Backfill-Migration: ASCII-Umlaute in articles korrigieren

Idempotente Migration mit --db Parameter, Backup vor Lauf, Sample-Output.
Behandelt headline_de + content_de bei allen Artikeln; bei language=de
zusaetzlich headline + content_original. Nutzt das gleiche hunspell-Dict
wie der Live-QC.

migrations/migrate_umlauts_2026-05-03.py

@@ -0,0 +1,127 @@
+"""Backfill 2026-05-03: ASCII-Umlaute in articles korrigieren.
+
+Ursache: Zwei verschiedene Pfade lassen ASCII-Umlaute in der DB landen:
+1. Quellen wie dpa-AFX/finanznachrichten.de oder Telegram-Kanaele liefern
+   Headlines schon als "Gespraeche" statt "Gespräche".
+2. LLM-Uebersetzungen (analyzer.py / orchestrator.py) erzeugen gelegentlich
+   ASCII-Umlaute in headline_de/content_de trotz Prompt-Anweisung.
+
+Diese Migration normalisiert rueckwirkend alle articles per
+normalize_german_umlauts (hunspell-Dict mit ~150k deutschen Woertern).
+
+Felder pro Artikel:
+- headline_de, content_de  (immer)
+- headline, content_original (nur wenn language='de')
+
+Idempotent: nur Aenderungen werden geschrieben.
+"""
+import argparse
+import shutil
+import sqlite3
+import sys
+from datetime import datetime
+
+# normalize_german_umlauts aus dem Monitor-Repo importieren
+sys.path.insert(0, "/home/claude-dev/AegisSight-Monitor/src")
+try:
+    from services.post_refresh_qc import normalize_german_umlauts
+except ImportError:
+    sys.path.insert(0, "/home/claude-dev/AegisSight-Monitor-staging/src")
+    from services.post_refresh_qc import normalize_german_umlauts
+
+
+def main():
+    ap = argparse.ArgumentParser()
+    ap.add_argument("--db", required=True)
+    ap.add_argument("--no-backup", action="store_true")
+    ap.add_argument("--dry-run", action="store_true")
+    args = ap.parse_args()
+
+    if not args.no_backup and not args.dry_run:
+        ts = datetime.now().strftime("%Y-%m-%d-%H%M")
+        backup_path = f"{args.db}.umlaut-bak-{ts}"
+        shutil.copy2(args.db, backup_path)
+        print(f"Backup angelegt: {backup_path}")
+
+    db = sqlite3.connect(args.db)
+    db.row_factory = sqlite3.Row
+
+    cur = db.execute(
+        """SELECT id, language, headline, headline_de, content_original, content_de
+           FROM articles"""
+    )
+    rows = cur.fetchall()
+    print(f"Artikel gesamt: {len(rows)}")
+
+    affected = 0
+    total_replacements = 0
+    by_field = {"headline": 0, "headline_de": 0, "content_original": 0, "content_de": 0}
+    samples = []
+
+    for r in rows:
+        rid = r["id"]
+        is_de = (r["language"] or "").lower() == "de"
+        updates = {}
+
+        if r["headline_de"]:
+            new, n = normalize_german_umlauts(r["headline_de"])
+            if n > 0:
+                updates["headline_de"] = new
+                by_field["headline_de"] += n
+                total_replacements += n
+        if r["content_de"]:
+            new, n = normalize_german_umlauts(r["content_de"])
+            if n > 0:
+                updates["content_de"] = new
+                by_field["content_de"] += n
+                total_replacements += n
+        if is_de:
+            if r["headline"]:
+                new, n = normalize_german_umlauts(r["headline"])
+                if n > 0:
+                    updates["headline"] = new
+                    by_field["headline"] += n
+                    total_replacements += n
+            if r["content_original"]:
+                new, n = normalize_german_umlauts(r["content_original"])
+                if n > 0:
+                    updates["content_original"] = new
+                    by_field["content_original"] += n
+                    total_replacements += n
+
+        if updates:
+            affected += 1
+            if len(samples) < 5:
+                # Sample fuer Bericht
+                first_field = next(iter(updates))
+                old_val = r[first_field]
+                new_val = updates[first_field]
+                samples.append((rid, first_field, old_val[:120], new_val[:120]))
+            if not args.dry_run:
+                set_clause = ", ".join(f"{k} = ?" for k in updates)
+                values = list(updates.values()) + [rid]
+                db.execute(f"UPDATE articles SET {set_clause} WHERE id = ?", values)
+
+    if not args.dry_run:
+        db.commit()
+
+    print()
+    print(f"Betroffene Artikel: {affected}")
+    print(f"Gesamt-Wortersetzungen: {total_replacements}")
+    for k, v in by_field.items():
+        print(f"  {k}: {v}")
+    print()
+    print("=== Sample (vorher -> nachher) ===")
+    for rid, field, old, new in samples:
+        print(f"  [{rid}] {field}:")
+        print(f"    BEFORE: {old!r}")
+        print(f"    AFTER:  {new!r}")
+
+    db.close()
+    print()
+    print("Fertig.")
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())