feat(pii): Stage 1b curated German gazetteer packs (names, cities)

internal/config/validate.go

@@ -387,6 +387,7 @@ func (c *Config) validate() error {
 		knownPacks := map[string]bool{
 			"company-forms": true,
 			"de-cities":     true,
+			"de-firstnames": true,
 		}
 		for _, packName := range gaz.Packs {
 			if !knownPacks[packName] {

internal/pii/gazetteer.go

@@ -296,6 +296,7 @@ func isWordBoundary(runes []rune, startRune, endRune int) bool {
 var embeddedPackRegistry = map[string]string{
 	"company-forms": "gazetteer_packs/company-forms.txt",
 	"de-cities":     "gazetteer_packs/de-cities.txt",
+	"de-firstnames": "gazetteer_packs/de-firstnames.txt",
 }
 
 // loadEmbeddedPack loads a Gazetteer from an embedded pack file by name.

internal/pii/gazetteer_packs/de-cities.txt

@@ -1,25 +1,54 @@
 # locale: de
 # type: CITY
 #
-# Curated set of unambiguous German city names for use as a template pack.
-# Only cities whose names are not common German words are included here to
-# minimise false positives. Avoid common-word cities such as "Essen", "Hof",
-# "Landsberg", "Neustadt", "Bad" (prefix), or "Berg" (suffix component).
+# Curated list of major German cities compiled from public knowledge and
+# general reference sources. This file is an original curated work — it is
+# NOT a copy of any licensed dataset (no GeoNames, OSM, or proprietary
+# database). Conservative curation: cities whose names are also common
+# everyday German words where the non-city meaning dominates are excluded to
+# minimise false positives. When in doubt, the city is left out.
+# Precision over recall.
 #
-# This pack is intentionally small. Operators should supplement it with their
-# own domain-specific gazetteer files via pii.gazetteer.dirs.
+# This pack is OPT-IN: it is only loaded when "de-cities" is listed in
+# settings.pii.gazetteer.packs.
+#
+# Excluded (common word meaning dominates or high FP risk):
+#   Essen (essen = to eat)
+#   Hof (Hof = courtyard/farm)
+#   Berg (Berg = mountain)
+#   Bad (Bad = bath/spa — prefix in many city names, but standalone "Bad" is a word)
+#   Neustadt (Neue Stadt = new city, very generic)
+#   Landsberg (Landsberg = generic compound, ambiguous)
+#   Hameln (marginal, very small)
+#   Siegen (siegen = to win/prevail — verb meaning dominates in prose)
+#   Singen (singen = to sing — verb meaning dominates in prose)
+#
+# German spelling with umlauts is used (the gazetteer folds case per-rune).
+# ASCII variants (Muenchen, Nuernberg, Koeln, Duesseldorf) are intentionally
+# omitted — they are rarely used in German-language text and add FP surface.
+#
+# ── Top tier (largest cities, unambiguous names) ─────────────────────────────
+Berlin
+Hamburg
+München
+Köln
+Frankfurt am Main
+Stuttgart
 Düsseldorf
 Dortmund
-Stuttgart
+Leipzig
+Dresden
 Hannover
 Nürnberg
+Bremen
 Duisburg
 Bochum
 Wuppertal
 Bielefeld
 Bonn
 Mannheim
 Karlsruhe
+# ── Large cities ─────────────────────────────────────────────────────────────
 Wiesbaden
 Münster
 Augsburg
@@ -67,7 +96,6 @@ Jena
 Remscheid
 Erlangen
 Moers
-Siegen
 Salzgitter
 Gütersloh
 Hildesheim
@@ -80,3 +108,52 @@ Witten
 Zwickau
 Dessau
 Wilhelmshaven
+# ── Mid-tier cities (unambiguous names) ─────────────────────────────────────
+Wolfenbüttel
+Ratingen
+Lünen
+Velbert
+Minden
+Villingen-Schwenningen
+Hanau
+Koblenz
+Tübingen
+Flensburg
+Gießen
+Reutlingen
+Esslingen am Neckar
+Offenburg
+Bayreuth
+Bamberg
+Aschaffenburg
+Straubing
+Passau
+Landshut
+Rosenheim
+Kempten
+Konstanz
+Friedrichshafen
+Ravensburg
+Sindelfingen
+Böblingen
+Ludwigsburg
+Schwäbisch Gmünd
+Aalen
+Heidenheim
+Weimar
+Plauen
+Wismar
+Greifswald
+Stralsund
+Frankfurt an der Oder
+Brandenburg an der Havel
+Dessau-Roßlau
+Halle an der Saale
+Halberstadt
+Stendal
+Paderborn
+Detmold
+Soest
+Arnsberg
+Herne
+Unna