Faktencheck: Ursprungsquellen bei fehlender URL anzeigen statt blind herabstufen

src/agents/factchecker.py

@@ -469,7 +469,7 @@ class FactCheckerAgent:
 
         try:
             result, usage = await call_claude(prompt)
-            facts = self._parse_response(result)
+            facts = self._parse_response(result, articles=articles)
             logger.info(f"Faktencheck: {len(facts)} Fakten geprüft")
             return facts, usage
         except TimeoutError:
@@ -511,7 +511,7 @@ class FactCheckerAgent:
 
         try:
             result, usage = await call_claude(prompt)
-            facts = self._parse_response(result)
+            facts = self._parse_response(result, articles=new_articles)
             logger.info(f"Inkrementeller Faktencheck: {len(facts)} Fakten (neu + aktualisiert)")
             return facts, usage
         except TimeoutError:
@@ -705,27 +705,66 @@ class FactCheckerAgent:
         logger.warning("Konnte Triage-Antwort nicht als JSON parsen")
         return None
 
-    def _validate_facts(self, facts: list[dict]) -> list[dict]:
-        """Validiert Fakten: confirmed/established ohne URL wird herabgestuft."""
+    def _validate_facts(self, facts: list[dict], articles: list[dict] = None) -> list[dict]:
+        """Validiert Fakten: Bei fehlender URL werden Ursprungsquellen aus den Artikeln ergaenzt."""
         url_pattern = re.compile(r'https?://')
+        # Verfuegbare Artikel-URLs sammeln
+        article_sources = []
+        if articles:
+            for a in articles:
+                url = a.get("source_url", "")
+                source = a.get("source", "")
+                headline = a.get("headline_de") or a.get("headline", "")
+                if url:
+                    article_sources.append({"url": url, "source": source, "headline": headline})
+
         for fact in facts:
             status = fact.get("status", "")
             evidence = fact.get("evidence") or ""
             if status in ("confirmed", "established") and not url_pattern.search(evidence):
+                # Passende Ursprungsquellen finden (Keyword-Match auf Claim)
+                claim_lower = (fact.get("claim") or "").lower()
+                claim_words = [w for w in claim_lower.split() if len(w) >= 4][:8]
+                matched_sources = []
+                for src in article_sources:
+                    src_text = (src["headline"] + " " + src["source"]).lower()
+                    matches = sum(1 for w in claim_words if w in src_text)
+                    if matches >= max(1, len(claim_words) // 4):
+                        matched_sources.append(src)
+                    if len(matched_sources) >= 3:
+                        break
+
+                if matched_sources:
+                    # Ursprungsquellen anhaengen statt herabstufen
+                    source_refs = "; ".join(
+                        f"{s['source']} ({s['url']})" for s in matched_sources
+                    )
+                    fact["evidence"] = (
+                        evidence.rstrip(". ") +
+                        ". [Ursprungsquellen: " + source_refs +
+                        " — Quellenlinks zum Zeitpunkt der Recherche moeglicherweise nicht mehr verfuegbar]"
+                    )
+                    logger.info(
+                        f"Fakt '{fact.get('claim', '')[:50]}...' ergaenzt mit "
+                        f"{len(matched_sources)} Ursprungsquelle(n)"
+                    )
+                else:
+                    # Keine passende Quelle gefunden -> herabstufen
                     old_status = status
                     fact["status"] = "unconfirmed" if status == "confirmed" else "unverified"
                     logger.warning(
                         f"Fakt herabgestuft ({old_status} -> {fact['status']}): "
-                    f"keine URL in Evidenz: '{fact.get('claim', '')[:60]}...'"
+                        f"keine URL in Evidenz und keine passende Ursprungsquelle: "
+                        f"'{fact.get('claim', '')[:60]}...'"
                     )
         return facts
 
-    def _parse_response(self, response: str) -> list[dict]:
+    def _parse_response(self, response: str, articles: list[dict] = None) -> list[dict]:
         """Parst die Claude-Antwort als JSON-Array."""
         try:
             data = json.loads(response)
             if isinstance(data, list):
-                return self._validate_facts(data)
+                return self._validate_facts(data, articles=articles)
         except json.JSONDecodeError:
             pass
 

src/config.py

@@ -78,7 +78,7 @@ MAGIC_LINK_EXPIRE_MINUTES = 10
 MAGIC_LINK_BASE_URL = os.environ.get("MAGIC_LINK_BASE_URL", "https://monitor.aegis-sight.de")
 
 # Telegram (Telethon)
-TELEGRAM_API_ID = int(os.environ.get("TELEGRAM_API_ID", "2040"))
-TELEGRAM_API_HASH = os.environ.get("TELEGRAM_API_HASH", "b18441a1ff607e10a989891a5462e627")
+TELEGRAM_API_ID = int(os.environ.get("TELEGRAM_API_ID", "31330502"))
+TELEGRAM_API_HASH = os.environ.get("TELEGRAM_API_HASH", "842db7220ad2d5371269d6d88cde6a84")
 TELEGRAM_SESSION_PATH = os.environ.get("TELEGRAM_SESSION_PATH", "/home/claude-dev/.telegram/telegram_session")
 

src/static/css/style.css

@@ -4557,16 +4557,16 @@ a.map-popup-article:hover {
 .tg-cat-grid {
   display: grid;
   grid-template-columns: 1fr 1fr;
-  gap: 6px 16px;
+  gap: 8px 24px;
 }
 .tg-cat-item {
   display: flex;
   align-items: center;
-  gap: 6px;
+  gap: 10px;
   font-size: 13px;
   color: var(--text-primary);
   cursor: pointer;
-  padding: 3px 0;
+  padding: 5px 0;
 }
 .tg-cat-item input[type="checkbox"] {
   accent-color: var(--accent);