diff --git a/components/settings/AiSections.tsx b/components/settings/AiSections.tsx
index 74ce6a14..47e20eed 100644
--- a/components/settings/AiSections.tsx
+++ b/components/settings/AiSections.tsx
@@ -21,6 +21,7 @@ import { AdaptiveAiHardwarePanel } from './AdaptiveAiHardwarePanel';
 import { AiProviderCard } from './AiProviderCard';
 import { GpuMetricsPanel } from './GpuMetricsPanel';
 import { LocalAiDownloadProgress } from './LocalAiDownloadProgress';
+import { OllamaDevicePull } from './OllamaDevicePull';
 import { ToggleSwitch } from './SettingsShared';
 
 const KNOWN_OLLAMA_MODELS = new Set([
@@ -334,10 +335,23 @@ export const AdvancedAiSection: FC = () => {
               {t('settings.advancedAi.modelDescription')}
             </p>
             {settings.advancedAi.provider === 'ollama' && (
-              <p className="text-xs text-[var(--sc-text-muted)] mb-3">
-                {t('settings.advancedAi.modelRecommendationsIntro')}{' '}
-                {RECOMMENDED_OLLAMA_MODEL_IDS.join(', ')}
-              </p>
+              <>
+                <p className="text-xs text-[var(--sc-text-muted)] mb-3">
+                  {t('settings.advancedAi.modelRecommendationsIntro')}{' '}
+                  {RECOMMENDED_OLLAMA_MODEL_IDS.join(', ')}
+                </p>
+                {/* QNBS-v3: device-aware recommendation + one-click pull (PR E). */}
+                <OllamaDevicePull
+                  baseUrl={settings.advancedAi.ollamaBaseUrl}
+                  onUseModel={(tag) => {
+                    handleSettingChange('advancedAi', { ...settings.advancedAi, model: tag });
+                    // QNBS-v3: recommended tags aren't in the known-options list, so applying one takes
+                    // the custom-model path — sync the custom input so the selector isn't left stale.
+                    if (isCustomOllamaModel(tag)) setCustomModelInput(tag.replace(/^ollama\//, ''));
+                  }}
+                  t={t}
+                />
+              </>
             )}
             <Select
               id="settings-ai-model"
diff --git a/components/settings/OllamaDevicePull.tsx b/components/settings/OllamaDevicePull.tsx
new file mode 100644
index 00000000..7946877e
--- /dev/null
+++ b/components/settings/OllamaDevicePull.tsx
@@ -0,0 +1,166 @@
+/**
+ * OllamaDevicePull — device-aware Ollama model recommendation + one-click pull.
+ * QNBS-v3: profiles the device (getHealthReport) → getOllamaModelForDevice picks a tag, then
+ * pullOllamaModel streams /api/pull progress with cancel + error-retry. Self-contained so AiSections
+ * stays lean; all network failure modes (unreachable server, inline error, abort) are surfaced.
+ */
+
+import { type FC, useEffect, useRef, useState } from 'react';
+import type { useTranslation } from '../../hooks/useTranslation';
+import { getHealthReport } from '../../services/ai/deviceHealthService';
+import {
+  getOllamaModelForDevice,
+  type OllamaDeviceRecommendation,
+} from '../../services/ai/modelRecommendations';
+import { logger } from '../../services/logger';
+import { type OllamaPullProgress, pullOllamaModel } from '../../services/ollamaService';
+import { Badge } from '../ui/Badge';
+import { Button } from '../ui/Button';
+
+type Translate = ReturnType<typeof useTranslation>['t'];
+
+interface OllamaDevicePullProps {
+  baseUrl: string;
+  /** Apply the pulled model as the active Ollama model (e.g. set advancedAi.model). */
+  onUseModel: (ollamaModelTag: string) => void;
+  t: Translate;
+}
+
+type PullState = 'idle' | 'pulling' | 'success' | 'error';
+
+export const OllamaDevicePull: FC<OllamaDevicePullProps> = ({ baseUrl, onUseModel, t }) => {
+  const [rec, setRec] = useState<OllamaDeviceRecommendation | null>(null);
+  const [state, setState] = useState<PullState>('idle');
+  const [progress, setProgress] = useState(0);
+  const [statusText, setStatusText] = useState('');
+  const [error, setError] = useState('');
+  const abortRef = useRef<AbortController | null>(null);
+
+  // QNBS-v3: profile the device once on mount; failure is non-fatal (just no recommendation chip).
+  useEffect(() => {
+    let active = true;
+    void getHealthReport()
+      .then((report) => {
+        if (active) setRec(getOllamaModelForDevice(report));
+      })
+      .catch((err: unknown) => logger.warn('Device profile for Ollama recommendation failed', err));
+    return () => {
+      active = false;
+      abortRef.current?.abort();
+    };
+  }, []);
+
+  if (!rec) return null;
+
+  const onProgress = (p: OllamaPullProgress) => {
+    setStatusText(p.status);
+    if (typeof p.progress === 'number') setProgress(p.progress);
+  };
+
+  const startPull = async () => {
+    // QNBS-v3: in-flight guard — never start a second pull while one is running, so two controllers
+    // can't race (and so a stale finally can't clear a newer request's abort ref).
+    if (abortRef.current) return;
+    const controller = new AbortController();
+    abortRef.current = controller;
+    setState('pulling');
+    setProgress(0);
+    setError('');
+    try {
+      await pullOllamaModel(rec.modelId, { baseUrl, signal: controller.signal, onProgress });
+      setState('success');
+    } catch (err) {
+      // QNBS-v3: an aborted pull returns to idle (user-initiated cancel), not the error state.
+      if ((err as { name?: string })?.name === 'AbortError') {
+        setState('idle');
+        return;
+      }
+      setError(err instanceof Error ? err.message : String(err));
+      setState('error');
+    } finally {
+      // QNBS-v3: only clear the ref if it still points to THIS controller — defends against a late
+      // resolution nulling a newer pull's controller and breaking its Cancel.
+      if (abortRef.current === controller) abortRef.current = null;
+    }
+  };
+
+  const pct = Math.round(progress * 100);
+
+  return (
+    <div className="mb-3 rounded-sc-md border border-[var(--sc-border-subtle)] p-3 space-y-2">
+      <div className="flex items-center gap-2 flex-wrap">
+        <span className="text-xs font-medium text-[var(--sc-text-secondary)]">
+          {t('settings.advancedAi.ollamaPull.recommendedFor')}
+        </span>
+        <code className="text-xs text-[var(--sc-text-primary)]">{rec.modelId}</code>
+        <Badge variant="new">{t(`settings.advancedAi.ollamaPull.tier.${rec.tier}`)}</Badge>
+        <span className="text-[10px] text-[var(--sc-text-muted)]">{rec.sizeHint}</span>
+      </div>
+
+      {rec.downgradedForBattery && (
+        <p className="text-[10px] text-[var(--sc-warning-fg)]">
+          {t('settings.advancedAi.ollamaPull.batteryNote')}
+        </p>
+      )}
+
+      {state === 'pulling' && (
+        <div className="space-y-1">
+          <div
+            className="h-1.5 w-full overflow-hidden rounded-full bg-[var(--sc-surface-overlay)]/40"
+            role="progressbar"
+            aria-valuenow={pct}
+            aria-valuemin={0}
+            aria-valuemax={100}
+            aria-label={t('settings.advancedAi.ollamaPull.pulling')}
+          >
+            <div
+              className="h-full bg-[var(--sc-accent)] transition-[width]"
+              style={{ width: `${pct}%` }}
+            />
+          </div>
+          <p className="text-[10px] text-[var(--sc-text-muted)]">
+            {statusText} {pct > 0 ? `· ${pct}%` : ''}
+          </p>
+        </div>
+      )}
+
+      {state === 'error' && (
+        <p className="text-[10px] text-[var(--sc-danger-fg)]" role="alert">
+          {error}
+        </p>
+      )}
+
+      <div className="flex items-center gap-2">
+        {state === 'idle' && (
+          <Button size="sm" variant="secondary" onClick={() => void startPull()}>
+            {t('settings.advancedAi.ollamaPull.pull')}
+          </Button>
+        )}
+        {state === 'pulling' && (
+          <Button size="sm" variant="ghost" onClick={() => abortRef.current?.abort()}>
+            {t('common.cancel')}
+          </Button>
+        )}
+        {state === 'error' && (
+          <Button size="sm" variant="secondary" onClick={() => void startPull()}>
+            {t('settings.advancedAi.ollamaPull.retry')}
+          </Button>
+        )}
+        {state === 'success' && (
+          <>
+            <span className="text-xs text-[var(--sc-success-fg)]">
+              {t('settings.advancedAi.ollamaPull.installed')}
+            </span>
+            <Button
+              size="sm"
+              variant="secondary"
+              onClick={() => onUseModel(`ollama/${rec.modelId}`)}
+            >
+              {t('settings.advancedAi.ollamaPull.useModel')}
+            </Button>
+          </>
+        )}
+      </div>
+    </div>
+  );
+};
diff --git a/locales/ar/settings.json b/locales/ar/settings.json
index 6b72bfba..383ff21e 100644
--- a/locales/ar/settings.json
+++ b/locales/ar/settings.json
@@ -64,6 +64,16 @@
   "settings.advancedAi.model": "نموذج الذكاء الاصطناعي",
   "settings.advancedAi.modelDescription": "نموذج الذكاء الاصطناعي المُستخدَم لتوليد المشاهد والتحليل والميزات الأخرى المدعومة بالذكاء الاصطناعي. يعتمد توفّر النموذج على المزوّد المختار.",
   "settings.advancedAi.modelRecommendationsIntro": "نقاط تفتيش محلية شائعة (تحقّق من الأسماء في بيئة تشغيلك):",
+  "settings.advancedAi.ollamaPull.recommendedFor": "موصى به لجهازك:",
+  "settings.advancedAi.ollamaPull.pull": "تنزيل النموذج",
+  "settings.advancedAi.ollamaPull.pulling": "جارٍ تنزيل النموذج…",
+  "settings.advancedAi.ollamaPull.retry": "إعادة المحاولة",
+  "settings.advancedAi.ollamaPull.installed": "مثبَّت",
+  "settings.advancedAi.ollamaPull.useModel": "استخدام هذا النموذج",
+  "settings.advancedAi.ollamaPull.batteryNote": "تم تقليل الحجم درجة واحدة لتوفير البطارية.",
+  "settings.advancedAi.ollamaPull.tier.high-end": "فائق الأداء",
+  "settings.advancedAi.ollamaPull.tier.mid-range": "متوسط",
+  "settings.advancedAi.ollamaPull.tier.low-end": "أساسي",
   "settings.advancedAi.ragModeHint": "يمزج الوضع الهجين مطابقة الكلمات المفتاحية بالفهم الدلالي — الأفضل لمعظم القصص. أعد بناء الفهرس بعد التعديلات الكبيرة.",
   "settings.advancedAi.ragModeHybrid": "هجين (دلالي + معجمي، موصى به)",
   "settings.advancedAi.ragModeLabel": "وضع استرجاع السياق",
diff --git a/locales/de/settings.json b/locales/de/settings.json
index 019cd01c..1dc6b92f 100644
--- a/locales/de/settings.json
+++ b/locales/de/settings.json
@@ -63,6 +63,16 @@
   "settings.advancedAi.model": "KI-Modell",
   "settings.advancedAi.modelDescription": "Das KI-Modell für Szenengeneration, Analyse und andere KI-Funktionen. Die Verfügbarkeit hängt vom gewählten Anbieter ab.",
   "settings.advancedAi.modelRecommendationsIntro": "Beliebte lokale Checkpoints (Namen im Runtime überprüfen):",
+  "settings.advancedAi.ollamaPull.recommendedFor": "Für dein Gerät empfohlen:",
+  "settings.advancedAi.ollamaPull.pull": "Modell laden",
+  "settings.advancedAi.ollamaPull.pulling": "Modell wird geladen…",
+  "settings.advancedAi.ollamaPull.retry": "Erneut versuchen",
+  "settings.advancedAi.ollamaPull.installed": "Installiert",
+  "settings.advancedAi.ollamaPull.useModel": "Dieses Modell verwenden",
+  "settings.advancedAi.ollamaPull.batteryNote": "Eine Größe kleiner gewählt, um Akku zu sparen.",
+  "settings.advancedAi.ollamaPull.tier.high-end": "High-End",
+  "settings.advancedAi.ollamaPull.tier.mid-range": "Mittelklasse",
+  "settings.advancedAi.ollamaPull.tier.low-end": "Einsteiger",
   "settings.advancedAi.ragModeHint": "Der Hybrid-Modus verbindet Keyword-Suche mit semantischem Verständnis — ideal für die meisten Geschichten. Index nach großen Änderungen neu aufbauen.",
   "settings.advancedAi.ragModeHybrid": "Hybrid (semantisch + lexikalisch, empfohlen)",
   "settings.advancedAi.ragModeLabel": "Kontext-Abrufmodus",
diff --git a/locales/el/settings.json b/locales/el/settings.json
index e9220941..8227d7ca 100644
--- a/locales/el/settings.json
+++ b/locales/el/settings.json
@@ -64,6 +64,16 @@
   "settings.advancedAi.model": "Μοντέλο AI",
   "settings.advancedAi.modelDescription": "Το μοντέλο AI χρησιμοποιείται για τη δημιουργία σκηνής, την ανάλυση και άλλα χαρακτηριστικά που τροφοδοτούνται από AI. Η διαθεσιμότητα του μοντέλου εξαρτάται από τον πάροχο που έχετε επιλέξει.",
   "settings.advancedAi.modelRecommendationsIntro": "Δημοφιλή τοπικά σημεία ελέγχου (επαληθεύστε τα ονόματα στο χρόνο εκτέλεσης):",
+  "settings.advancedAi.ollamaPull.recommendedFor": "Προτείνεται για τη συσκευή σας:",
+  "settings.advancedAi.ollamaPull.pull": "Λήψη μοντέλου",
+  "settings.advancedAi.ollamaPull.pulling": "Λήψη μοντέλου…",
+  "settings.advancedAi.ollamaPull.retry": "Επανάληψη",
+  "settings.advancedAi.ollamaPull.installed": "Εγκαταστάθηκε",
+  "settings.advancedAi.ollamaPull.useModel": "Χρήση αυτού του μοντέλου",
+  "settings.advancedAi.ollamaPull.batteryNote": "Μειώθηκε ένα μέγεθος για εξοικονόμηση μπαταρίας.",
+  "settings.advancedAi.ollamaPull.tier.high-end": "Υψηλών επιδόσεων",
+  "settings.advancedAi.ollamaPull.tier.mid-range": "Μεσαία",
+  "settings.advancedAi.ollamaPull.tier.low-end": "Βασική",
   "settings.advancedAi.ragModeHint": "Η υβριδική λειτουργία συνδυάζει την αντιστοίχιση λέξεων-κλειδιών με τη σημασιολογική κατανόηση — το καλύτερο για τις περισσότερες ιστορίες. Δημιουργήστε ξανά το ευρετήριο μετά από μεγάλες επεξεργασίες.",
   "settings.advancedAi.ragModeHybrid": "Υβριδικό (σημασιολογικό + λεξιλογικό, συνιστώμενο)",
   "settings.advancedAi.ragModeLabel": "Λειτουργία ανάκτησης περιβάλλοντος",
diff --git a/locales/en/settings.json b/locales/en/settings.json
index a6ba5db3..e1a241e0 100644
--- a/locales/en/settings.json
+++ b/locales/en/settings.json
@@ -63,6 +63,16 @@
   "settings.advancedAi.model": "AI Model",
   "settings.advancedAi.modelDescription": "The AI model used for scene generation, analysis, and other AI-powered features. Model availability depends on your selected provider.",
   "settings.advancedAi.modelRecommendationsIntro": "Popular local checkpoints (verify names in your runtime):",
+  "settings.advancedAi.ollamaPull.recommendedFor": "Recommended for your device:",
+  "settings.advancedAi.ollamaPull.pull": "Pull model",
+  "settings.advancedAi.ollamaPull.pulling": "Downloading model…",
+  "settings.advancedAi.ollamaPull.retry": "Retry",
+  "settings.advancedAi.ollamaPull.installed": "Installed",
+  "settings.advancedAi.ollamaPull.useModel": "Use this model",
+  "settings.advancedAi.ollamaPull.batteryNote": "Stepped down a size to save battery.",
+  "settings.advancedAi.ollamaPull.tier.high-end": "High-end",
+  "settings.advancedAi.ollamaPull.tier.mid-range": "Mid-range",
+  "settings.advancedAi.ollamaPull.tier.low-end": "Low-end",
   "settings.advancedAi.ragModeHint": "Hybrid mode blends keyword matching with semantic understanding — best for most stories. Rebuild the index after large edits.",
   "settings.advancedAi.ragModeHybrid": "Hybrid (semantic + lexical, recommended)",
   "settings.advancedAi.ragModeLabel": "Context retrieval mode",
diff --git a/locales/es/settings.json b/locales/es/settings.json
index d8c8c3d4..b028e91f 100644
--- a/locales/es/settings.json
+++ b/locales/es/settings.json
@@ -63,6 +63,16 @@
   "settings.advancedAi.model": "Modelo de IA",
   "settings.advancedAi.modelDescription": "El modelo de IA usado para generación de escenas, análisis y otras funciones. La disponibilidad depende del proveedor seleccionado.",
   "settings.advancedAi.modelRecommendationsIntro": "Puntos de control locales populares (verifica los nombres en tu entorno de ejecución):",
+  "settings.advancedAi.ollamaPull.recommendedFor": "Recomendado para tu dispositivo:",
+  "settings.advancedAi.ollamaPull.pull": "Descargar modelo",
+  "settings.advancedAi.ollamaPull.pulling": "Descargando modelo…",
+  "settings.advancedAi.ollamaPull.retry": "Reintentar",
+  "settings.advancedAi.ollamaPull.installed": "Instalado",
+  "settings.advancedAi.ollamaPull.useModel": "Usar este modelo",
+  "settings.advancedAi.ollamaPull.batteryNote": "Se eligió un tamaño menor para ahorrar batería.",
+  "settings.advancedAi.ollamaPull.tier.high-end": "Gama alta",
+  "settings.advancedAi.ollamaPull.tier.mid-range": "Gama media",
+  "settings.advancedAi.ollamaPull.tier.low-end": "Gama baja",
   "settings.advancedAi.ragModeHint": "El modo híbrido combina búsqueda por palabras clave con comprensión semántica — ideal para la mayoría de historias. Reconstruye el índice tras grandes cambios.",
   "settings.advancedAi.ragModeHybrid": "Híbrido (semántico + léxico, recomendado)",
   "settings.advancedAi.ragModeLabel": "Modo de recuperación de contexto",
diff --git a/locales/eu/settings.json b/locales/eu/settings.json
index 914d7878..8022cb1b 100644
--- a/locales/eu/settings.json
+++ b/locales/eu/settings.json
@@ -64,6 +64,16 @@
   "settings.advancedAi.model": "AI eredua",
   "settings.advancedAi.modelDescription": "Eszena sortzeko, analisirako eta AI bidezko beste funtzio batzuetarako erabiltzen den AI eredua. Ereduaren erabilgarritasuna hautatutako hornitzailearen araberakoa da.",
   "settings.advancedAi.modelRecommendationsIntro": "Tokiko kontrol-puntu ezagunak (egiaztatu izenak zure exekuzioan):",
+  "settings.advancedAi.ollamaPull.recommendedFor": "Zure gailurako gomendatua:",
+  "settings.advancedAi.ollamaPull.pull": "Deskargatu eredua",
+  "settings.advancedAi.ollamaPull.pulling": "Eredua deskargatzen…",
+  "settings.advancedAi.ollamaPull.retry": "Saiatu berriro",
+  "settings.advancedAi.ollamaPull.installed": "Instalatuta",
+  "settings.advancedAi.ollamaPull.useModel": "Erabili eredu hau",
+  "settings.advancedAi.ollamaPull.batteryNote": "Tamaina bat jaitsi da bateria aurrezteko.",
+  "settings.advancedAi.ollamaPull.tier.high-end": "Goi-mailakoa",
+  "settings.advancedAi.ollamaPull.tier.mid-range": "Erdi-mailakoa",
+  "settings.advancedAi.ollamaPull.tier.low-end": "Oinarrizkoa",
   "settings.advancedAi.ragModeHint": "Modu hibridoak gako-hitzekin bat etortzea ulermen semantikoarekin nahasten du, istorio gehienetarako onena. Berreraiki indizea aldaketa handien ondoren.",
   "settings.advancedAi.ragModeHybrid": "Hibridoa (semantikoa + lexikoa, gomendagarria)",
   "settings.advancedAi.ragModeLabel": "Testuingurua berreskuratzeko modua",
diff --git a/locales/fa/settings.json b/locales/fa/settings.json
index 48c639e3..cb9b276c 100644
--- a/locales/fa/settings.json
+++ b/locales/fa/settings.json
@@ -64,6 +64,16 @@
   "settings.advancedAi.model": "مدل هوش مصنوعی",
   "settings.advancedAi.modelDescription": "مدل هوش مصنوعی برای تولید صحنه، تحلیل و سایر ویژگی‌های مبتنی بر هوش مصنوعی استفاده می‌شود. در دسترس بودن مدل به ارائه دهنده انتخابی شما بستگی دارد.",
   "settings.advancedAi.modelRecommendationsIntro": "نقاط بازرسی محلی محبوب (اسامی را در زمان اجرا تأیید کنید):",
+  "settings.advancedAi.ollamaPull.recommendedFor": "پیشنهادشده برای دستگاه شما:",
+  "settings.advancedAi.ollamaPull.pull": "دریافت مدل",
+  "settings.advancedAi.ollamaPull.pulling": "در حال دانلود مدل…",
+  "settings.advancedAi.ollamaPull.retry": "تلاش دوباره",
+  "settings.advancedAi.ollamaPull.installed": "نصب‌شده",
+  "settings.advancedAi.ollamaPull.useModel": "استفاده از این مدل",
+  "settings.advancedAi.ollamaPull.batteryNote": "برای صرفه‌جویی در باتری یک اندازه کوچک‌تر شد.",
+  "settings.advancedAi.ollamaPull.tier.high-end": "رده‌بالا",
+  "settings.advancedAi.ollamaPull.tier.mid-range": "میان‌رده",
+  "settings.advancedAi.ollamaPull.tier.low-end": "رده‌پایین",
   "settings.advancedAi.ragModeHint": "حالت ترکیبی تطبیق کلمات کلیدی را با درک معنایی ترکیب می‌کند - بهترین برای اکثر داستان‌ها. پس از ویرایش های بزرگ، فهرست را دوباره بسازید.",
   "settings.advancedAi.ragModeHybrid": "ترکیبی (معنی + واژگانی، توصیه شده)",
   "settings.advancedAi.ragModeLabel": "حالت بازیابی متن",
diff --git a/locales/fi/settings.json b/locales/fi/settings.json
index 35d4f0d6..6e3bf1c9 100644
--- a/locales/fi/settings.json
+++ b/locales/fi/settings.json
@@ -64,6 +64,16 @@
   "settings.advancedAi.model": "AI malli",
   "settings.advancedAi.modelDescription": "Tekoälymalli, jota käytetään kohtausten luomiseen, analysointiin ja muihin tekoälyyn perustuviin ominaisuuksiin. Mallin saatavuus riippuu valitsemastasi toimittajasta.",
   "settings.advancedAi.modelRecommendationsIntro": "Suositut paikalliset tarkistuspisteet (tarkista nimet ajon aikana):",
+  "settings.advancedAi.ollamaPull.recommendedFor": "Suositus laitteellesi:",
+  "settings.advancedAi.ollamaPull.pull": "Lataa malli",
+  "settings.advancedAi.ollamaPull.pulling": "Ladataan mallia…",
+  "settings.advancedAi.ollamaPull.retry": "Yritä uudelleen",
+  "settings.advancedAi.ollamaPull.installed": "Asennettu",
+  "settings.advancedAi.ollamaPull.useModel": "Käytä tätä mallia",
+  "settings.advancedAi.ollamaPull.batteryNote": "Pienennetty kokoa akun säästämiseksi.",
+  "settings.advancedAi.ollamaPull.tier.high-end": "Tehokas",
+  "settings.advancedAi.ollamaPull.tier.mid-range": "Keskitaso",
+  "settings.advancedAi.ollamaPull.tier.low-end": "Kevyt",
   "settings.advancedAi.ragModeHint": "Hybriditila yhdistää avainsanahaun semanttiseen ymmärrykseen – paras useimpiin tarinoihin. Rakenna hakemisto uudelleen suurten muokkausten jälkeen.",
   "settings.advancedAi.ragModeHybrid": "Hybridi (semanttinen + leksikaalinen, suositeltava)",
   "settings.advancedAi.ragModeLabel": "Kontekstin hakutila",
diff --git a/locales/fr/settings.json b/locales/fr/settings.json
index a80ff67c..177cc1fc 100644
--- a/locales/fr/settings.json
+++ b/locales/fr/settings.json
@@ -63,6 +63,16 @@
   "settings.advancedAi.model": "Modèle d'IA",
   "settings.advancedAi.modelDescription": "Le modèle IA utilisé pour la génération de scènes, l'analyse et d'autres fonctionnalités IA. La disponibilité dépend du fournisseur sélectionné.",
   "settings.advancedAi.modelRecommendationsIntro": "Points de contrôle locaux populaires (vérifiez les noms dans votre environnement d'exécution) :",
+  "settings.advancedAi.ollamaPull.recommendedFor": "Recommandé pour votre appareil :",
+  "settings.advancedAi.ollamaPull.pull": "Télécharger le modèle",
+  "settings.advancedAi.ollamaPull.pulling": "Téléchargement du modèle…",
+  "settings.advancedAi.ollamaPull.retry": "Réessayer",
+  "settings.advancedAi.ollamaPull.installed": "Installé",
+  "settings.advancedAi.ollamaPull.useModel": "Utiliser ce modèle",
+  "settings.advancedAi.ollamaPull.batteryNote": "Taille réduite pour économiser la batterie.",
+  "settings.advancedAi.ollamaPull.tier.high-end": "Haut de gamme",
+  "settings.advancedAi.ollamaPull.tier.mid-range": "Milieu de gamme",
+  "settings.advancedAi.ollamaPull.tier.low-end": "Entrée de gamme",
   "settings.advancedAi.ragModeHint": "Le mode hybride associe recherche par mots-clés et compréhension sémantique — idéal pour la plupart des histoires. Reconstruisez l'index après de grandes modifications.",
   "settings.advancedAi.ragModeHybrid": "Hybride (sémantique + lexical, recommandé)",
   "settings.advancedAi.ragModeLabel": "Mode de récupération du contexte",
diff --git a/locales/he/settings.json b/locales/he/settings.json
index c4eb7f6a..62437475 100644
--- a/locales/he/settings.json
+++ b/locales/he/settings.json
@@ -64,6 +64,16 @@
   "settings.advancedAi.model": "מודל AI",
   "settings.advancedAi.modelDescription": "מודל ה‑AI המשמש ליצירת סצנות, ניתוח ותכונות מופעלות‑AI נוספות. זמינות המודל תלויה בספק שבחרתם.",
   "settings.advancedAi.modelRecommendationsIntro": "נקודות ביקורת מקומיות פופולריות (אמתו שמות בזמן הריצה שלכם):",
+  "settings.advancedAi.ollamaPull.recommendedFor": "מומלץ למכשיר שלך:",
+  "settings.advancedAi.ollamaPull.pull": "הורדת מודל",
+  "settings.advancedAi.ollamaPull.pulling": "מוריד מודל…",
+  "settings.advancedAi.ollamaPull.retry": "נסה שוב",
+  "settings.advancedAi.ollamaPull.installed": "מותקן",
+  "settings.advancedAi.ollamaPull.useModel": "השתמש במודל הזה",
+  "settings.advancedAi.ollamaPull.batteryNote": "הוקטן בגודל אחד כדי לחסוך בסוללה.",
+  "settings.advancedAi.ollamaPull.tier.high-end": "ביצועים גבוהים",
+  "settings.advancedAi.ollamaPull.tier.mid-range": "בינוני",
+  "settings.advancedAi.ollamaPull.tier.low-end": "בסיסי",
   "settings.advancedAi.ragModeHint": "מצב היברידי משלב התאמת מילות מפתח עם הבנה סמנטית — הטוב ביותר לרוב הסיפורים. בנו מחדש את האינדקס לאחר עריכות גדולות.",
   "settings.advancedAi.ragModeHybrid": "היברידי (סמנטי + לקסיקלי, מומלץ)",
   "settings.advancedAi.ragModeLabel": "מצב אחזור הקשר",
diff --git a/locales/hu/settings.json b/locales/hu/settings.json
index bed1aa02..882d81e5 100644
--- a/locales/hu/settings.json
+++ b/locales/hu/settings.json
@@ -64,6 +64,16 @@
   "settings.advancedAi.model": "AI modell",
   "settings.advancedAi.modelDescription": "A jelenetek generálásához, elemzéséhez és más mesterséges intelligencia által vezérelt funkciókhoz használt AI-modell. A modell elérhetősége a kiválasztott szolgáltatótól függ.",
   "settings.advancedAi.modelRecommendationsIntro": "Népszerű helyi ellenőrzőpontok (ellenőrizze a neveket a futásidőben):",
+  "settings.advancedAi.ollamaPull.recommendedFor": "Ajánlott az eszközödhöz:",
+  "settings.advancedAi.ollamaPull.pull": "Modell letöltése",
+  "settings.advancedAi.ollamaPull.pulling": "Modell letöltése…",
+  "settings.advancedAi.ollamaPull.retry": "Újrapróbálkozás",
+  "settings.advancedAi.ollamaPull.installed": "Telepítve",
+  "settings.advancedAi.ollamaPull.useModel": "Ez a modell használata",
+  "settings.advancedAi.ollamaPull.batteryNote": "Egy mérettel kisebbet választottunk az akkumulátor kímélésére.",
+  "settings.advancedAi.ollamaPull.tier.high-end": "Csúcskategória",
+  "settings.advancedAi.ollamaPull.tier.mid-range": "Középkategória",
+  "settings.advancedAi.ollamaPull.tier.low-end": "Belépőszint",
   "settings.advancedAi.ragModeHint": "A hibrid mód ötvözi a kulcsszóegyezést a szemantikai megértéssel – ez a legjobb a legtöbb történethez. Nagyobb szerkesztések után építse újra az indexet.",
   "settings.advancedAi.ragModeHybrid": "Hibrid (szemantikai + lexikális, ajánlott)",
   "settings.advancedAi.ragModeLabel": "Kontextus visszakeresési mód",
diff --git a/locales/is/settings.json b/locales/is/settings.json
index 3bd8487b..bde2aa6e 100644
--- a/locales/is/settings.json
+++ b/locales/is/settings.json
@@ -64,6 +64,16 @@
   "settings.advancedAi.model": "AI líkan",
   "settings.advancedAi.modelDescription": "Gervigreind líkanið sem er notað fyrir senuframleiðslu, greiningu og aðra gervigreindar eiginleika. Framboð á gerðum fer eftir þjónustuveitunni sem þú hefur valið.",
   "settings.advancedAi.modelRecommendationsIntro": "Vinsælir staðbundnir eftirlitsstöðvar (staðfestu nöfn á keyrslutíma þínum):",
+  "settings.advancedAi.ollamaPull.recommendedFor": "Mælt með fyrir tækið þitt:",
+  "settings.advancedAi.ollamaPull.pull": "Sækja líkan",
+  "settings.advancedAi.ollamaPull.pulling": "Sæki líkan…",
+  "settings.advancedAi.ollamaPull.retry": "Reyna aftur",
+  "settings.advancedAi.ollamaPull.installed": "Uppsett",
+  "settings.advancedAi.ollamaPull.useModel": "Nota þetta líkan",
+  "settings.advancedAi.ollamaPull.batteryNote": "Lækkað um eitt þrep til að spara rafhlöðu.",
+  "settings.advancedAi.ollamaPull.tier.high-end": "Hágæða",
+  "settings.advancedAi.ollamaPull.tier.mid-range": "Millistig",
+  "settings.advancedAi.ollamaPull.tier.low-end": "Grunnstig",
   "settings.advancedAi.ragModeHint": "Hybrid háttur blandar saman leitarorðasamsvörun við merkingarskilning - best fyrir flestar sögur. Endurbyggðu vísitöluna eftir miklar breytingar.",
   "settings.advancedAi.ragModeHybrid": "Blendingur (merkingartækni + orðafræði, mælt með)",
   "settings.advancedAi.ragModeLabel": "Samhengisleitarstilling",
diff --git a/locales/it/settings.json b/locales/it/settings.json
index 0c75c55c..152d2bde 100644
--- a/locales/it/settings.json
+++ b/locales/it/settings.json
@@ -63,6 +63,16 @@
   "settings.advancedAi.model": "Modello IA",
   "settings.advancedAi.modelDescription": "Il modello IA usato per la generazione di scene, analisi e altre funzionalità IA. La disponibilità dipende dal provider selezionato.",
   "settings.advancedAi.modelRecommendationsIntro": "Checkpoint locali popolari (verifica i nomi nel tuo ambiente di esecuzione):",
+  "settings.advancedAi.ollamaPull.recommendedFor": "Consigliato per il tuo dispositivo:",
+  "settings.advancedAi.ollamaPull.pull": "Scarica modello",
+  "settings.advancedAi.ollamaPull.pulling": "Download del modello…",
+  "settings.advancedAi.ollamaPull.retry": "Riprova",
+  "settings.advancedAi.ollamaPull.installed": "Installato",
+  "settings.advancedAi.ollamaPull.useModel": "Usa questo modello",
+  "settings.advancedAi.ollamaPull.batteryNote": "Ridotta la dimensione per risparmiare batteria.",
+  "settings.advancedAi.ollamaPull.tier.high-end": "Fascia alta",
+  "settings.advancedAi.ollamaPull.tier.mid-range": "Fascia media",
+  "settings.advancedAi.ollamaPull.tier.low-end": "Fascia bassa",
   "settings.advancedAi.ragModeHint": "La modalità ibrida combina ricerca per parole chiave e comprensione semantica — ideale per la maggior parte delle storie. Ricostruisci l'indice dopo modifiche importanti.",
   "settings.advancedAi.ragModeHybrid": "Ibrido (semantico + lessicale, consigliato)",
   "settings.advancedAi.ragModeLabel": "Modalità di recupero del contesto",
diff --git a/locales/ja/settings.json b/locales/ja/settings.json
index fb3c457c..a9901cc2 100644
--- a/locales/ja/settings.json
+++ b/locales/ja/settings.json
@@ -64,6 +64,16 @@
   "settings.advancedAi.model": "AIモデル",
   "settings.advancedAi.modelDescription": "シーンの生成、分析、その他の AI を利用した機能に使用される AI モデル。利用可能なモデルは、選択したプロバイダーによって異なります。",
   "settings.advancedAi.modelRecommendationsIntro": "一般的なローカル チェックポイント (ランタイムで名前を確認):",
+  "settings.advancedAi.ollamaPull.recommendedFor": "お使いのデバイスにおすすめ:",
+  "settings.advancedAi.ollamaPull.pull": "モデルを取得",
+  "settings.advancedAi.ollamaPull.pulling": "モデルをダウンロード中…",
+  "settings.advancedAi.ollamaPull.retry": "再試行",
+  "settings.advancedAi.ollamaPull.installed": "インストール済み",
+  "settings.advancedAi.ollamaPull.useModel": "このモデルを使用",
+  "settings.advancedAi.ollamaPull.batteryNote": "バッテリー節約のためサイズを一段下げました。",
+  "settings.advancedAi.ollamaPull.tier.high-end": "ハイエンド",
+  "settings.advancedAi.ollamaPull.tier.mid-range": "ミッドレンジ",
+  "settings.advancedAi.ollamaPull.tier.low-end": "ローエンド",
   "settings.advancedAi.ragModeHint": "ハイブリッド モードでは、キーワードの一致と意味の理解が融合され、ほとんどのストーリーに最適です。大規模な編集後にインデックスを再構築します。",
   "settings.advancedAi.ragModeHybrid": "ハイブリッド (意味論 + 語彙論、推奨)",
   "settings.advancedAi.ragModeLabel": "コンテキスト取得モード",
diff --git a/locales/pt/settings.json b/locales/pt/settings.json
index 908e5b38..50faeb64 100644
--- a/locales/pt/settings.json
+++ b/locales/pt/settings.json
@@ -64,6 +64,16 @@
   "settings.advancedAi.model": "IA Model",
   "settings.advancedAi.modelDescription": "The IA model used for scene generation, analysis, and other IA-powered features. Model availability depends on your selected provider.",
   "settings.advancedAi.modelRecommendationsIntro": "Pontos de verificação locais populares (verifique os nomes em seu tempo de execução):",
+  "settings.advancedAi.ollamaPull.recommendedFor": "Recomendado para o seu dispositivo:",
+  "settings.advancedAi.ollamaPull.pull": "Baixar modelo",
+  "settings.advancedAi.ollamaPull.pulling": "Baixando modelo…",
+  "settings.advancedAi.ollamaPull.retry": "Tentar novamente",
+  "settings.advancedAi.ollamaPull.installed": "Instalado",
+  "settings.advancedAi.ollamaPull.useModel": "Usar este modelo",
+  "settings.advancedAi.ollamaPull.batteryNote": "Reduzido um tamanho para poupar bateria.",
+  "settings.advancedAi.ollamaPull.tier.high-end": "Alto desempenho",
+  "settings.advancedAi.ollamaPull.tier.mid-range": "Intermediário",
+  "settings.advancedAi.ollamaPull.tier.low-end": "Básico",
   "settings.advancedAi.ragModeHint": "O modo híbrido combina correspondência de palavras-chave com compreensão semântica – melhor para a maioria das histórias. Reconstrua o índice após grandes edições.",
   "settings.advancedAi.ragModeHybrid": "Híbrido (semântico + lexical, recomendado)",
   "settings.advancedAi.ragModeLabel": "Modo de recuperação de contexto",
diff --git a/locales/sv/settings.json b/locales/sv/settings.json
index 945cf2fb..3539cf16 100644
--- a/locales/sv/settings.json
+++ b/locales/sv/settings.json
@@ -64,6 +64,16 @@
   "settings.advancedAi.model": "AI-modell",
   "settings.advancedAi.modelDescription": "AI-modellen som används för scengenerering, analys och andra AI-drivna funktioner. Modellens tillgänglighet beror på din valda leverantör.",
   "settings.advancedAi.modelRecommendationsIntro": "Populära lokala kontrollpunkter (verifiera namn under din körning):",
+  "settings.advancedAi.ollamaPull.recommendedFor": "Rekommenderas för din enhet:",
+  "settings.advancedAi.ollamaPull.pull": "Hämta modell",
+  "settings.advancedAi.ollamaPull.pulling": "Hämtar modell…",
+  "settings.advancedAi.ollamaPull.retry": "Försök igen",
+  "settings.advancedAi.ollamaPull.installed": "Installerad",
+  "settings.advancedAi.ollamaPull.useModel": "Använd den här modellen",
+  "settings.advancedAi.ollamaPull.batteryNote": "Sänkte en storlek för att spara batteri.",
+  "settings.advancedAi.ollamaPull.tier.high-end": "Högpresterande",
+  "settings.advancedAi.ollamaPull.tier.mid-range": "Mellanklass",
+  "settings.advancedAi.ollamaPull.tier.low-end": "Instegsnivå",
   "settings.advancedAi.ragModeHint": "Hybridläge blandar sökordsmatchning med semantisk förståelse – bäst för de flesta artiklar. Bygg om indexet efter stora redigeringar.",
   "settings.advancedAi.ragModeHybrid": "Hybrid (semantisk + lexikal, rekommenderas)",
   "settings.advancedAi.ragModeLabel": "Läge för kontexthämtning",
diff --git a/locales/zh/settings.json b/locales/zh/settings.json
index 8e258476..8e2c5f73 100644
--- a/locales/zh/settings.json
+++ b/locales/zh/settings.json
@@ -64,6 +64,16 @@
   "settings.advancedAi.model": "人工智能模型",
   "settings.advancedAi.modelDescription": "用于场景生成、分析和其他人工智能功能的人工智能模型。型号可用性取决于您选择的提供商。",
   "settings.advancedAi.modelRecommendationsIntro": "流行的本地检查点（在运行时验证名称）：",
+  "settings.advancedAi.ollamaPull.recommendedFor": "为您的设备推荐：",
+  "settings.advancedAi.ollamaPull.pull": "拉取模型",
+  "settings.advancedAi.ollamaPull.pulling": "正在下载模型…",
+  "settings.advancedAi.ollamaPull.retry": "重试",
+  "settings.advancedAi.ollamaPull.installed": "已安装",
+  "settings.advancedAi.ollamaPull.useModel": "使用此模型",
+  "settings.advancedAi.ollamaPull.batteryNote": "已降低一档以节省电量。",
+  "settings.advancedAi.ollamaPull.tier.high-end": "高端",
+  "settings.advancedAi.ollamaPull.tier.mid-range": "中端",
+  "settings.advancedAi.ollamaPull.tier.low-end": "低端",
   "settings.advancedAi.ragModeHint": "混合模式将关键词匹配与语义理解相结合——最适合大多数故事。大量编辑后重建索引。",
   "settings.advancedAi.ragModeHybrid": "混合（语义+词汇，推荐）",
   "settings.advancedAi.ragModeLabel": "上下文检索模式",
diff --git a/public/locales/ar/bundle.json b/public/locales/ar/bundle.json
index 7f90c364..db98e3a3 100644
--- a/public/locales/ar/bundle.json
+++ b/public/locales/ar/bundle.json
@@ -1578,6 +1578,16 @@
   "settings.advancedAi.model": "نموذج الذكاء الاصطناعي",
   "settings.advancedAi.modelDescription": "نموذج الذكاء الاصطناعي المُستخدَم لتوليد المشاهد والتحليل والميزات الأخرى المدعومة بالذكاء الاصطناعي. يعتمد توفّر النموذج على المزوّد المختار.",
   "settings.advancedAi.modelRecommendationsIntro": "نقاط تفتيش محلية شائعة (تحقّق من الأسماء في بيئة تشغيلك):",
+  "settings.advancedAi.ollamaPull.recommendedFor": "موصى به لجهازك:",
+  "settings.advancedAi.ollamaPull.pull": "تنزيل النموذج",
+  "settings.advancedAi.ollamaPull.pulling": "جارٍ تنزيل النموذج…",
+  "settings.advancedAi.ollamaPull.retry": "إعادة المحاولة",
+  "settings.advancedAi.ollamaPull.installed": "مثبَّت",
+  "settings.advancedAi.ollamaPull.useModel": "استخدام هذا النموذج",
+  "settings.advancedAi.ollamaPull.batteryNote": "تم تقليل الحجم درجة واحدة لتوفير البطارية.",
+  "settings.advancedAi.ollamaPull.tier.high-end": "فائق الأداء",
+  "settings.advancedAi.ollamaPull.tier.mid-range": "متوسط",
+  "settings.advancedAi.ollamaPull.tier.low-end": "أساسي",
   "settings.advancedAi.ragModeHint": "يمزج الوضع الهجين مطابقة الكلمات المفتاحية بالفهم الدلالي — الأفضل لمعظم القصص. أعد بناء الفهرس بعد التعديلات الكبيرة.",
   "settings.advancedAi.ragModeHybrid": "هجين (دلالي + معجمي، موصى به)",
   "settings.advancedAi.ragModeLabel": "وضع استرجاع السياق",
diff --git a/public/locales/de/bundle.json b/public/locales/de/bundle.json
index de2e6e0c..5f85dfb3 100644
--- a/public/locales/de/bundle.json
+++ b/public/locales/de/bundle.json
@@ -1577,6 +1577,16 @@
   "settings.advancedAi.model": "KI-Modell",
   "settings.advancedAi.modelDescription": "Das KI-Modell für Szenengeneration, Analyse und andere KI-Funktionen. Die Verfügbarkeit hängt vom gewählten Anbieter ab.",
   "settings.advancedAi.modelRecommendationsIntro": "Beliebte lokale Checkpoints (Namen im Runtime überprüfen):",
+  "settings.advancedAi.ollamaPull.recommendedFor": "Für dein Gerät empfohlen:",
+  "settings.advancedAi.ollamaPull.pull": "Modell laden",
+  "settings.advancedAi.ollamaPull.pulling": "Modell wird geladen…",
+  "settings.advancedAi.ollamaPull.retry": "Erneut versuchen",
+  "settings.advancedAi.ollamaPull.installed": "Installiert",
+  "settings.advancedAi.ollamaPull.useModel": "Dieses Modell verwenden",
+  "settings.advancedAi.ollamaPull.batteryNote": "Eine Größe kleiner gewählt, um Akku zu sparen.",
+  "settings.advancedAi.ollamaPull.tier.high-end": "High-End",
+  "settings.advancedAi.ollamaPull.tier.mid-range": "Mittelklasse",
+  "settings.advancedAi.ollamaPull.tier.low-end": "Einsteiger",
   "settings.advancedAi.ragModeHint": "Der Hybrid-Modus verbindet Keyword-Suche mit semantischem Verständnis — ideal für die meisten Geschichten. Index nach großen Änderungen neu aufbauen.",
   "settings.advancedAi.ragModeHybrid": "Hybrid (semantisch + lexikalisch, empfohlen)",
   "settings.advancedAi.ragModeLabel": "Kontext-Abrufmodus",
diff --git a/public/locales/el/bundle.json b/public/locales/el/bundle.json
index c99c9e69..b90b6a56 100644
--- a/public/locales/el/bundle.json
+++ b/public/locales/el/bundle.json
@@ -1578,6 +1578,16 @@
   "settings.advancedAi.model": "Μοντέλο AI",
   "settings.advancedAi.modelDescription": "Το μοντέλο AI χρησιμοποιείται για τη δημιουργία σκηνής, την ανάλυση και άλλα χαρακτηριστικά που τροφοδοτούνται από AI. Η διαθεσιμότητα του μοντέλου εξαρτάται από τον πάροχο που έχετε επιλέξει.",
   "settings.advancedAi.modelRecommendationsIntro": "Δημοφιλή τοπικά σημεία ελέγχου (επαληθεύστε τα ονόματα στο χρόνο εκτέλεσης):",
+  "settings.advancedAi.ollamaPull.recommendedFor": "Προτείνεται για τη συσκευή σας:",
+  "settings.advancedAi.ollamaPull.pull": "Λήψη μοντέλου",
+  "settings.advancedAi.ollamaPull.pulling": "Λήψη μοντέλου…",
+  "settings.advancedAi.ollamaPull.retry": "Επανάληψη",
+  "settings.advancedAi.ollamaPull.installed": "Εγκαταστάθηκε",
+  "settings.advancedAi.ollamaPull.useModel": "Χρήση αυτού του μοντέλου",
+  "settings.advancedAi.ollamaPull.batteryNote": "Μειώθηκε ένα μέγεθος για εξοικονόμηση μπαταρίας.",
+  "settings.advancedAi.ollamaPull.tier.high-end": "Υψηλών επιδόσεων",
+  "settings.advancedAi.ollamaPull.tier.mid-range": "Μεσαία",
+  "settings.advancedAi.ollamaPull.tier.low-end": "Βασική",
   "settings.advancedAi.ragModeHint": "Η υβριδική λειτουργία συνδυάζει την αντιστοίχιση λέξεων-κλειδιών με τη σημασιολογική κατανόηση — το καλύτερο για τις περισσότερες ιστορίες. Δημιουργήστε ξανά το ευρετήριο μετά από μεγάλες επεξεργασίες.",
   "settings.advancedAi.ragModeHybrid": "Υβριδικό (σημασιολογικό + λεξιλογικό, συνιστώμενο)",
   "settings.advancedAi.ragModeLabel": "Λειτουργία ανάκτησης περιβάλλοντος",
diff --git a/public/locales/en/bundle.json b/public/locales/en/bundle.json
index 7d2846a5..40524c01 100644
--- a/public/locales/en/bundle.json
+++ b/public/locales/en/bundle.json
@@ -1577,6 +1577,16 @@
   "settings.advancedAi.model": "AI Model",
   "settings.advancedAi.modelDescription": "The AI model used for scene generation, analysis, and other AI-powered features. Model availability depends on your selected provider.",
   "settings.advancedAi.modelRecommendationsIntro": "Popular local checkpoints (verify names in your runtime):",
+  "settings.advancedAi.ollamaPull.recommendedFor": "Recommended for your device:",
+  "settings.advancedAi.ollamaPull.pull": "Pull model",
+  "settings.advancedAi.ollamaPull.pulling": "Downloading model…",
+  "settings.advancedAi.ollamaPull.retry": "Retry",
+  "settings.advancedAi.ollamaPull.installed": "Installed",
+  "settings.advancedAi.ollamaPull.useModel": "Use this model",
+  "settings.advancedAi.ollamaPull.batteryNote": "Stepped down a size to save battery.",
+  "settings.advancedAi.ollamaPull.tier.high-end": "High-end",
+  "settings.advancedAi.ollamaPull.tier.mid-range": "Mid-range",
+  "settings.advancedAi.ollamaPull.tier.low-end": "Low-end",
   "settings.advancedAi.ragModeHint": "Hybrid mode blends keyword matching with semantic understanding — best for most stories. Rebuild the index after large edits.",
   "settings.advancedAi.ragModeHybrid": "Hybrid (semantic + lexical, recommended)",
   "settings.advancedAi.ragModeLabel": "Context retrieval mode",
diff --git a/public/locales/es/bundle.json b/public/locales/es/bundle.json
index 114c7748..72a2f937 100644
--- a/public/locales/es/bundle.json
+++ b/public/locales/es/bundle.json
@@ -1577,6 +1577,16 @@
   "settings.advancedAi.model": "Modelo de IA",
   "settings.advancedAi.modelDescription": "El modelo de IA usado para generación de escenas, análisis y otras funciones. La disponibilidad depende del proveedor seleccionado.",
   "settings.advancedAi.modelRecommendationsIntro": "Puntos de control locales populares (verifica los nombres en tu entorno de ejecución):",
+  "settings.advancedAi.ollamaPull.recommendedFor": "Recomendado para tu dispositivo:",
+  "settings.advancedAi.ollamaPull.pull": "Descargar modelo",
+  "settings.advancedAi.ollamaPull.pulling": "Descargando modelo…",
+  "settings.advancedAi.ollamaPull.retry": "Reintentar",
+  "settings.advancedAi.ollamaPull.installed": "Instalado",
+  "settings.advancedAi.ollamaPull.useModel": "Usar este modelo",
+  "settings.advancedAi.ollamaPull.batteryNote": "Se eligió un tamaño menor para ahorrar batería.",
+  "settings.advancedAi.ollamaPull.tier.high-end": "Gama alta",
+  "settings.advancedAi.ollamaPull.tier.mid-range": "Gama media",
+  "settings.advancedAi.ollamaPull.tier.low-end": "Gama baja",
   "settings.advancedAi.ragModeHint": "El modo híbrido combina búsqueda por palabras clave con comprensión semántica — ideal para la mayoría de historias. Reconstruye el índice tras grandes cambios.",
   "settings.advancedAi.ragModeHybrid": "Híbrido (semántico + léxico, recomendado)",
   "settings.advancedAi.ragModeLabel": "Modo de recuperación de contexto",
diff --git a/public/locales/eu/bundle.json b/public/locales/eu/bundle.json
index 285d3ae7..46f6357d 100644
--- a/public/locales/eu/bundle.json
+++ b/public/locales/eu/bundle.json
@@ -1578,6 +1578,16 @@
   "settings.advancedAi.model": "AI eredua",
   "settings.advancedAi.modelDescription": "Eszena sortzeko, analisirako eta AI bidezko beste funtzio batzuetarako erabiltzen den AI eredua. Ereduaren erabilgarritasuna hautatutako hornitzailearen araberakoa da.",
   "settings.advancedAi.modelRecommendationsIntro": "Tokiko kontrol-puntu ezagunak (egiaztatu izenak zure exekuzioan):",
+  "settings.advancedAi.ollamaPull.recommendedFor": "Zure gailurako gomendatua:",
+  "settings.advancedAi.ollamaPull.pull": "Deskargatu eredua",
+  "settings.advancedAi.ollamaPull.pulling": "Eredua deskargatzen…",
+  "settings.advancedAi.ollamaPull.retry": "Saiatu berriro",
+  "settings.advancedAi.ollamaPull.installed": "Instalatuta",
+  "settings.advancedAi.ollamaPull.useModel": "Erabili eredu hau",
+  "settings.advancedAi.ollamaPull.batteryNote": "Tamaina bat jaitsi da bateria aurrezteko.",
+  "settings.advancedAi.ollamaPull.tier.high-end": "Goi-mailakoa",
+  "settings.advancedAi.ollamaPull.tier.mid-range": "Erdi-mailakoa",
+  "settings.advancedAi.ollamaPull.tier.low-end": "Oinarrizkoa",
   "settings.advancedAi.ragModeHint": "Modu hibridoak gako-hitzekin bat etortzea ulermen semantikoarekin nahasten du, istorio gehienetarako onena. Berreraiki indizea aldaketa handien ondoren.",
   "settings.advancedAi.ragModeHybrid": "Hibridoa (semantikoa + lexikoa, gomendagarria)",
   "settings.advancedAi.ragModeLabel": "Testuingurua berreskuratzeko modua",
diff --git a/public/locales/fa/bundle.json b/public/locales/fa/bundle.json
index 41151864..07a4dd15 100644
--- a/public/locales/fa/bundle.json
+++ b/public/locales/fa/bundle.json
@@ -1578,6 +1578,16 @@
   "settings.advancedAi.model": "مدل هوش مصنوعی",
   "settings.advancedAi.modelDescription": "مدل هوش مصنوعی برای تولید صحنه، تحلیل و سایر ویژگی‌های مبتنی بر هوش مصنوعی استفاده می‌شود. در دسترس بودن مدل به ارائه دهنده انتخابی شما بستگی دارد.",
   "settings.advancedAi.modelRecommendationsIntro": "نقاط بازرسی محلی محبوب (اسامی را در زمان اجرا تأیید کنید):",
+  "settings.advancedAi.ollamaPull.recommendedFor": "پیشنهادشده برای دستگاه شما:",
+  "settings.advancedAi.ollamaPull.pull": "دریافت مدل",
+  "settings.advancedAi.ollamaPull.pulling": "در حال دانلود مدل…",
+  "settings.advancedAi.ollamaPull.retry": "تلاش دوباره",
+  "settings.advancedAi.ollamaPull.installed": "نصب‌شده",
+  "settings.advancedAi.ollamaPull.useModel": "استفاده از این مدل",
+  "settings.advancedAi.ollamaPull.batteryNote": "برای صرفه‌جویی در باتری یک اندازه کوچک‌تر شد.",
+  "settings.advancedAi.ollamaPull.tier.high-end": "رده‌بالا",
+  "settings.advancedAi.ollamaPull.tier.mid-range": "میان‌رده",
+  "settings.advancedAi.ollamaPull.tier.low-end": "رده‌پایین",
   "settings.advancedAi.ragModeHint": "حالت ترکیبی تطبیق کلمات کلیدی را با درک معنایی ترکیب می‌کند - بهترین برای اکثر داستان‌ها. پس از ویرایش های بزرگ، فهرست را دوباره بسازید.",
   "settings.advancedAi.ragModeHybrid": "ترکیبی (معنی + واژگانی، توصیه شده)",
   "settings.advancedAi.ragModeLabel": "حالت بازیابی متن",
diff --git a/public/locales/fi/bundle.json b/public/locales/fi/bundle.json
index 3f97d1de..d3068e18 100644
--- a/public/locales/fi/bundle.json
+++ b/public/locales/fi/bundle.json
@@ -1578,6 +1578,16 @@
   "settings.advancedAi.model": "AI malli",
   "settings.advancedAi.modelDescription": "Tekoälymalli, jota käytetään kohtausten luomiseen, analysointiin ja muihin tekoälyyn perustuviin ominaisuuksiin. Mallin saatavuus riippuu valitsemastasi toimittajasta.",
   "settings.advancedAi.modelRecommendationsIntro": "Suositut paikalliset tarkistuspisteet (tarkista nimet ajon aikana):",
+  "settings.advancedAi.ollamaPull.recommendedFor": "Suositus laitteellesi:",
+  "settings.advancedAi.ollamaPull.pull": "Lataa malli",
+  "settings.advancedAi.ollamaPull.pulling": "Ladataan mallia…",
+  "settings.advancedAi.ollamaPull.retry": "Yritä uudelleen",
+  "settings.advancedAi.ollamaPull.installed": "Asennettu",
+  "settings.advancedAi.ollamaPull.useModel": "Käytä tätä mallia",
+  "settings.advancedAi.ollamaPull.batteryNote": "Pienennetty kokoa akun säästämiseksi.",
+  "settings.advancedAi.ollamaPull.tier.high-end": "Tehokas",
+  "settings.advancedAi.ollamaPull.tier.mid-range": "Keskitaso",
+  "settings.advancedAi.ollamaPull.tier.low-end": "Kevyt",
   "settings.advancedAi.ragModeHint": "Hybriditila yhdistää avainsanahaun semanttiseen ymmärrykseen – paras useimpiin tarinoihin. Rakenna hakemisto uudelleen suurten muokkausten jälkeen.",
   "settings.advancedAi.ragModeHybrid": "Hybridi (semanttinen + leksikaalinen, suositeltava)",
   "settings.advancedAi.ragModeLabel": "Kontekstin hakutila",
diff --git a/public/locales/fr/bundle.json b/public/locales/fr/bundle.json
index 9dc66e5d..ea968c91 100644
--- a/public/locales/fr/bundle.json
+++ b/public/locales/fr/bundle.json
@@ -1577,6 +1577,16 @@
   "settings.advancedAi.model": "Modèle d'IA",
   "settings.advancedAi.modelDescription": "Le modèle IA utilisé pour la génération de scènes, l'analyse et d'autres fonctionnalités IA. La disponibilité dépend du fournisseur sélectionné.",
   "settings.advancedAi.modelRecommendationsIntro": "Points de contrôle locaux populaires (vérifiez les noms dans votre environnement d'exécution) :",
+  "settings.advancedAi.ollamaPull.recommendedFor": "Recommandé pour votre appareil :",
+  "settings.advancedAi.ollamaPull.pull": "Télécharger le modèle",
+  "settings.advancedAi.ollamaPull.pulling": "Téléchargement du modèle…",
+  "settings.advancedAi.ollamaPull.retry": "Réessayer",
+  "settings.advancedAi.ollamaPull.installed": "Installé",
+  "settings.advancedAi.ollamaPull.useModel": "Utiliser ce modèle",
+  "settings.advancedAi.ollamaPull.batteryNote": "Taille réduite pour économiser la batterie.",
+  "settings.advancedAi.ollamaPull.tier.high-end": "Haut de gamme",
+  "settings.advancedAi.ollamaPull.tier.mid-range": "Milieu de gamme",
+  "settings.advancedAi.ollamaPull.tier.low-end": "Entrée de gamme",
   "settings.advancedAi.ragModeHint": "Le mode hybride associe recherche par mots-clés et compréhension sémantique — idéal pour la plupart des histoires. Reconstruisez l'index après de grandes modifications.",
   "settings.advancedAi.ragModeHybrid": "Hybride (sémantique + lexical, recommandé)",
   "settings.advancedAi.ragModeLabel": "Mode de récupération du contexte",
diff --git a/public/locales/he/bundle.json b/public/locales/he/bundle.json
index 0fdfb67d..9a9911fb 100644
--- a/public/locales/he/bundle.json
+++ b/public/locales/he/bundle.json
@@ -1578,6 +1578,16 @@
   "settings.advancedAi.model": "מודל AI",
   "settings.advancedAi.modelDescription": "מודל ה‑AI המשמש ליצירת סצנות, ניתוח ותכונות מופעלות‑AI נוספות. זמינות המודל תלויה בספק שבחרתם.",
   "settings.advancedAi.modelRecommendationsIntro": "נקודות ביקורת מקומיות פופולריות (אמתו שמות בזמן הריצה שלכם):",
+  "settings.advancedAi.ollamaPull.recommendedFor": "מומלץ למכשיר שלך:",
+  "settings.advancedAi.ollamaPull.pull": "הורדת מודל",
+  "settings.advancedAi.ollamaPull.pulling": "מוריד מודל…",
+  "settings.advancedAi.ollamaPull.retry": "נסה שוב",
+  "settings.advancedAi.ollamaPull.installed": "מותקן",
+  "settings.advancedAi.ollamaPull.useModel": "השתמש במודל הזה",
+  "settings.advancedAi.ollamaPull.batteryNote": "הוקטן בגודל אחד כדי לחסוך בסוללה.",
+  "settings.advancedAi.ollamaPull.tier.high-end": "ביצועים גבוהים",
+  "settings.advancedAi.ollamaPull.tier.mid-range": "בינוני",
+  "settings.advancedAi.ollamaPull.tier.low-end": "בסיסי",
   "settings.advancedAi.ragModeHint": "מצב היברידי משלב התאמת מילות מפתח עם הבנה סמנטית — הטוב ביותר לרוב הסיפורים. בנו מחדש את האינדקס לאחר עריכות גדולות.",
   "settings.advancedAi.ragModeHybrid": "היברידי (סמנטי + לקסיקלי, מומלץ)",
   "settings.advancedAi.ragModeLabel": "מצב אחזור הקשר",
diff --git a/public/locales/hu/bundle.json b/public/locales/hu/bundle.json
index 5466f7a9..b6995b66 100644
--- a/public/locales/hu/bundle.json
+++ b/public/locales/hu/bundle.json
@@ -1578,6 +1578,16 @@
   "settings.advancedAi.model": "AI modell",
   "settings.advancedAi.modelDescription": "A jelenetek generálásához, elemzéséhez és más mesterséges intelligencia által vezérelt funkciókhoz használt AI-modell. A modell elérhetősége a kiválasztott szolgáltatótól függ.",
   "settings.advancedAi.modelRecommendationsIntro": "Népszerű helyi ellenőrzőpontok (ellenőrizze a neveket a futásidőben):",
+  "settings.advancedAi.ollamaPull.recommendedFor": "Ajánlott az eszközödhöz:",
+  "settings.advancedAi.ollamaPull.pull": "Modell letöltése",
+  "settings.advancedAi.ollamaPull.pulling": "Modell letöltése…",
+  "settings.advancedAi.ollamaPull.retry": "Újrapróbálkozás",
+  "settings.advancedAi.ollamaPull.installed": "Telepítve",
+  "settings.advancedAi.ollamaPull.useModel": "Ez a modell használata",
+  "settings.advancedAi.ollamaPull.batteryNote": "Egy mérettel kisebbet választottunk az akkumulátor kímélésére.",
+  "settings.advancedAi.ollamaPull.tier.high-end": "Csúcskategória",
+  "settings.advancedAi.ollamaPull.tier.mid-range": "Középkategória",
+  "settings.advancedAi.ollamaPull.tier.low-end": "Belépőszint",
   "settings.advancedAi.ragModeHint": "A hibrid mód ötvözi a kulcsszóegyezést a szemantikai megértéssel – ez a legjobb a legtöbb történethez. Nagyobb szerkesztések után építse újra az indexet.",
   "settings.advancedAi.ragModeHybrid": "Hibrid (szemantikai + lexikális, ajánlott)",
   "settings.advancedAi.ragModeLabel": "Kontextus visszakeresési mód",
diff --git a/public/locales/is/bundle.json b/public/locales/is/bundle.json
index 0433816c..b9acfdc8 100644
--- a/public/locales/is/bundle.json
+++ b/public/locales/is/bundle.json
@@ -1578,6 +1578,16 @@
   "settings.advancedAi.model": "AI líkan",
   "settings.advancedAi.modelDescription": "Gervigreind líkanið sem er notað fyrir senuframleiðslu, greiningu og aðra gervigreindar eiginleika. Framboð á gerðum fer eftir þjónustuveitunni sem þú hefur valið.",
   "settings.advancedAi.modelRecommendationsIntro": "Vinsælir staðbundnir eftirlitsstöðvar (staðfestu nöfn á keyrslutíma þínum):",
+  "settings.advancedAi.ollamaPull.recommendedFor": "Mælt með fyrir tækið þitt:",
+  "settings.advancedAi.ollamaPull.pull": "Sækja líkan",
+  "settings.advancedAi.ollamaPull.pulling": "Sæki líkan…",
+  "settings.advancedAi.ollamaPull.retry": "Reyna aftur",
+  "settings.advancedAi.ollamaPull.installed": "Uppsett",
+  "settings.advancedAi.ollamaPull.useModel": "Nota þetta líkan",
+  "settings.advancedAi.ollamaPull.batteryNote": "Lækkað um eitt þrep til að spara rafhlöðu.",
+  "settings.advancedAi.ollamaPull.tier.high-end": "Hágæða",
+  "settings.advancedAi.ollamaPull.tier.mid-range": "Millistig",
+  "settings.advancedAi.ollamaPull.tier.low-end": "Grunnstig",
   "settings.advancedAi.ragModeHint": "Hybrid háttur blandar saman leitarorðasamsvörun við merkingarskilning - best fyrir flestar sögur. Endurbyggðu vísitöluna eftir miklar breytingar.",
   "settings.advancedAi.ragModeHybrid": "Blendingur (merkingartækni + orðafræði, mælt með)",
   "settings.advancedAi.ragModeLabel": "Samhengisleitarstilling",
diff --git a/public/locales/it/bundle.json b/public/locales/it/bundle.json
index f306beb2..6efedfff 100644
--- a/public/locales/it/bundle.json
+++ b/public/locales/it/bundle.json
@@ -1577,6 +1577,16 @@
   "settings.advancedAi.model": "Modello IA",
   "settings.advancedAi.modelDescription": "Il modello IA usato per la generazione di scene, analisi e altre funzionalità IA. La disponibilità dipende dal provider selezionato.",
   "settings.advancedAi.modelRecommendationsIntro": "Checkpoint locali popolari (verifica i nomi nel tuo ambiente di esecuzione):",
+  "settings.advancedAi.ollamaPull.recommendedFor": "Consigliato per il tuo dispositivo:",
+  "settings.advancedAi.ollamaPull.pull": "Scarica modello",
+  "settings.advancedAi.ollamaPull.pulling": "Download del modello…",
+  "settings.advancedAi.ollamaPull.retry": "Riprova",
+  "settings.advancedAi.ollamaPull.installed": "Installato",
+  "settings.advancedAi.ollamaPull.useModel": "Usa questo modello",
+  "settings.advancedAi.ollamaPull.batteryNote": "Ridotta la dimensione per risparmiare batteria.",
+  "settings.advancedAi.ollamaPull.tier.high-end": "Fascia alta",
+  "settings.advancedAi.ollamaPull.tier.mid-range": "Fascia media",
+  "settings.advancedAi.ollamaPull.tier.low-end": "Fascia bassa",
   "settings.advancedAi.ragModeHint": "La modalità ibrida combina ricerca per parole chiave e comprensione semantica — ideale per la maggior parte delle storie. Ricostruisci l'indice dopo modifiche importanti.",
   "settings.advancedAi.ragModeHybrid": "Ibrido (semantico + lessicale, consigliato)",
   "settings.advancedAi.ragModeLabel": "Modalità di recupero del contesto",
diff --git a/public/locales/ja/bundle.json b/public/locales/ja/bundle.json
index 434088c9..e38044bc 100644
--- a/public/locales/ja/bundle.json
+++ b/public/locales/ja/bundle.json
@@ -1578,6 +1578,16 @@
   "settings.advancedAi.model": "AIモデル",
   "settings.advancedAi.modelDescription": "シーンの生成、分析、その他の AI を利用した機能に使用される AI モデル。利用可能なモデルは、選択したプロバイダーによって異なります。",
   "settings.advancedAi.modelRecommendationsIntro": "一般的なローカル チェックポイント (ランタイムで名前を確認):",
+  "settings.advancedAi.ollamaPull.recommendedFor": "お使いのデバイスにおすすめ:",
+  "settings.advancedAi.ollamaPull.pull": "モデルを取得",
+  "settings.advancedAi.ollamaPull.pulling": "モデルをダウンロード中…",
+  "settings.advancedAi.ollamaPull.retry": "再試行",
+  "settings.advancedAi.ollamaPull.installed": "インストール済み",
+  "settings.advancedAi.ollamaPull.useModel": "このモデルを使用",
+  "settings.advancedAi.ollamaPull.batteryNote": "バッテリー節約のためサイズを一段下げました。",
+  "settings.advancedAi.ollamaPull.tier.high-end": "ハイエンド",
+  "settings.advancedAi.ollamaPull.tier.mid-range": "ミッドレンジ",
+  "settings.advancedAi.ollamaPull.tier.low-end": "ローエンド",
   "settings.advancedAi.ragModeHint": "ハイブリッド モードでは、キーワードの一致と意味の理解が融合され、ほとんどのストーリーに最適です。大規模な編集後にインデックスを再構築します。",
   "settings.advancedAi.ragModeHybrid": "ハイブリッド (意味論 + 語彙論、推奨)",
   "settings.advancedAi.ragModeLabel": "コンテキスト取得モード",
diff --git a/public/locales/pt/bundle.json b/public/locales/pt/bundle.json
index f2e38ab0..855f9755 100644
--- a/public/locales/pt/bundle.json
+++ b/public/locales/pt/bundle.json
@@ -1578,6 +1578,16 @@
   "settings.advancedAi.model": "IA Model",
   "settings.advancedAi.modelDescription": "The IA model used for scene generation, analysis, and other IA-powered features. Model availability depends on your selected provider.",
   "settings.advancedAi.modelRecommendationsIntro": "Pontos de verificação locais populares (verifique os nomes em seu tempo de execução):",
+  "settings.advancedAi.ollamaPull.recommendedFor": "Recomendado para o seu dispositivo:",
+  "settings.advancedAi.ollamaPull.pull": "Baixar modelo",
+  "settings.advancedAi.ollamaPull.pulling": "Baixando modelo…",
+  "settings.advancedAi.ollamaPull.retry": "Tentar novamente",
+  "settings.advancedAi.ollamaPull.installed": "Instalado",
+  "settings.advancedAi.ollamaPull.useModel": "Usar este modelo",
+  "settings.advancedAi.ollamaPull.batteryNote": "Reduzido um tamanho para poupar bateria.",
+  "settings.advancedAi.ollamaPull.tier.high-end": "Alto desempenho",
+  "settings.advancedAi.ollamaPull.tier.mid-range": "Intermediário",
+  "settings.advancedAi.ollamaPull.tier.low-end": "Básico",
   "settings.advancedAi.ragModeHint": "O modo híbrido combina correspondência de palavras-chave com compreensão semântica – melhor para a maioria das histórias. Reconstrua o índice após grandes edições.",
   "settings.advancedAi.ragModeHybrid": "Híbrido (semântico + lexical, recomendado)",
   "settings.advancedAi.ragModeLabel": "Modo de recuperação de contexto",
diff --git a/public/locales/sv/bundle.json b/public/locales/sv/bundle.json
index 3b0d3d20..56f3391e 100644
--- a/public/locales/sv/bundle.json
+++ b/public/locales/sv/bundle.json
@@ -1578,6 +1578,16 @@
   "settings.advancedAi.model": "AI-modell",
   "settings.advancedAi.modelDescription": "AI-modellen som används för scengenerering, analys och andra AI-drivna funktioner. Modellens tillgänglighet beror på din valda leverantör.",
   "settings.advancedAi.modelRecommendationsIntro": "Populära lokala kontrollpunkter (verifiera namn under din körning):",
+  "settings.advancedAi.ollamaPull.recommendedFor": "Rekommenderas för din enhet:",
+  "settings.advancedAi.ollamaPull.pull": "Hämta modell",
+  "settings.advancedAi.ollamaPull.pulling": "Hämtar modell…",
+  "settings.advancedAi.ollamaPull.retry": "Försök igen",
+  "settings.advancedAi.ollamaPull.installed": "Installerad",
+  "settings.advancedAi.ollamaPull.useModel": "Använd den här modellen",
+  "settings.advancedAi.ollamaPull.batteryNote": "Sänkte en storlek för att spara batteri.",
+  "settings.advancedAi.ollamaPull.tier.high-end": "Högpresterande",
+  "settings.advancedAi.ollamaPull.tier.mid-range": "Mellanklass",
+  "settings.advancedAi.ollamaPull.tier.low-end": "Instegsnivå",
   "settings.advancedAi.ragModeHint": "Hybridläge blandar sökordsmatchning med semantisk förståelse – bäst för de flesta artiklar. Bygg om indexet efter stora redigeringar.",
   "settings.advancedAi.ragModeHybrid": "Hybrid (semantisk + lexikal, rekommenderas)",
   "settings.advancedAi.ragModeLabel": "Läge för kontexthämtning",
diff --git a/public/locales/zh/bundle.json b/public/locales/zh/bundle.json
index 1b27e317..9406abf6 100644
--- a/public/locales/zh/bundle.json
+++ b/public/locales/zh/bundle.json
@@ -1578,6 +1578,16 @@
   "settings.advancedAi.model": "人工智能模型",
   "settings.advancedAi.modelDescription": "用于场景生成、分析和其他人工智能功能的人工智能模型。型号可用性取决于您选择的提供商。",
   "settings.advancedAi.modelRecommendationsIntro": "流行的本地检查点（在运行时验证名称）：",
+  "settings.advancedAi.ollamaPull.recommendedFor": "为您的设备推荐：",
+  "settings.advancedAi.ollamaPull.pull": "拉取模型",
+  "settings.advancedAi.ollamaPull.pulling": "正在下载模型…",
+  "settings.advancedAi.ollamaPull.retry": "重试",
+  "settings.advancedAi.ollamaPull.installed": "已安装",
+  "settings.advancedAi.ollamaPull.useModel": "使用此模型",
+  "settings.advancedAi.ollamaPull.batteryNote": "已降低一档以节省电量。",
+  "settings.advancedAi.ollamaPull.tier.high-end": "高端",
+  "settings.advancedAi.ollamaPull.tier.mid-range": "中端",
+  "settings.advancedAi.ollamaPull.tier.low-end": "低端",
   "settings.advancedAi.ragModeHint": "混合模式将关键词匹配与语义理解相结合——最适合大多数故事。大量编辑后重建索引。",
   "settings.advancedAi.ragModeHybrid": "混合（语义+词汇，推荐）",
   "settings.advancedAi.ragModeLabel": "上下文检索模式",
diff --git a/services/ai/modelRecommendations.ts b/services/ai/modelRecommendations.ts
index bf4aa8ea..c9015ad0 100644
--- a/services/ai/modelRecommendations.ts
+++ b/services/ai/modelRecommendations.ts
@@ -13,6 +13,54 @@ export const RECOMMENDED_OLLAMA_MODEL_IDS = [
 export const RECOMMENDED_OPENAI_COMPAT_CLOUD_HINT =
   'OpenRouter/Groq/Azure OpenAI: Basis-URL + Modell-ID wie beim Anbieter dokumentiert (Stand 2026 — Preise/SLAs auf der Anbieterseite prüfen).';
 
+export type OllamaDeviceTier = 'high-end' | 'mid-range' | 'low-end';
+
+export interface OllamaDeviceRecommendation {
+  /** Ollama model tag to `ollama pull` (and the one-click pull button installs). */
+  modelId: string;
+  /** The device tier this recommendation targets (after any battery downgrade). */
+  tier: OllamaDeviceTier;
+  /** Human-readable approximate download size for the UI. */
+  sizeHint: string;
+  /** True when the tier was stepped down one level because the device is on low battery. */
+  downgradedForBattery: boolean;
+}
+
+// QNBS-v3: curated device→Ollama catalog (2025/2026 tags). Bigger models need more RAM/VRAM, so the
+// recommendation tracks the profiler's deviceClass; on low battery we step down one tier to cut the
+// sustained compute/thermal load of a one-click local run.
+const OLLAMA_DEVICE_CATALOG: Record<OllamaDeviceTier, { modelId: string; sizeHint: string }> = {
+  'high-end': { modelId: 'qwen2.5:7b', sizeHint: '~4.7 GB' },
+  'mid-range': { modelId: 'llama3.2:3b', sizeHint: '~2.0 GB' },
+  'low-end': { modelId: 'llama3.2:1b', sizeHint: '~1.3 GB' },
+};
+
+const LOW_BATTERY_THRESHOLD = 0.2;
+
+/**
+ * Pick the recommended Ollama model for the current device from a curated, tiered catalog.
+ * QNBS-v3: device-aware sibling of getModelRecommendationForTask, but for the user-run Ollama server
+ * (so it returns concrete `ollama pull` tags, not WebLLM/ONNX ids). Unknown class → conservative low-end.
+ */
+export function getOllamaModelForDevice(report: DeviceHealthReport): OllamaDeviceRecommendation {
+  let tier: OllamaDeviceTier =
+    report.deviceClass === 'high-end'
+      ? 'high-end'
+      : report.deviceClass === 'mid-range'
+        ? 'mid-range'
+        : 'low-end'; // 'low-end' + 'unknown' → conservative default
+
+  const lowBattery = report.batteryLevel !== null && report.batteryLevel < LOW_BATTERY_THRESHOLD;
+  let downgradedForBattery = false;
+  if (lowBattery && tier !== 'low-end') {
+    tier = tier === 'high-end' ? 'mid-range' : 'low-end';
+    downgradedForBattery = true;
+  }
+
+  const { modelId, sizeHint } = OLLAMA_DEVICE_CATALOG[tier];
+  return { modelId, tier, sizeHint, downgradedForBattery };
+}
+
 // QNBS-v3: Eco-mode model: smallest viable ONNX model for battery-constrained devices (2025 update).
 export const ECO_TEXT_GEN_MODEL = 'HuggingFaceTB/SmolLM2-135M-Instruct';
 
diff --git a/services/ollamaService.ts b/services/ollamaService.ts
index 226287b7..78a1223b 100644
--- a/services/ollamaService.ts
+++ b/services/ollamaService.ts
@@ -35,6 +35,107 @@ export interface OllamaStreamCallbacks {
   onError?: (error: Error) => void;
 }
 
+// QNBS-v3: shape-based AbortError check — DOMException is not an instanceof Error in some runtimes.
+function isAbortError(err: unknown): boolean {
+  return (
+    typeof err === 'object' && err !== null && (err as { name?: string }).name === 'AbortError'
+  );
+}
+
+export interface OllamaPullProgress {
+  /** Ollama status line, e.g. "pulling manifest", "downloading …", "verifying", "success". */
+  status: string;
+  /** 0–1 download fraction for the active layer, when total/completed bytes are present. */
+  progress?: number;
+  completedBytes?: number;
+  totalBytes?: number;
+}
+
+export interface OllamaPullOptions {
+  baseUrl?: string;
+  signal?: AbortSignal;
+  onProgress?: (p: OllamaPullProgress) => void;
+}
+
+/**
+ * One-click model install: POST {baseUrl}/api/pull and stream the NDJSON progress until the model is
+ * ready. QNBS-v3: mirrors streamOllama's reader/abort handling. Ollama reports failures inline as an
+ * `{error}` line (not via HTTP status), so we surface those as a thrown Error; an aborted signal
+ * propagates as the original AbortError so callers can distinguish cancel from failure.
+ */
+export async function pullOllamaModel(name: string, opts: OllamaPullOptions = {}): Promise<void> {
+  const baseUrl = normalizeBaseUrl(opts.baseUrl);
+  const model = name.trim().replace(/^ollama\//, '');
+  if (!model) throw new Error('Ollama pull: model name is required');
+
+  let response: Response;
+  try {
+    response = await fetch(`${baseUrl}/api/pull`, {
+      method: 'POST',
+      headers: { 'Content-Type': 'application/json' },
+      body: JSON.stringify({ name: model, stream: true }),
+      signal: opts.signal ?? null,
+    });
+  } catch (err) {
+    if (isAbortError(err)) throw err;
+    throw new Error(
+      `Ollama not reachable (${baseUrl}). Make sure Ollama is running: ollama serve`,
+      { cause: err as Error },
+    );
+  }
+
+  if (!response.ok) {
+    throw new Error(`Ollama pull failed ${response.status}: ${response.statusText}`);
+  }
+
+  const reader = response.body?.getReader();
+  if (!reader) throw new Error('Ollama pull: no response body');
+
+  const decoder = new TextDecoder();
+  let buffer = '';
+
+  const handleLine = (line: string): void => {
+    const trimmed = line.trim();
+    if (!trimmed) return;
+    let payload: { status?: string; total?: number; completed?: number; error?: string };
+    try {
+      payload = JSON.parse(trimmed);
+    } catch {
+      return; // ignore non-JSON keepalive/noise lines
+    }
+    if (payload.error) throw new Error(`Ollama pull error: ${payload.error}`);
+    const hasBytes =
+      typeof payload.total === 'number' &&
+      payload.total > 0 &&
+      typeof payload.completed === 'number';
+    opts.onProgress?.({
+      status: payload.status ?? 'pulling',
+      ...(hasBytes
+        ? { progress: Math.min(1, (payload.completed as number) / (payload.total as number)) }
+        : {}),
+      ...(typeof payload.completed === 'number' ? { completedBytes: payload.completed } : {}),
+      ...(typeof payload.total === 'number' ? { totalBytes: payload.total } : {}),
+    });
+  };
+
+  // QNBS-v3: cancel the reader in finally so an inline-{error} throw (or any parse error) does not
+  // leave the HTTP body streaming in the background. cancel() is a no-op once the stream is drained.
+  try {
+    while (true) {
+      const { done, value } = await reader.read();
+      if (done) break;
+      buffer += decoder.decode(value, { stream: true });
+      const lines = buffer.split(/\r?\n/);
+      buffer = lines.pop() ?? '';
+      for (const line of lines) handleLine(line);
+    }
+    // Flush any trailing buffered line (final "success" status without a newline).
+    if (buffer.trim()) handleLine(buffer);
+  } finally {
+    await reader.cancel().catch(() => {});
+  }
+}
+
 export async function listOllamaModels(baseUrl?: string): Promise<string[]> {
   const url = `${normalizeBaseUrl(baseUrl)}/api/tags`;
   try {
diff --git a/tests/unit/ai/modelRecommendations.test.ts b/tests/unit/ai/modelRecommendations.test.ts
index 4faba76d..e8b09261 100644
--- a/tests/unit/ai/modelRecommendations.test.ts
+++ b/tests/unit/ai/modelRecommendations.test.ts
@@ -8,6 +8,7 @@ import type { DeviceHealthReport } from '../../../services/ai/deviceHealthServic
 import {
   ECO_TEXT_GEN_MODEL,
   getModelRecommendationForTask,
+  getOllamaModelForDevice,
   getProviderSpeedEstimate,
   RECOMMENDED_OLLAMA_MODEL_IDS,
 } from '../../../services/ai/modelRecommendations';
@@ -110,3 +111,64 @@ describe('getProviderSpeedEstimate', () => {
     expect(latency).toBe(Number.POSITIVE_INFINITY);
   });
 });
+
+describe('getOllamaModelForDevice', () => {
+  it('recommends the 7B model for a high-end device', () => {
+    const rec = getOllamaModelForDevice({ ...baseReport, deviceClass: 'high-end' });
+    expect(rec).toMatchObject({
+      modelId: 'qwen2.5:7b',
+      tier: 'high-end',
+      downgradedForBattery: false,
+    });
+  });
+
+  it('recommends the 3B model for a mid-range device', () => {
+    const rec = getOllamaModelForDevice({ ...baseReport, deviceClass: 'mid-range' });
+    expect(rec.modelId).toBe('llama3.2:3b');
+    expect(rec.tier).toBe('mid-range');
+  });
+
+  it('recommends the smallest model for low-end and unknown devices', () => {
+    expect(getOllamaModelForDevice({ ...baseReport, deviceClass: 'low-end' }).modelId).toBe(
+      'llama3.2:1b',
+    );
+    expect(getOllamaModelForDevice({ ...baseReport, deviceClass: 'unknown' }).modelId).toBe(
+      'llama3.2:1b',
+    );
+  });
+
+  it('exposes a human-readable size hint', () => {
+    expect(getOllamaModelForDevice({ ...baseReport, deviceClass: 'high-end' }).sizeHint).toMatch(
+      /GB/,
+    );
+  });
+
+  it('steps down one tier on low battery and flags the downgrade', () => {
+    const rec = getOllamaModelForDevice({
+      ...baseReport,
+      deviceClass: 'high-end',
+      batteryLevel: 0.1,
+    });
+    expect(rec.tier).toBe('mid-range');
+    expect(rec.modelId).toBe('llama3.2:3b');
+    expect(rec.downgradedForBattery).toBe(true);
+  });
+
+  it('does not downgrade below low-end and does not flag when battery is unknown', () => {
+    const lowEndLowBattery = getOllamaModelForDevice({
+      ...baseReport,
+      deviceClass: 'low-end',
+      batteryLevel: 0.05,
+    });
+    expect(lowEndLowBattery.tier).toBe('low-end');
+    expect(lowEndLowBattery.downgradedForBattery).toBe(false);
+
+    const noBattery = getOllamaModelForDevice({
+      ...baseReport,
+      deviceClass: 'high-end',
+      batteryLevel: null,
+    });
+    expect(noBattery.tier).toBe('high-end');
+    expect(noBattery.downgradedForBattery).toBe(false);
+  });
+});
diff --git a/tests/unit/ollamaService.test.ts b/tests/unit/ollamaService.test.ts
index 33f1b6db..5339747e 100644
--- a/tests/unit/ollamaService.test.ts
+++ b/tests/unit/ollamaService.test.ts
@@ -1,5 +1,23 @@
 import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest';
-import { listOllamaModels, streamOllama, testOllamaConnection } from '../../services/ollamaService';
+import {
+  listOllamaModels,
+  type OllamaPullProgress,
+  pullOllamaModel,
+  streamOllama,
+  testOllamaConnection,
+} from '../../services/ollamaService';
+
+/** Build a streaming NDJSON Response (one JSON object per line) for /api/pull mocks. */
+function ndjsonResponse(objects: object[], status = 200): Response {
+  const enc = new TextEncoder();
+  const stream = new ReadableStream<Uint8Array>({
+    start(controller) {
+      for (const o of objects) controller.enqueue(enc.encode(`${JSON.stringify(o)}\n`));
+      controller.close();
+    },
+  });
+  return new Response(stream, { status });
+}
 
 beforeEach(() => {
   vi.stubGlobal('fetch', vi.fn());
@@ -149,3 +167,70 @@ describe('streamOllama', () => {
     expect(chunks).toEqual(['ok']);
   });
 });
+
+// ─── pullOllamaModel ──────────────────────────────────────────────────────────
+
+describe('pullOllamaModel', () => {
+  it('streams progress and resolves on success', async () => {
+    vi.mocked(fetch).mockResolvedValueOnce(
+      ndjsonResponse([
+        { status: 'pulling manifest' },
+        { status: 'downloading', total: 1000, completed: 250 },
+        { status: 'downloading', total: 1000, completed: 1000 },
+        { status: 'success' },
+      ]),
+    );
+    const progress: OllamaPullProgress[] = [];
+    await pullOllamaModel('llama3.2:1b', { onProgress: (p) => progress.push(p) });
+
+    expect(progress.map((p) => p.status)).toEqual([
+      'pulling manifest',
+      'downloading',
+      'downloading',
+      'success',
+    ]);
+    // mid-download fraction surfaced
+    expect(progress[1]).toMatchObject({ progress: 0.25, completedBytes: 250, totalBytes: 1000 });
+    expect(progress[2]?.progress).toBe(1);
+    // manifest/success lines (no byte counts) omit the progress fraction
+    expect(progress[0]?.progress).toBeUndefined();
+  });
+
+  it('POSTs to /api/pull with the stripped model name and stream:true', async () => {
+    vi.mocked(fetch).mockResolvedValueOnce(ndjsonResponse([{ status: 'success' }]));
+    await pullOllamaModel('ollama/llama3.2:3b', { baseUrl: 'http://host:11434/' });
+    const [url, init] = vi.mocked(fetch).mock.calls[0] as [string, RequestInit];
+    expect(url).toBe('http://host:11434/api/pull');
+    expect(init.method).toBe('POST');
+    expect(JSON.parse(init.body as string)).toEqual({ name: 'llama3.2:3b', stream: true });
+  });
+
+  it('throws on an inline {error} line (Ollama reports failures in-band)', async () => {
+    vi.mocked(fetch).mockResolvedValueOnce(
+      ndjsonResponse([{ status: 'pulling manifest' }, { error: 'manifest unknown' }]),
+    );
+    await expect(pullOllamaModel('bogus:latest')).rejects.toThrow(/manifest unknown/);
+  });
+
+  it('throws a reachability error on network failure', async () => {
+    vi.mocked(fetch).mockRejectedValueOnce(new TypeError('Network error'));
+    await expect(pullOllamaModel('llama3.2:1b')).rejects.toThrow(/not reachable/);
+  });
+
+  it('throws on a non-200 response', async () => {
+    vi.mocked(fetch).mockResolvedValueOnce(
+      new Response('', { status: 500, statusText: 'Server Error' }),
+    );
+    await expect(pullOllamaModel('llama3.2:1b')).rejects.toThrow(/pull failed 500/);
+  });
+
+  it('propagates an AbortError unchanged (cancel, not failure)', async () => {
+    const abortErr = Object.assign(new Error('aborted'), { name: 'AbortError' });
+    vi.mocked(fetch).mockRejectedValueOnce(abortErr);
+    await expect(pullOllamaModel('llama3.2:1b')).rejects.toMatchObject({ name: 'AbortError' });
+  });
+
+  it('rejects an empty model name', async () => {
+    await expect(pullOllamaModel('   ')).rejects.toThrow(/model name is required/);
+  });
+});
diff --git a/tests/unit/settings/OllamaDevicePull.test.tsx b/tests/unit/settings/OllamaDevicePull.test.tsx
new file mode 100644
index 00000000..14344b52
--- /dev/null
+++ b/tests/unit/settings/OllamaDevicePull.test.tsx
@@ -0,0 +1,119 @@
+/**
+ * Tests for components/settings/OllamaDevicePull.tsx — device-aware Ollama recommendation + one-click
+ * pull. Mocks the device profiler, the recommendation mapper, and pullOllamaModel.
+ */
+
+import { render, screen } from '@testing-library/react';
+import userEvent from '@testing-library/user-event';
+import { beforeEach, describe, expect, it, vi } from 'vitest';
+
+const { mockGetHealthReport, mockGetOllamaModelForDevice, mockPull } = vi.hoisted(() => ({
+  mockGetHealthReport: vi.fn(),
+  mockGetOllamaModelForDevice: vi.fn(),
+  mockPull: vi.fn(),
+}));
+
+vi.mock('../../../services/ai/deviceHealthService', () => ({
+  getHealthReport: mockGetHealthReport,
+}));
+vi.mock('../../../services/ai/modelRecommendations', () => ({
+  getOllamaModelForDevice: mockGetOllamaModelForDevice,
+}));
+vi.mock('../../../services/ollamaService', () => ({ pullOllamaModel: mockPull }));
+vi.mock('../../../services/logger', () => ({ logger: { warn: vi.fn(), error: vi.fn() } }));
+
+import { OllamaDevicePull } from '../../../components/settings/OllamaDevicePull';
+
+const t = ((k: string) => k) as never;
+
+function setRec(over: Record<string, unknown> = {}) {
+  mockGetOllamaModelForDevice.mockReturnValue({
+    modelId: 'llama3.2:3b',
+    tier: 'mid-range',
+    sizeHint: '~2.0 GB',
+    downgradedForBattery: false,
+    ...over,
+  });
+}
+
+beforeEach(() => {
+  vi.clearAllMocks();
+  mockGetHealthReport.mockResolvedValue({ deviceClass: 'mid-range' });
+  setRec();
+  mockPull.mockResolvedValue(undefined);
+});
+
+describe('OllamaDevicePull', () => {
+  it('renders the device recommendation after profiling', async () => {
+    render(<OllamaDevicePull baseUrl="http://localhost:11434" onUseModel={vi.fn()} t={t} />);
+    expect(await screen.findByText('llama3.2:3b')).toBeInTheDocument();
+    expect(screen.getByText('~2.0 GB')).toBeInTheDocument();
+    expect(screen.getByText('settings.advancedAi.ollamaPull.tier.mid-range')).toBeInTheDocument();
+  });
+
+  it('pulls the model and, on success, applies it via onUseModel', async () => {
+    const onUseModel = vi.fn();
+    const user = userEvent.setup();
+    render(<OllamaDevicePull baseUrl="http://localhost:11434" onUseModel={onUseModel} t={t} />);
+    await user.click(
+      await screen.findByRole('button', { name: 'settings.advancedAi.ollamaPull.pull' }),
+    );
+    expect(mockPull).toHaveBeenCalledWith(
+      'llama3.2:3b',
+      expect.objectContaining({ baseUrl: 'http://localhost:11434' }),
+    );
+    const useBtn = await screen.findByRole('button', {
+      name: 'settings.advancedAi.ollamaPull.useModel',
+    });
+    await user.click(useBtn);
+    expect(onUseModel).toHaveBeenCalledWith('ollama/llama3.2:3b');
+  });
+
+  it('surfaces an error and offers retry when the pull fails', async () => {
+    mockPull.mockRejectedValueOnce(new Error('Ollama not reachable (…)'));
+    const user = userEvent.setup();
+    render(<OllamaDevicePull baseUrl="http://localhost:11434" onUseModel={vi.fn()} t={t} />);
+    await user.click(
+      await screen.findByRole('button', { name: 'settings.advancedAi.ollamaPull.pull' }),
+    );
+    expect(await screen.findByRole('alert')).toHaveTextContent('not reachable');
+    expect(
+      screen.getByRole('button', { name: 'settings.advancedAi.ollamaPull.retry' }),
+    ).toBeInTheDocument();
+  });
+
+  it('cancels via the Cancel button (aborts the controller) and returns to idle', async () => {
+    // The mocked pull observes the AbortSignal and rejects with AbortError when it fires — so this
+    // exercises the REAL Cancel wiring (onClick → abortRef.abort() → signal). Broken wiring (onClick
+    // not aborting) would leave the pull pending and the idle Pull button would never return.
+    mockPull.mockImplementationOnce(
+      (_name: string, opts: { signal?: AbortSignal }) =>
+        new Promise((_resolve, reject) => {
+          opts.signal?.addEventListener('abort', () =>
+            reject(Object.assign(new Error('aborted'), { name: 'AbortError' })),
+          );
+        }),
+    );
+    const user = userEvent.setup();
+    render(<OllamaDevicePull baseUrl="http://localhost:11434" onUseModel={vi.fn()} t={t} />);
+    await user.click(
+      await screen.findByRole('button', { name: 'settings.advancedAi.ollamaPull.pull' }),
+    );
+    // pull is in flight → click the real Cancel button
+    await user.click(await screen.findByRole('button', { name: 'common.cancel' }));
+
+    // definitive transition: the idle Pull button returns (findBy waits) and no error alert appears
+    expect(
+      await screen.findByRole('button', { name: 'settings.advancedAi.ollamaPull.pull' }),
+    ).toBeInTheDocument();
+    expect(screen.queryByRole('alert')).toBeNull();
+  });
+
+  it('shows the battery-downgrade note when the tier was stepped down', async () => {
+    setRec({ downgradedForBattery: true });
+    render(<OllamaDevicePull baseUrl="http://localhost:11434" onUseModel={vi.fn()} t={t} />);
+    expect(
+      await screen.findByText('settings.advancedAi.ollamaPull.batteryNote'),
+    ).toBeInTheDocument();
+  });
+});