+                        </select>
                         <label for="sources-search" class="sr-only">Quellen durchsuchen</label>
                         <input type="text" id="sources-search" class="timeline-filter-input sources-search-input" placeholder="Suche..." oninput="App.filterSources()">
                     </div>
@@ -537,6 +627,89 @@
                                 <input type="text" id="src-notes" placeholder="Optional">
                             </div>
                         </div>
+                        <div class="sources-classification-section">
+                            <div class="sources-classification-header">Einordnung</div>
+                            <div class="sources-add-form-grid">
+                                <div class="form-group">
+                                    <label for="src-political">Politische Ausrichtung</label>
+                                    <select id="src-political">
+                                        <option value="na">Nicht eingeordnet</option>
+                                        <option value="links_extrem">Links (extrem)</option>
+                                        <option value="links">Links</option>
+                                        <option value="mitte_links">Mitte-Links</option>
+                                        <option value="liberal">Liberal</option>
+                                        <option value="mitte">Mitte</option>
+                                        <option value="konservativ">Konservativ</option>
+                                        <option value="mitte_rechts">Mitte-Rechts</option>
+                                        <option value="rechts">Rechts</option>
+                                        <option value="rechts_extrem">Rechts (extrem)</option>
+                                    </select>
+                                </div>
+                                <div class="form-group">
+                                    <label for="src-mediatype">Medientyp</label>
+                                    <select id="src-mediatype">
+                                        <option value="sonstige">Sonstige</option>
+                                        <option value="tageszeitung">Tageszeitung</option>
+                                        <option value="wochenzeitung">Wochenzeitung</option>
+                                        <option value="magazin">Magazin</option>
+                                        <option value="tv_sender">TV-Sender</option>
+                                        <option value="radio">Radio</option>
+                                        <option value="oeffentlich_rechtlich">Öffentlich-Rechtlich</option>
+                                        <option value="nachrichtenagentur">Nachrichtenagentur</option>
+                                        <option value="online_only">Online-only</option>
+                                        <option value="blog">Blog</option>
+                                        <option value="telegram_kanal">Telegram-Kanal</option>
+                                        <option value="telegram_bot">Telegram-Bot</option>
+                                        <option value="podcast">Podcast</option>
+                                        <option value="social_media">Social Media</option>
+                                        <option value="imageboard">Imageboard</option>
+                                        <option value="think_tank">Think Tank</option>
+                                        <option value="ngo">NGO</option>
+                                        <option value="behoerde">Behörde</option>
+                                        <option value="staatsmedium">Staatsmedium</option>
+                                        <option value="fachmedium">Fachmedium</option>
+                                    </select>
+                                </div>
+                                <div class="form-group">
+                                    <label for="src-reliability">Glaubwürdigkeit</label>
+                                    <select id="src-reliability">
+                                        <option value="na">Nicht eingeordnet</option>
+                                        <option value="sehr_hoch">Sehr hoch</option>
+                                        <option value="hoch">Hoch</option>
+                                        <option value="gemischt">Gemischt</option>
+                                        <option value="niedrig">Niedrig</option>
+                                        <option value="sehr_niedrig">Sehr niedrig</option>
+                                    </select>
+                                </div>
+                                <div class="form-group">
+                                    <label for="src-country">Land (ISO 3166)</label>
+                                    <input type="text" id="src-country" maxlength="2" placeholder="z.B. DE, RU, US" style="text-transform:uppercase;">
+                                </div>
+                                <div class="form-group">
+                                    <label class="checkbox-label" style="display:flex;align-items:center;gap:8px;">
+                                        <input type="checkbox" id="src-state-affiliated">
+                                        <span>Staatsnah/-kontrolliert</span>
+                                    </label>
+                                </div>
+                            </div>
+                            <div class="form-group" style="margin-top:8px;">
+                                <label>Geopolitische Nähe (Mehrfachauswahl)</label>
+                                <div id="src-alignments-chips" class="alignment-chips" onclick="App.handleAlignmentChipClick(event)">
+                                    <button type="button" class="alignment-chip" data-alignment="prorussisch">prorussisch</button>
+                                    <button type="button" class="alignment-chip" data-alignment="proiranisch">proiranisch</button>
+                                    <button type="button" class="alignment-chip" data-alignment="prowestlich">prowestlich</button>
+                                    <button type="button" class="alignment-chip" data-alignment="proukrainisch">proukrainisch</button>
+                                    <button type="button" class="alignment-chip" data-alignment="prochinesisch">prochinesisch</button>
+                                    <button type="button" class="alignment-chip" data-alignment="projapanisch">projapanisch</button>
+                                    <button type="button" class="alignment-chip" data-alignment="proisraelisch">proisraelisch</button>
+                                    <button type="button" class="alignment-chip" data-alignment="propalaestinensisch">propalästinensisch</button>
+                                    <button type="button" class="alignment-chip" data-alignment="protuerkisch">protürkisch</button>
+                                    <button type="button" class="alignment-chip" data-alignment="panarabisch">panarabisch</button>
+                                    <button type="button" class="alignment-chip" data-alignment="neutral">neutral</button>
+                                    <button type="button" class="alignment-chip" data-alignment="sonstige">sonstige</button>
+                                </div>
+                            </div>
+                        </div>
                         <div class="sources-discovery-actions">
                             <button class="btn btn-primary btn-small" onclick="App.saveSource()">Speichern</button>
                             <button class="btn btn-secondary btn-small" onclick="App.toggleSourceForm(false)">Abbrechen</button>
@@ -548,6 +721,36 @@
                 <div class="sources-list" id="sources-list">
                     <div class="empty-state-text" style="padding:var(--sp-3xl);text-align:center;">Lade Quellen...</div>
                 </div>
+
+                </div>
+                <!-- /sources-list-view -->
+
+                <!-- View: Klassifikations-Review (Admin-only) -->
+                <div id="sources-review-view" style="display:none;">
+                    <div class="review-toolbar">
+                        <div class="review-toolbar-info">
+                            <span><strong id="review-pending-count">0</strong> Vorschlaege ausstehend</span>
+                            <label class="review-conf-filter">
+                                Mindest-Konfidenz:
+                                <select id="review-min-confidence" onchange="App.loadClassificationQueue()">
+                                    <option value="0">alle</option>
+                                    <option value="0.5">0.5+</option>
+                                    <option value="0.7">0.7+</option>
+                                    <option value="0.85">0.85+</option>
+                                    <option value="0.9">0.9+</option>
+                                </select>
+                            </label>
+                        </div>
+                        <div class="review-toolbar-actions">
+                            <button class="btn btn-small btn-secondary" onclick="App.triggerExternalReputationSync()" title="IFCN-Faktenchecker-Liste und EUvsDisinfo-Daten synchronisieren">Externe Daten syncen</button>
+                            <button class="btn btn-small btn-secondary" onclick="App.triggerBulkClassify()" title="LLM-Klassifikation fuer noch unklassifizierte Quellen starten">+ Klassifikation starten</button>
+                            <button class="btn btn-small btn-primary" onclick="App.bulkApproveHighConfidence()" title="Alle Vorschlaege ueber dem Konfidenz-Schwellwert genehmigen">Alle &ge; 0.85 genehmigen</button>
+                        </div>
+                    </div>
+                    <div class="review-list" id="sources-review-list">
+                        <div class="empty-state-text" style="padding:var(--sp-3xl);text-align:center;">Lade Review-Queue...</div>
+                    </div>
+                </div>
             </div>
         </div>
     </div>
@@ -738,5 +941,6 @@
     </div>
 
     <script src="/static/js/update-system.js"></script>
+    <script src="/static/js/ai-disclaimer.js"></script>
 </body>
 </html>

src/static/js/ai-disclaimer.js

@@ -0,0 +1,195 @@
+/**
+ * AI-Hallucination-Disclaimer fuer den AegisSight Monitor.
+ *
+ * Zeigt:
+ *   1) Beim ersten Besuch (oder bei neuem v-Bump) ein Modal mit Hinweisen
+ *      zur Fehlbarkeit von KI-Modellen.
+ *   2) Im Header-User-Dropdown immer einen Eintrag "Ueber KI-Inhalte",
+ *      ueber den der User das Modal jederzeit erneut oeffnen kann.
+ *
+ * Persistenz:
+ *   localStorage 'aegis_ai_disclaimer_seen' -> Versionsstring (z.B. "v1").
+ *   Wenn die Version sich aendert (Wortlaut-Update), erscheint das Modal
+ *   beim naechsten Login erneut.
+ */
+(function () {
+    'use strict';
+
+    const STORAGE_KEY = 'aegis_ai_disclaimer_seen';
+    const CURRENT_VERSION = 'v1';
+
+    // ---- DOM-Helpers (analog zu update-system.js) ----
+    function el(tag, attrs, ...children) {
+        const e = document.createElement(tag);
+        for (const k in (attrs || {})) {
+            if (k === 'class') e.className = attrs[k];
+            else if (k === 'html') e.innerHTML = attrs[k];
+            else if (k.startsWith('on')) e.addEventListener(k.slice(2), attrs[k]);
+            else e.setAttribute(k, attrs[k]);
+        }
+        for (const c of children) {
+            if (c == null) continue;
+            e.appendChild(typeof c === 'string' ? document.createTextNode(c) : c);
+        }
+        return e;
+    }
+
+    function injectStyles() {
+        if (document.getElementById('aegis-aidisc-styles')) return;
+        const css = `
+        #aegis-aidisc-overlay {
+            position: fixed; inset: 0; background: rgba(0,0,0,0.55); z-index: 99998;
+            backdrop-filter: blur(3px);
+            display: flex; align-items: center; justify-content: center; padding: 24px;
+            animation: aegis-aidisc-fade 0.25s ease;
+        }
+        @keyframes aegis-aidisc-fade { from { opacity: 0; } to { opacity: 1; } }
+        #aegis-aidisc-modal {
+            background: var(--bg-card);
+            color: var(--text-primary);
+            border-radius: 14px;
+            border: 1px solid var(--border);
+            box-shadow: 0 24px 80px rgba(0,0,0,0.4);
+            font-family: 'Inter', -apple-system, sans-serif;
+            max-width: 580px; width: 100%; max-height: 85vh; overflow: hidden;
+            display: flex; flex-direction: column;
+        }
+        #aegis-aidisc-modal header {
+            padding: 22px 28px 18px; border-bottom: 1px solid var(--border);
+            display: flex; align-items: center; gap: 12px;
+        }
+        #aegis-aidisc-modal header svg { color: var(--accent); flex-shrink: 0; }
+        #aegis-aidisc-modal h2 { margin: 0; color: var(--accent); font-size: 1.25rem; font-weight: 700; }
+        #aegis-aidisc-modal .body { padding: 18px 28px; overflow-y: auto; line-height: 1.55; }
+        #aegis-aidisc-modal .body p { margin: 0 0 12px; color: var(--text-primary); font-size: 0.94rem; }
+        #aegis-aidisc-modal .body strong { color: var(--accent); }
+        #aegis-aidisc-modal .body ul { margin: 8px 0 14px; padding-left: 22px; }
+        #aegis-aidisc-modal .body li { margin-bottom: 6px; color: var(--text-secondary); font-size: 0.92rem; }
+        #aegis-aidisc-modal .footnote {
+            margin-top: 10px; padding-top: 12px; border-top: 1px solid var(--border);
+            color: var(--text-tertiary); font-size: 0.82rem;
+        }
+        #aegis-aidisc-modal footer {
+            padding: 14px 28px 20px; border-top: 1px solid var(--border);
+            display: flex; justify-content: flex-end; gap: 10px;
+        }
+        #aegis-aidisc-modal footer button {
+            background: var(--accent); color: #fff; border: 0; padding: 10px 22px;
+            border-radius: 6px; font: inherit; font-size: 0.92rem; font-weight: 600;
+            cursor: pointer;
+        }
+        #aegis-aidisc-modal footer button:hover { background: var(--accent-hover); }
+        #aegis-aidisc-modal footer button.secondary {
+            background: transparent; color: var(--text-secondary); border: 1px solid var(--border);
+        }
+        #aegis-aidisc-modal footer button.secondary:hover {
+            background: var(--bg-hover, rgba(255,255,255,0.04)); color: var(--text-primary);
+        }`;
+        document.head.appendChild(el('style', { id: 'aegis-aidisc-styles', html: css }));
+    }
+
+    // ---- Modal-Aufbau ----
+    function buildModal(opts) {
+        const isFromUser = !!(opts && opts.fromUserAction);
+
+        // Lucide info-Icon (gleiches Pattern wie .info-icon im Repo)
+        const headerIcon = el('span', {
+            html: '<svg xmlns="http://www.w3.org/2000/svg" width="22" height="22" '
+                + 'viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" '
+                + 'stroke-linecap="round" stroke-linejoin="round">'
+                + '<circle cx="12" cy="12" r="10"/>'
+                + '<path d="M12 16v-4"/><path d="M12 8h.01"/></svg>'
+        });
+
+        const body = el('div', { class: 'body' });
+        body.appendChild(el('p', null,
+            'Der AegisSight Monitor nutzt Künstliche Intelligenz '
+            + 'zur Analyse, Übersetzung und Zusammenfassung von Nachrichten.'));
+
+        const warn = el('p');
+        warn.innerHTML = '<strong>KI-Modelle können Fehler machen</strong> '
+            + '(sogenannte „Halluzinationen"): erfundene Details, falsche Verbindungen oder '
+            + 'ungenaue Zusammenfassungen sind möglich, auch wenn der Text plausibel klingt.';
+        body.appendChild(warn);
+
+        body.appendChild(el('p', null, 'Wir empfehlen daher:'));
+        body.appendChild(el('ul', null,
+            el('li', null, 'Wichtige Informationen mit den verlinkten Quellen verifizieren'),
+            el('li', null, 'Bei kritischen Entscheidungen die Originalartikel prüfen'),
+            el('li', null, 'Faktenchecks als Hinweis verstehen, nicht als endgültige Wahrheit')
+        ));
+
+        body.appendChild(el('p', { class: 'footnote' },
+            'Diesen Hinweis findest du jederzeit wieder im Menü oben rechts unter „Über KI-Inhalte".'));
+
+        const closeAndStore = () => {
+            try { localStorage.setItem(STORAGE_KEY, CURRENT_VERSION); } catch (e) {}
+            overlay.remove();
+            document.removeEventListener('keydown', escHandler);
+        };
+        const closeOnly = () => {
+            overlay.remove();
+            document.removeEventListener('keydown', escHandler);
+        };
+
+        const footer = el('footer', null);
+        if (!isFromUser) {
+            footer.appendChild(el('button', { class: 'secondary', onclick: closeOnly }, 'Später nochmal'));
+        }
+        footer.appendChild(el('button', { onclick: closeAndStore }, 'Verstanden'));
+
+        const overlay = el('div', { id: 'aegis-aidisc-overlay' },
+            el('div', { id: 'aegis-aidisc-modal' },
+                el('header', null, headerIcon, el('h2', null, 'Hinweis zu KI-generierten Inhalten')),
+                body,
+                footer
+            )
+        );
+
+        function escHandler(ev) {
+            if (ev.key === 'Escape' && document.getElementById('aegis-aidisc-overlay')) {
+                // ESC = wie "Verstanden" beim erstmaligen Anzeigen, sonst nur schliessen
+                if (isFromUser) closeOnly(); else closeAndStore();
+            }
+        }
+        overlay.addEventListener('click', (ev) => {
+            if (ev.target === overlay) {
+                if (isFromUser) closeOnly(); else closeAndStore();
+            }
+        });
+        document.addEventListener('keydown', escHandler);
+
+        return overlay;
+    }
+
+    function show(opts) {
+        if (document.getElementById('aegis-aidisc-overlay')) return;
+        injectStyles();
+        document.body.appendChild(buildModal(opts));
+    }
+
+    function init() {
+        // Nur auf der Dashboard-Seite zeigen, nicht auf der Login-Seite
+        if (!document.body || document.body.classList.contains('login-page')) return;
+
+        injectStyles();
+        let seenVersion = '';
+        try { seenVersion = localStorage.getItem(STORAGE_KEY) || ''; } catch (e) {}
+        if (seenVersion !== CURRENT_VERSION) {
+            // Etwas verzoegern, damit Hauptdashboard sichtbar ist bevor Modal kommt
+            setTimeout(() => show({ fromUserAction: false }), 600);
+        }
+    }
+
+    // Globaler Zugriff zum manuellen Oeffnen aus dem Header-Dropdown
+    window.AIDisclaimer = {
+        show: () => show({ fromUserAction: true }),
+        VERSION: CURRENT_VERSION,
+    };
+
+    if (document.readyState === 'loading') {
+        document.addEventListener('DOMContentLoaded', init);
+    } else {
+        init();
+    }
+})();

src/static/js/api.js

@@ -198,10 +198,46 @@ const API = {
         if (params.source_type) query.set('source_type', params.source_type);
         if (params.category) query.set('category', params.category);
         if (params.source_status) query.set('source_status', params.source_status);
+        if (params.political_orientation) query.set('political_orientation', params.political_orientation);
+        if (params.media_type) query.set('media_type', params.media_type);
+        if (params.reliability) query.set('reliability', params.reliability);
+        if (params.alignment) query.set('alignment', params.alignment);
+        if (params.state_affiliated !== undefined && params.state_affiliated !== null) {
+            query.set('state_affiliated', String(params.state_affiliated));
+        }
         const qs = query.toString();
         return this._request('GET', `/sources${qs ? '?' + qs : ''}`);
     },
 
+    // Sources: Klassifikations-Review (LLM)
+    getClassificationStats() {
+        return this._request('GET', '/sources/classification/stats');
+    },
+    getClassificationQueue(limit = 50, minConfidence = 0.0) {
+        const qs = new URLSearchParams({ limit: String(limit), min_confidence: String(minConfidence) }).toString();
+        return this._request('GET', `/sources/classification/queue?${qs}`);
+    },
+    approveClassification(id) {
+        return this._request('POST', `/sources/${id}/classification/approve`);
+    },
+    rejectClassification(id) {
+        return this._request('POST', `/sources/${id}/classification/reject`);
+    },
+    reclassifySource(id) {
+        return this._request('POST', `/sources/${id}/classification/reclassify`);
+    },
+    triggerBulkClassify(limit = 50, onlyUnclassified = true) {
+        const qs = new URLSearchParams({ limit: String(limit), only_unclassified: String(onlyUnclassified) }).toString();
+        return this._request('POST', `/sources/classification/bulk-classify?${qs}`);
+    },
+    bulkApproveClassifications(minConfidence = 0.85) {
+        const qs = new URLSearchParams({ min_confidence: String(minConfidence) }).toString();
+        return this._request('POST', `/sources/classification/bulk-approve?${qs}`);
+    },
+    triggerExternalReputationSync() {
+        return this._request('POST', '/sources/external-reputation/sync');
+    },
+
     createSource(data) {
         return this._request('POST', '/sources', data);
     },

src/static/js/app.js

@@ -523,7 +523,7 @@ const App = {
                     let text = 'Nur Lesezugriff';
                     const reason = user.read_only_reason;
                     if (reason === 'budget_exceeded') {
-                        text = 'Token-Budget aufgebraucht – nur Lesezugriff. Bitte Verwaltung kontaktieren.';
+                        text = 'Token-Budget aufgebraucht – nur Lesezugriff. Für Aufstockung oder Upgrade bitte info@aegis-sight.de kontaktieren.';
                     } else if (reason === 'expired') {
                         text = 'Lizenz abgelaufen – nur Lesezugriff';
                     } else if (reason === 'no_license') {
@@ -618,6 +618,10 @@ const App = {
                     const inc = this.incidents.find(i => i.id === id);
                     const isFirst = inc && !inc.has_summary;
                     UI.showProgress('queued', { queue_position: idx + 1 }, id, isFirst);
+                    // Pipeline-Reset auch nach F5: aktive Lage in Queue -> Icons grau
+                    if (id === this.currentIncidentId && typeof Pipeline !== 'undefined' && Pipeline.beginQueue) {
+                        Pipeline.beginQueue(id);
+                    }
                 });
             }
 
@@ -1926,6 +1930,11 @@ async handleRefresh() {
             this._updateRefreshButton(true);
             // showProgress called via handleStatusUpdate
             const result = await API.refreshIncident(this.currentIncidentId);
+            // Pipeline auf "pending" setzen, damit alte gruene Haekchen nicht
+            // faelschlich "schon fertig" suggerieren waehrend die Lage in der Queue steht
+            if (typeof Pipeline !== 'undefined' && Pipeline.beginQueue) {
+                Pipeline.beginQueue(this.currentIncidentId);
+            }
             if (result && result.status === 'skipped') {
                 UI.showToast('Aktualisierung ist in der Warteschlange und wird ausgefuehrt, sobald die aktuelle Recherche abgeschlossen ist.', 'info');
             } else {
@@ -2693,6 +2702,12 @@ async handleRefresh() {
     async openSourceManagement() {
         openModal('modal-sources');
         await this.loadSources();
+        // Admin sieht den Review-Tab
+        const reviewTab = document.getElementById('sources-tab-review');
+        if (reviewTab && this.user && this.user.role === 'org_admin') {
+            reviewTab.style.display = '';
+            this._refreshReviewBadge().catch(() => {});
+        }
     },
 
     async loadSources() {
@@ -2713,6 +2728,122 @@ async handleRefresh() {
         }
     },
 
+    async _refreshReviewBadge() {
+        try {
+            const stats = await API.getClassificationStats();
+            const badge = document.getElementById('sources-review-count');
+            if (badge) badge.textContent = String(stats.pending_review || 0);
+        } catch (_) { /* still ok */ }
+    },
+
+    switchSourcesTab(tab) {
+        const listView = document.getElementById('sources-list-view');
+        const reviewView = document.getElementById('sources-review-view');
+        const tabList = document.getElementById('sources-tab-list');
+        const tabReview = document.getElementById('sources-tab-review');
+        if (!listView || !reviewView) return;
+        if (tab === 'review') {
+            listView.style.display = 'none';
+            reviewView.style.display = '';
+            if (tabList) { tabList.classList.remove('active'); tabList.setAttribute('aria-selected', 'false'); }
+            if (tabReview) { tabReview.classList.add('active'); tabReview.setAttribute('aria-selected', 'true'); }
+            this.loadClassificationQueue();
+        } else {
+            listView.style.display = '';
+            reviewView.style.display = 'none';
+            if (tabList) { tabList.classList.add('active'); tabList.setAttribute('aria-selected', 'true'); }
+            if (tabReview) { tabReview.classList.remove('active'); tabReview.setAttribute('aria-selected', 'false'); }
+        }
+    },
+
+    async loadClassificationQueue() {
+        const list = document.getElementById('sources-review-list');
+        if (!list) return;
+        const minConf = parseFloat(document.getElementById('review-min-confidence')?.value || '0');
+        list.innerHTML = '<div class="empty-state-text" style="padding:var(--sp-3xl);text-align:center;">Lade...</div>';
+        try {
+            const items = await API.getClassificationQueue(200, minConf);
+            this._reviewItems = items;
+            const countEl = document.getElementById('review-pending-count');
+            if (countEl) countEl.textContent = String(items.length);
+            if (items.length === 0) {
+                list.innerHTML = '<div class="empty-state-text" style="padding:var(--sp-3xl);text-align:center;">Keine ausstehenden Vorschlaege.</div>';
+                return;
+            }
+            list.innerHTML = items.map(item => UI.renderClassificationQueueItem(item)).join('');
+        } catch (err) {
+            list.innerHTML = `<div class="empty-state-text" style="padding:var(--sp-3xl);text-align:center;color:var(--danger);">Fehler: ${err.message}</div>`;
+        }
+    },
+
+    async approveClassification(id) {
+        try {
+            await API.approveClassification(id);
+            UI.showToast('Klassifikation uebernommen.', 'success');
+            await this.loadClassificationQueue();
+            this._refreshReviewBadge();
+        } catch (err) {
+            UI.showToast('Approve fehlgeschlagen: ' + err.message, 'error');
+        }
+    },
+
+    async rejectClassification(id) {
+        try {
+            await API.rejectClassification(id);
+            UI.showToast('Vorschlag verworfen.', 'success');
+            await this.loadClassificationQueue();
+            this._refreshReviewBadge();
+        } catch (err) {
+            UI.showToast('Reject fehlgeschlagen: ' + err.message, 'error');
+        }
+    },
+
+    async reclassifySource(id) {
+        const btn = document.querySelector(`[data-reclassify-id="${id}"]`);
+        if (btn) { btn.disabled = true; btn.textContent = '...'; }
+        try {
+            await API.reclassifySource(id);
+            UI.showToast('Neu klassifiziert.', 'success');
+            await this.loadClassificationQueue();
+        } catch (err) {
+            UI.showToast('Reclassify fehlgeschlagen: ' + err.message, 'error');
+        } finally {
+            if (btn) { btn.disabled = false; btn.textContent = 'Neu klassifizieren'; }
+        }
+    },
+
+    async triggerBulkClassify() {
+        if (!confirm('Bulk-Klassifikation aller noch nicht klassifizierten Quellen starten? Lauft im Hintergrund (~3-5 Sek pro Quelle, ~0.02 USD pro Quelle).')) return;
+        try {
+            const r = await API.triggerBulkClassify(500, true);
+            UI.showToast(`Bulk-Klassifikation gestartet (limit=${r.limit}). Nachschauen mit Reload.`, 'info');
+        } catch (err) {
+            UI.showToast('Start fehlgeschlagen: ' + err.message, 'error');
+        }
+    },
+
+    async bulkApproveHighConfidence() {
+        if (!confirm('Alle Vorschlaege mit Konfidenz >= 0.85 genehmigen?')) return;
+        try {
+            const r = await API.bulkApproveClassifications(0.85);
+            UI.showToast(`${r.approved_count} Vorschlaege uebernommen.`, 'success');
+            await this.loadClassificationQueue();
+            this._refreshReviewBadge();
+        } catch (err) {
+            UI.showToast('Bulk-Approve fehlgeschlagen: ' + err.message, 'error');
+        }
+    },
+
+    async triggerExternalReputationSync() {
+        if (!confirm('IFCN- und EUvsDisinfo-Datenbanken jetzt syncen? Lauft im Hintergrund (~30 Sek).')) return;
+        try {
+            await API.triggerExternalReputationSync();
+            UI.showToast('Externer Sync gestartet. Quellenliste in 30 Sek neu laden.', 'info');
+        } catch (err) {
+            UI.showToast('Sync fehlgeschlagen: ' + err.message, 'error');
+        }
+    },
+
     renderSourceStats(stats) {
         const bar = document.getElementById('sources-stats-bar');
         if (!bar) return;
@@ -2741,6 +2872,11 @@ async handleRefresh() {
         // Filter anwenden
         const typeFilter = document.getElementById('sources-filter-type')?.value || '';
         const catFilter = document.getElementById('sources-filter-category')?.value || '';
+        const politicalFilter = document.getElementById('sources-filter-political')?.value || '';
+        const mediaTypeFilter = document.getElementById('sources-filter-mediatype')?.value || '';
+        const reliabilityFilter = document.getElementById('sources-filter-reliability')?.value || '';
+        const alignmentFilter = document.getElementById('sources-filter-alignment')?.value || '';
+        const externFilter = document.getElementById('sources-filter-extern')?.value || '';
         const search = (document.getElementById('sources-search')?.value || '').toLowerCase();
 
         // Alle Quellen nach Domain gruppieren
@@ -2791,6 +2927,25 @@ async handleRefresh() {
                 if (!hasMatchingCat) continue;
             }
 
+            // Klassifikations-Filter
+            if (politicalFilter) {
+                if (!feeds.some(f => (f.political_orientation || 'na') === politicalFilter)) continue;
+            }
+            if (mediaTypeFilter) {
+                if (!feeds.some(f => (f.media_type || 'sonstige') === mediaTypeFilter)) continue;
+            }
+            if (reliabilityFilter) {
+                if (!feeds.some(f => (f.reliability || 'na') === reliabilityFilter)) continue;
+            }
+            if (alignmentFilter) {
+                if (!feeds.some(f => Array.isArray(f.alignments) && f.alignments.includes(alignmentFilter))) continue;
+            }
+            if (externFilter === 'ifcn') {
+                if (!feeds.some(f => f.ifcn_signatory)) continue;
+            } else if (externFilter === 'eu_disinfo') {
+                if (!feeds.some(f => f.eu_disinfo_listed)) continue;
+            }
+
             // Suche
             if (search) {
                 const groupText = feeds.map(f =>
@@ -3045,6 +3200,13 @@ async handleRefresh() {
             document.getElementById('src-discover-btn').disabled = false;
             document.getElementById('src-discover-btn').textContent = 'Erkennen';
             document.getElementById('src-type-select').value = 'rss_feed';
+            // Klassifikations-Felder auf Default zurücksetzen
+            const polEl = document.getElementById('src-political'); if (polEl) polEl.value = 'na';
+            const mtEl = document.getElementById('src-mediatype'); if (mtEl) mtEl.value = 'sonstige';
+            const relEl = document.getElementById('src-reliability'); if (relEl) relEl.value = 'na';
+            const ccEl = document.getElementById('src-country'); if (ccEl) ccEl.value = '';
+            const saEl = document.getElementById('src-state-affiliated'); if (saEl) saEl.checked = false;
+            this._setAlignmentChips([]);
             // Save-Button Text zurücksetzen
             const saveBtn = document.querySelector('#src-discovery-result .sources-discovery-actions .btn-primary');
             if (saveBtn) saveBtn.textContent = 'Speichern';
@@ -3226,6 +3388,19 @@ async handleRefresh() {
             rss_url: source.url,
         };
 
+        // Klassifikations-Felder setzen
+        const polEl = document.getElementById('src-political');
+        if (polEl) polEl.value = source.political_orientation || 'na';
+        const mtEl = document.getElementById('src-mediatype');
+        if (mtEl) mtEl.value = source.media_type || 'sonstige';
+        const relEl = document.getElementById('src-reliability');
+        if (relEl) relEl.value = source.reliability || 'na';
+        const ccEl = document.getElementById('src-country');
+        if (ccEl) ccEl.value = source.country_code || '';
+        const saEl = document.getElementById('src-state-affiliated');
+        if (saEl) saEl.checked = !!source.state_affiliated;
+        this._setAlignmentChips(source.alignments || []);
+
         // Submit-Button-Text ändern
         const saveBtn = document.querySelector('#src-discovery-result .sources-discovery-actions .btn-primary');
         if (saveBtn) saveBtn.textContent = 'Quelle speichern';
@@ -3234,6 +3409,27 @@ async handleRefresh() {
         if (form) form.scrollIntoView({ behavior: 'smooth', block: 'start' });
     },
 
+    _setAlignmentChips(active) {
+        const chips = document.querySelectorAll('#src-alignments-chips .alignment-chip');
+        const set = new Set((active || []).map(a => (a || '').toLowerCase()));
+        chips.forEach(chip => {
+            if (set.has(chip.dataset.alignment)) chip.classList.add('active');
+            else chip.classList.remove('active');
+        });
+    },
+
+    _getAlignmentChips() {
+        return Array.from(document.querySelectorAll('#src-alignments-chips .alignment-chip.active'))
+            .map(chip => chip.dataset.alignment);
+    },
+
+    handleAlignmentChipClick(e) {
+        const chip = e.target.closest('.alignment-chip');
+        if (!chip) return;
+        e.preventDefault();
+        chip.classList.toggle('active');
+    },
+
     async saveSource() {
         const name = document.getElementById('src-name').value.trim();
         if (!name) {
@@ -3249,6 +3445,12 @@ async handleRefresh() {
             url: discovered.rss_url || (discovered.source_type === 'telegram_channel' ? (document.getElementById('src-domain').value || null) : null),
             domain: document.getElementById('src-domain').value.trim() || discovered.domain || null,
             notes: document.getElementById('src-notes').value.trim() || null,
+            political_orientation: document.getElementById('src-political')?.value || 'na',
+            media_type: document.getElementById('src-mediatype')?.value || 'sonstige',
+            reliability: document.getElementById('src-reliability')?.value || 'na',
+            country_code: (document.getElementById('src-country')?.value || '').trim().toUpperCase() || null,
+            state_affiliated: !!document.getElementById('src-state-affiliated')?.checked,
+            alignments: this._getAlignmentChips(),
         };
 
         if (!data.domain && discovered.domain) {

src/static/js/components.js

@@ -354,9 +354,22 @@ const UI = {
         const minBtn = document.getElementById('progress-popup-minimize');
         if (minBtn) minBtn.style.display = state.isFirst ? 'none' : '';
 
-        // Title
+        // Title - haengt von Status ab (queued = wartet, cancelling = bricht ab, sonst laeuft)
         const titleEl = document.getElementById('progress-popup-title');
-        if (titleEl) titleEl.textContent = state.isFirst ? 'Erste Recherche l\u00e4uft' : 'Aktualisierung l\u00e4uft';
+        if (titleEl) {
+            let title;
+            if (status === 'queued') {
+                const pos = (state && state._queuePos) ? ' (#' + state._queuePos + ')' : '';
+                title = 'In Warteschlange' + pos;
+            } else if (status === 'cancelling') {
+                title = 'Wird abgebrochen\u2026';
+            } else if (state.isFirst) {
+                title = 'Erste Recherche l\u00e4uft';
+            } else {
+                title = 'Aktualisierung l\u00e4uft';
+            }
+            titleEl.textContent = title;
+        }
 
         // Multi-pass info
         const passEl = document.getElementById('progress-popup-pass');
@@ -1049,6 +1062,163 @@ const UI = {
         'sonstige': 'Sonstige',
     },
 
+    _politicalLabels: {
+        links_extrem: { short: 'L+', full: 'Links (extrem)' },
+        links: { short: 'L', full: 'Links' },
+        mitte_links: { short: 'ML', full: 'Mitte-Links' },
+        liberal: { short: 'LIB', full: 'Liberal' },
+        mitte: { short: 'M', full: 'Mitte' },
+        konservativ: { short: 'KON', full: 'Konservativ' },
+        mitte_rechts: { short: 'MR', full: 'Mitte-Rechts' },
+        rechts: { short: 'R', full: 'Rechts' },
+        rechts_extrem: { short: 'R+', full: 'Rechts (extrem)' },
+        na: { short: '?', full: 'Nicht eingeordnet' },
+    },
+    _reliabilityLabels: {
+        sehr_hoch: 'Sehr hoch',
+        hoch: 'Hoch',
+        gemischt: 'Gemischt',
+        niedrig: 'Niedrig',
+        sehr_niedrig: 'Sehr niedrig',
+        na: 'Nicht eingeordnet',
+    },
+    _mediaTypeLabels: {
+        tageszeitung: 'Tageszeitung',
+        wochenzeitung: 'Wochenzeitung',
+        magazin: 'Magazin',
+        tv_sender: 'TV-Sender',
+        radio: 'Radio',
+        oeffentlich_rechtlich: 'Öffentlich-Rechtlich',
+        nachrichtenagentur: 'Nachrichtenagentur',
+        online_only: 'Online-only',
+        blog: 'Blog',
+        telegram_kanal: 'Telegram-Kanal',
+        telegram_bot: 'Telegram-Bot',
+        podcast: 'Podcast',
+        social_media: 'Social Media',
+        imageboard: 'Imageboard',
+        think_tank: 'Think Tank',
+        ngo: 'NGO',
+        behoerde: 'Behörde',
+        staatsmedium: 'Staatsmedium',
+        fachmedium: 'Fachmedium',
+        sonstige: 'Sonstige',
+    },
+    _alignmentLabels: {
+        prorussisch: 'prorussisch',
+        proiranisch: 'proiranisch',
+        prowestlich: 'prowestlich',
+        proukrainisch: 'proukrainisch',
+        prochinesisch: 'prochinesisch',
+        projapanisch: 'projapanisch',
+        proisraelisch: 'proisraelisch',
+        propalaestinensisch: 'propalästinensisch',
+        protuerkisch: 'protürkisch',
+        panarabisch: 'panarabisch',
+        neutral: 'neutral',
+        sonstige: 'sonstige',
+    },
+
+    /**
+     * Eintrag in der Klassifikations-Review-Queue.
+     * Zeigt Diff zwischen aktuellem Wert und LLM-Vorschlag.
+     */
+    renderClassificationQueueItem(item) {
+        const cur = item.current || {};
+        const prop = item.proposed || {};
+        const conf = prop.confidence || 0;
+        const confPct = Math.round(conf * 100);
+        const confClass = conf >= 0.85 ? 'high' : (conf >= 0.7 ? 'medium' : 'low');
+
+        const diffRow = (label, currentVal, proposedVal, formatter) => {
+            const fmt = formatter || (v => v == null || v === '' ? '–' : String(v));
+            const c = fmt(currentVal);
+            const p = fmt(proposedVal);
+            const changed = c !== p;
+            return `<div class="review-diff-row${changed ? ' changed' : ''}">
+                <span class="review-diff-label">${this.escape(label)}</span>
+                <span class="review-diff-current">${this.escape(c)}</span>
+                <span class="review-diff-arrow">→</span>
+                <span class="review-diff-proposed">${this.escape(p)}</span>
+            </div>`;
+        };
+
+        const polFmt = v => (v && v !== 'na') ? (this._politicalLabels[v]?.full || v) : '–';
+        const mtFmt = v => (v && v !== 'sonstige') ? (this._mediaTypeLabels[v] || v) : (v === 'sonstige' ? 'Sonstige' : '–');
+        const relFmt = v => (v && v !== 'na') ? (this._reliabilityLabels[v] || v) : '–';
+        const stateFmt = v => v ? 'ja' : 'nein';
+        const ccFmt = v => v || '–';
+        const alignFmt = v => (Array.isArray(v) && v.length > 0)
+            ? v.map(a => this._alignmentLabels[a] || a).join(', ')
+            : '–';
+
+        const globalBadge = item.is_global ? '<span class="review-global-badge">Grundquelle</span>' : '';
+        const reasoning = prop.reasoning ? this.escape(prop.reasoning) : '';
+
+        return `<div class="review-card" data-source-id="${item.id}">
+            <div class="review-card-header">
+                <div class="review-card-title">
+                    <span class="review-card-name">${this.escape(item.name)}</span>
+                    ${globalBadge}
+                    <span class="review-card-domain">${this.escape(item.domain || '')}</span>
+                </div>
+                <div class="review-card-confidence conf-${confClass}" title="LLM-Konfidenz">
+                    <span class="conf-value">${confPct}%</span>
+                    <span class="conf-label">Konfidenz</span>
+                </div>
+            </div>
+            <div class="review-card-diff">
+                ${diffRow('Politik', cur.political_orientation, prop.political_orientation, polFmt)}
+                ${diffRow('Medientyp', cur.media_type, prop.media_type, mtFmt)}
+                ${diffRow('Glaubwürdigkeit', cur.reliability, prop.reliability, relFmt)}
+                ${diffRow('Staatsnah', cur.state_affiliated, prop.state_affiliated, stateFmt)}
+                ${diffRow('Land', cur.country_code, prop.country_code, ccFmt)}
+                ${diffRow('Geopol. Nähe', cur.alignments, prop.alignments, alignFmt)}
+            </div>
+            ${reasoning ? `<div class="review-card-reasoning"><strong>Begründung:</strong> ${reasoning}</div>` : ''}
+            <div class="review-card-actions">
+                <button class="btn btn-small btn-primary" onclick="App.approveClassification(${item.id})">Übernehmen</button>
+                <button class="btn btn-small btn-secondary" onclick="App.rejectClassification(${item.id})">Verwerfen</button>
+                <button class="btn btn-small btn-secondary" data-reclassify-id="${item.id}" onclick="App.reclassifySource(${item.id})">Neu klassifizieren</button>
+            </div>
+        </div>`;
+    },
+
+    _renderClassificationBadges(feed) {
+        const parts = [];
+        const pol = feed.political_orientation;
+        if (pol && pol !== 'na') {
+            const label = this._politicalLabels[pol] || { short: pol, full: pol };
+            parts.push(`<span class="source-political-badge pol-${this.escape(pol)}" title="${this.escape(label.full)}">${this.escape(label.short)}</span>`);
+        }
+        const rel = feed.reliability;
+        if (rel && rel !== 'na') {
+            const relLabel = this._reliabilityLabels[rel] || rel;
+            const relSource = feed.ifcn_signatory ? '(IFCN-Faktenchecker)'
+                : (feed.eu_disinfo_listed ? `(EU-Desinfo, ${feed.eu_disinfo_case_count || 0} Fälle)`
+                : '(LLM-Schätzung)');
+            const relTitle = `Glaubwürdigkeit: ${relLabel} ${relSource}`;
+            parts.push(`<span class="source-reliability-dot rel-${this.escape(rel)}" title="${this.escape(relTitle)}" aria-label="${this.escape(relTitle)}"></span>`);
+        }
+        if (feed.ifcn_signatory) {
+            parts.push(`<span class="source-ifcn-badge" title="IFCN-zertifizierter Faktenchecker" aria-label="IFCN-Faktenchecker">✓ IFCN</span>`);
+        }
+        if (feed.eu_disinfo_listed) {
+            const cnt = feed.eu_disinfo_case_count || 0;
+            const title = `EUvsDisinfo: ${cnt} dokumentierte Desinformations-Fälle`;
+            parts.push(`<span class="source-eu-disinfo-badge" title="${this.escape(title)}" aria-label="${this.escape(title)}">⚠ EU-Desinfo (${cnt})</span>`);
+        }
+        if (feed.state_affiliated) {
+            parts.push(`<span class="source-state-badge" title="Staatsnah/-kontrolliert" aria-label="Staatsnah">⚑</span>`);
+        }
+        const aligns = Array.isArray(feed.alignments) ? feed.alignments : [];
+        aligns.forEach(a => {
+            const label = this._alignmentLabels[a] || a;
+            parts.push(`<span class="source-alignment-chip-badge align-${this.escape(a)}">${this.escape(label)}</span>`);
+        });
+        return parts.join('');
+    },
+
     /**
      * Domain-Gruppe rendern (aufklappbar mit Feeds).
      */
@@ -1104,20 +1274,52 @@ const UI = {
             ? `<span class="source-feed-count">${feedCount} Feed${feedCount !== 1 ? 's' : ''}</span>`
             : '';
 
-        // Info-Button mit Tooltip (Typ, Sprache, Ausrichtung)
+        // Info-Button mit Tooltip (Typ, Sprache, Ausrichtung, Klassifikation)
         let infoButtonHtml = '';
         const firstFeed = feeds[0] || {};
-        const hasInfo = firstFeed.language || firstFeed.bias;
+        const hasInfo = firstFeed.language || firstFeed.bias
+            || (firstFeed.political_orientation && firstFeed.political_orientation !== 'na')
+            || (firstFeed.media_type && firstFeed.media_type !== 'sonstige')
+            || (firstFeed.reliability && firstFeed.reliability !== 'na')
+            || firstFeed.state_affiliated
+            || firstFeed.country_code
+            || (Array.isArray(firstFeed.alignments) && firstFeed.alignments.length > 0);
         if (hasInfo) {
-            const typeMap = { rss_feed: 'RSS-Feed', web_source: 'Web-Quelle', telegram_channel: 'Telegram-Kanal' };
+            const typeMap = { rss_feed: 'RSS-Feed', web_source: 'Web-Quelle', telegram_channel: 'Telegram-Kanal', podcast_feed: 'Podcast' };
             const lines = [];
             lines.push('Typ: ' + (typeMap[firstFeed.source_type] || firstFeed.source_type || 'Unbekannt'));
             if (firstFeed.language) lines.push('Sprache: ' + firstFeed.language);
-            if (firstFeed.bias) lines.push('Ausrichtung: ' + firstFeed.bias);
+            if (firstFeed.country_code) lines.push('Land: ' + firstFeed.country_code);
+            if (firstFeed.media_type && firstFeed.media_type !== 'sonstige') {
+                lines.push('Medientyp: ' + (this._mediaTypeLabels[firstFeed.media_type] || firstFeed.media_type));
+            }
+            if (firstFeed.political_orientation && firstFeed.political_orientation !== 'na') {
+                const pl = this._politicalLabels[firstFeed.political_orientation];
+                lines.push('Politisch: ' + (pl ? pl.full : firstFeed.political_orientation));
+            }
+            if (firstFeed.reliability && firstFeed.reliability !== 'na') {
+                const relLabel = this._reliabilityLabels[firstFeed.reliability] || firstFeed.reliability;
+                const relSrc = firstFeed.ifcn_signatory ? ' (IFCN-Faktenchecker)'
+                    : (firstFeed.eu_disinfo_listed ? ` (EU-Desinfo, ${firstFeed.eu_disinfo_case_count || 0} Fälle)`
+                    : ' (LLM-Schätzung)');
+                lines.push('Glaubwürdigkeit: ' + relLabel + relSrc);
+            }
+            if (firstFeed.ifcn_signatory) lines.push('IFCN-Faktenchecker: ja');
+            if (firstFeed.eu_disinfo_listed) {
+                lines.push(`EUvsDisinfo: ${firstFeed.eu_disinfo_case_count || 0} Fälle` + (firstFeed.eu_disinfo_last_seen ? ` (zuletzt ${firstFeed.eu_disinfo_last_seen})` : ''));
+            }
+            if (firstFeed.state_affiliated) lines.push('Staatsnah: ja');
+            if (Array.isArray(firstFeed.alignments) && firstFeed.alignments.length > 0) {
+                const labels = firstFeed.alignments.map(a => this._alignmentLabels[a] || a);
+                lines.push('Geopolitische Nähe: ' + labels.join(', '));
+            }
+            if (firstFeed.bias) lines.push('Notiz: ' + firstFeed.bias);
             const tooltipText = this.escape(lines.join('\n'));
             infoButtonHtml = ` <span class="info-icon tooltip-below" data-tooltip="${tooltipText}"><svg xmlns="http://www.w3.org/2000/svg" width="14" height="14" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round"><circle cx="12" cy="12" r="10"/><path d="M12 16v-4"/><path d="M12 8h.01"/></svg></span>`;
         }
 
+        const classificationBadges = this._renderClassificationBadges(firstFeed);
+
         return `<div class="source-group">
             <div class="source-group-header" ${toggleAttr}>
                 ${toggleIcon}
@@ -1125,6 +1327,7 @@ const UI = {
                     <span class="source-group-name">${this.escape(displayName)}</span>${infoButtonHtml}
                 </div>
                 <span class="source-category-badge cat-${feeds[0]?.category || 'sonstige'}">${catLabel}</span>
+                ${classificationBadges ? `<span class="source-classification-badges">${classificationBadges}</span>` : ''}
                 ${feedCountBadge}
                 <div class="source-group-actions" onclick="event.stopPropagation()">
                     ${!isGlobal && !hasMultiple && feeds[0]?.id ? `<button class="source-edit-btn" onclick="App.editSource(${feeds[0].id})" title="Bearbeiten" aria-label="Bearbeiten">&#9998;</button>` : ''}

src/static/js/pipeline.js

@@ -19,6 +19,7 @@ const Pipeline = {
     _incidentId: null,
     _definition: null,        // PIPELINE_STEPS vom Backend
     _stateByKey: {},          // step_key -> {status, count_value, count_secondary, pass_number}
+    _snapshotState: null,     // deep-copy von _stateByKey vor Refresh-Start (fuer Cancel-Restore)
     _isResearch: false,
     _passTotal: 1,
     _lastRefreshHeader: null,
@@ -42,10 +43,11 @@ const Pipeline = {
         if (this._wsBound) return;
         if (typeof WS !== 'undefined' && WS.on) {
             WS.on('pipeline_step', (msg) => this._onWsStep(msg));
-            // Bei Refresh-Complete den finalen Stand neu laden, damit Zahlen gefroren sichtbar bleiben
-            WS.on('refresh_complete', (msg) => this._onRefreshDone(msg));
-            WS.on('refresh_cancelled', (msg) => this._onRefreshDone(msg));
-            WS.on('refresh_error', (msg) => this._onRefreshDone(msg));
+            // Erfolg: API-State neu laden (finaler Stand sichtbar)
+            WS.on('refresh_complete', (msg) => this._onRefreshDoneSuccess(msg));
+            // Cancel/Error: vor-Refresh-Snapshot zurueckspielen, damit Pipeline nicht im Mix-Zustand stehen bleibt
+            WS.on('refresh_cancelled', (msg) => this._onRefreshDoneCancel(msg));
+            WS.on('refresh_error', (msg) => this._onRefreshDoneError(msg));
             this._wsBound = true;
         }
         // Hover-Tooltip-Element vorbereiten
@@ -68,6 +70,7 @@ const Pipeline = {
     async bindToIncident(incidentId) {
         this._incidentId = incidentId;
         this._stateByKey = {};
+        this._snapshotState = null;  // Snapshot ist immer lagen-spezifisch
         this._isResearch = false;
         this._passTotal = 1;
         this._lastRefreshHeader = null;
@@ -101,6 +104,20 @@ const Pipeline = {
 
             this._render();
             this._renderMini();
+
+            // Edge-Case: Lage ist gerade in Queue (z.B. via Lagen-Wechsel beim
+            // Klick in der Sidebar). API liefert den LETZTEN gespeicherten Stand
+            // (alles done = gruen), aber tatsaechlich wartet ein neuer Refresh.
+            // -> beginQueue() selbst ausloesen, damit Icons grau zeigen.
+            try {
+                if (typeof App !== 'undefined' && App._refreshingIncidents
+                    && App._refreshingIncidents.has(incidentId)
+                    && typeof UI !== 'undefined' && UI._progressState
+                    && UI._progressState[incidentId]
+                    && UI._progressState[incidentId].step === 'queued') {
+                    this.beginQueue(incidentId);
+                }
+            } catch (e) { /* tolerant */ }
         } catch (e) {
             console.warn('Pipeline laden fehlgeschlagen:', e);
             this._renderEmpty('Pipeline-Daten konnten nicht geladen werden.');
@@ -166,14 +183,65 @@ const Pipeline = {
         }
     },
 
-    _onRefreshDone(msg) {
+    /**
+     * Wird vom Frontend gerufen, wenn ein Refresh angestossen wurde (queued).
+     * Macht einen Snapshot des aktuellen Pipeline-Stands (zur spaeteren Wiederherstellung
+     * bei Cancel/Error) und setzt dann alle Steps auf "pending" - damit der User sieht:
+     * "neuer Refresh laeuft an, alte gruene Haekchen sind nicht mehr aktuell".
+     */
+    beginQueue(incidentId) {
+        if (this._incidentId !== incidentId) return;          // andere Lage offen
+        if (!this._definition) return;                        // noch keine Pipeline-Definition geladen
+        // Aktuellen Stand sichern (deep-copy). Bei Mehrfach-Refresh ohne Cancel
+        // dazwischen wird der Snapshot bewusst ueberschrieben - er soll immer
+        // der "Stand kurz vor diesem Refresh" sein.
+        this._snapshotState = JSON.parse(JSON.stringify(this._stateByKey));
+        // Alle Steps auf pending setzen
+        this._definition.forEach(s => {
+            if (this._stateByKey[s.key]) {
+                this._stateByKey[s.key].status = 'pending';
+            } else {
+                this._stateByKey[s.key] = { status: 'pending', count_value: null, count_secondary: null, pass_number: 1 };
+            }
+        });
+        this._render();
+        this._renderMini();
+    },
+
+    /** Restauriert den letzten Snapshot. Rueckgabe: true bei Erfolg, false wenn keiner da war. */
+    _restoreSnapshot() {
+        if (!this._snapshotState) return false;
+        this._stateByKey = this._snapshotState;
+        this._snapshotState = null;
+        this._render();
+        this._renderMini();
+        return true;
+    },
+
+    _onRefreshDoneSuccess(msg) {
         if (this._incidentId == null || (msg && msg.incident_id !== this._incidentId)) return;
+        this._snapshotState = null;  // verworfen, neuer Stand wird vom API geladen
         // Daten frisch nachladen, damit Header (Dauer) und finale Zahlen passen
         setTimeout(() => {
             if (this._incidentId != null) this.bindToIncident(this._incidentId);
         }, 600);
     },
 
+    _onRefreshDoneCancel(msg) {
+        if (this._incidentId == null || (msg && msg.incident_id !== this._incidentId)) return;
+        if (!this._restoreSnapshot()) {
+            // Kein Snapshot vorhanden (z.B. Page-Reload mitten im Refresh) -> wie bisher API-Reload
+            setTimeout(() => {
+                if (this._incidentId != null) this.bindToIncident(this._incidentId);
+            }, 600);
+        }
+    },
+
+    _onRefreshDoneError(msg) {
+        // Wie Cancel: vorheriger Stand zurueck (nicht im Mix-Zustand stehenbleiben)
+        this._onRefreshDoneCancel(msg);
+    },
+
     /** Vollbild-Pipeline (Tab "Analysepipeline") als 3x3-Snake rendern. */
     _render() {
         const stage = document.getElementById('pipeline-stage');