fix(suggestion): paste after the shortcut keys are released and keep typing detection stable

Ctrl+Alt+Enter pasted while Ctrl+Alt were still down, so the target app got
Ctrl+Alt+V; accepting now closes the panel and waits for the modifiers to be
released. Candidates are accepted on pointer press because the list is
redrawn as new candidates stream in, which swallowed clicks.

The typing gate identified the focused field by its bounds, so chat boxes
that grow while typing looked like a new field on every keystroke and were
reported as "not typing". Fields are now keyed by window, control type and
name, and a mouse click re-baselines the text instead. The decision log
includes both gate values.

The live-caption model selector moves to the caption section of the General
tab, next to the other caption settings.

apps/desktop/src/renderer/components/STTTab.tsx

@@ -51,9 +51,6 @@ import type {
 } from '@d3ro/core/types'
 import { CodexOAuthGuideModal } from './CodexOAuthGuideModal'
 
-/** 자막 모델 선택지 "받아쓰기와 같게" — 설정에는 null 로 저장한다 */
-const SAME_AS_DICTATION = '__same__'
-
 interface STTTabProps {
   config: Partial<AppConfig>
   updateConfig: (key: keyof AppConfig, value: AppConfig[keyof AppConfig]) => void
@@ -365,30 +362,6 @@ export function STTTab({ config, updateConfig }: STTTabProps): React.ReactElemen
               </Select>
             </FormControl>
 
-            {/* 실시간 자막 전용 모델 — 1초마다 다시 인식하므로 받아쓰기와 다른 모델이 나을 수 있다 */}
-            <FormControl size="small" fullWidth>
-              <InputLabel>{t('settings.captionModel')}</InputLabel>
-              <Select
-                label={t('settings.captionModel')}
-                value={config.captionSttModelId ?? SAME_AS_DICTATION}
-                onChange={(e) =>
-                  updateConfig('captionSttModelId', e.target.value === SAME_AS_DICTATION ? null : e.target.value)
-                }
-              >
-                <MenuItem value={SAME_AS_DICTATION}>{t('settings.captionModel.same')}</MenuItem>
-                {localModels
-                  .filter((m) => m.downloaded)
-                  .map((m) => (
-                    <MenuItem key={m.id} value={m.id}>
-                      {m.name}
-                    </MenuItem>
-                  ))}
-              </Select>
-              <Typography sx={{ mt: 0.5, fontSize: d3roTypo.meta.size, color: d3roPalette.text.secondary }}>
-                {t('settings.captionModel.desc')}
-              </Typography>
-            </FormControl>
-
             {/* 모델 다운로드 진행 바 또는 다운로드 버튼 */}
             {(() => {
               const selected = localModels.find((m) => m.id === (config.sttModelId ?? 'large-v3-turbo'))

apps/desktop/src/renderer/components/SettingsModal.tsx

@@ -35,7 +35,7 @@ import { useI18n, LOCALE_META } from '@d3ro/i18n'
 import type { Locale } from '@d3ro/i18n'
 import type { KeyBindingActionGroup } from '@d3ro/core/keybinding'
 import { auditKeyBindingMap, KEYBINDING_ACTIONS } from '@d3ro/core/keybinding'
-import type { ThemeMode, AppConfig, AudioDevice, LLMModel, LLMStatus } from '@d3ro/core/types'
+import type { ThemeMode, AppConfig, AudioDevice, LLMModel, LLMStatus, STTModel } from '@d3ro/core/types'
 import { useKeyBindingMap } from '../hooks/useKeyBindingMap'
 
 interface SettingsModalProps {
@@ -64,6 +64,9 @@ const ACTION_GROUP_LABEL_KEYS: Readonly<Record<KeyBindingActionGroup, string>> =
 
 const ACTION_GROUP_ORDER: readonly KeyBindingActionGroup[] = ['voice', 'window', 'input']
 
+/** 자막 모델 선택지 "받아쓰기와 같게" — 설정에는 null 로 저장한다 */
+const SAME_AS_DICTATION = '__same__'
+
 export function SettingsModal({ open, initialTab = 0, onClose }: SettingsModalProps): React.ReactElement {
   const { t, locale, setLocale } = useI18n()
   const [activeTab, setActiveTab] = useState(initialTab)
@@ -108,6 +111,15 @@ export function SettingsModal({ open, initialTab = 0, onClose }: SettingsModalPr
   const [micTesting, setMicTesting] = useState(false)
   const [micLevel, setMicLevel] = useState(0)
 
+  // 실시간 자막 모델 선택지 — 받아 둔 로컬 Whisper 모델만
+  const [downloadedSttModels, setDownloadedSttModels] = useState<STTModel[]>([])
+  useEffect(() => {
+    if (!open) return
+    window.electronAPI.stt.getModels().then((res) => {
+      if (res.success && res.data) setDownloadedSttModels(res.data.filter((m) => m.downloaded))
+    })
+  }, [open])
+
   // Ollama 상태
   const [ollamaStatus, setOllamaStatus] = useState<LLMStatus | null>(null)
   const [ollamaModels, setOllamaModels] = useState<LLMModel[]>([])
@@ -481,6 +493,27 @@ export function SettingsModal({ open, initialTab = 0, onClose }: SettingsModalPr
                 </Select>
               </FormControl>
 
+              <FormControl size="small">
+                <InputLabel>{t('settings.captionModel')}</InputLabel>
+                <Select
+                  label={t('settings.captionModel')}
+                  value={config.captionSttModelId ?? SAME_AS_DICTATION}
+                  onChange={(e) =>
+                    updateConfig('captionSttModelId', e.target.value === SAME_AS_DICTATION ? null : e.target.value)
+                  }
+                >
+                  <MenuItem value={SAME_AS_DICTATION}>{t('settings.captionModel.same')}</MenuItem>
+                  {downloadedSttModels.map((m) => (
+                    <MenuItem key={m.id} value={m.id}>
+                      {m.name}
+                    </MenuItem>
+                  ))}
+                </Select>
+                <Typography sx={{ mt: 0.5, fontSize: d3roTypo.meta.size, color: d3roPalette.text.secondary }}>
+                  {t('settings.captionModel.desc')}
+                </Typography>
+              </FormControl>
+
               <FormControlLabel
                 sx={{ mr: 0, alignItems: 'flex-start' }}
                 control={