release: v1.3.2 - Linux CUDA and paste fixes

release: v1.3.2 - Linux CUDA and paste fixes

Author: infiniV

installer/voiceflow.iss

@@ -2,7 +2,7 @@
 ; Creates a Windows installer from the PyInstaller --onedir output
 
 #define MyAppName "VoiceFlow"
-#define MyAppVersion "1.3.1"
+#define MyAppVersion "1.3.2"
 #define MyAppPublisher "VoiceFlow"
 #define MyAppURL "https://github.com/your-repo/voiceflow"
 #define MyAppExeName "VoiceFlow.exe"

package.json

@@ -1,7 +1,7 @@
 {
   "name": "voiceflow",
   "private": true,
-  "version": "1.3.1",
+  "version": "1.3.2",
   "type": "module",
   "scripts": {
     "dev": "npm-run-all --parallel vite pyloid",

src-pyloid/app_controller.py

@@ -174,6 +174,10 @@ def transcribe():
                 info(f"Transcription result: '{text}'")
 
                 if text:
+                    # Prepend space if enabled (useful for continuous dictation)
+                    if settings.prepend_space:
+                        text = " " + text
+
                     # Paste at cursor
                     info("Pasting text at cursor...")
                     self.clipboard_service.paste_at_cursor(text)
@@ -235,6 +239,7 @@ def get_settings(self) -> dict:
             "holdHotkeyEnabled": settings.hold_hotkey_enabled,
             "toggleHotkey": settings.toggle_hotkey,
             "toggleHotkeyEnabled": settings.toggle_hotkey_enabled,
+            "prependSpace": settings.prepend_space,
         }
 
     def update_settings(self, **kwargs) -> dict:
@@ -249,6 +254,8 @@ def update_settings(self, **kwargs) -> dict:
             mapped["save_audio_to_history"] = kwargs["saveAudioToHistory"]
         if "showPopup" in kwargs:
             mapped["show_popup"] = kwargs["showPopup"]
+        if "prependSpace" in kwargs:
+            mapped["prepend_space"] = kwargs["prependSpace"]
         # Hotkey settings (camelCase to snake_case)
         if "holdHotkey" in kwargs:
             mapped["hold_hotkey"] = kwargs["holdHotkey"]

src-pyloid/services/clipboard.py

@@ -67,12 +67,32 @@ def _wl_copy(self, text: str):
             import pyperclip
             pyperclip.copy(text)
 
+    def _type_text_directly(self, text: str) -> bool:
+        """Type text directly using Wayland input tool. Returns True on success."""
+        try:
+            if self._paste_tool == 'wtype':
+                subprocess.run(['wtype', '--', text], check=True, timeout=10)
+            elif self._paste_tool == 'dotool':
+                # dotool type command types text directly
+                subprocess.run(['dotool'], input=f'type {text}\n'.encode(),
+                               check=True, timeout=10)
+            elif self._paste_tool == 'ydotool':
+                subprocess.run(['ydotool', 'type', '--', text],
+                               check=True, timeout=10)
+            else:
+                return False
+            log.debug("Text typed directly via", tool=self._paste_tool)
+            return True
+        except (subprocess.SubprocessError, OSError) as e:
+            log.warning("Direct type failed", tool=self._paste_tool, error=str(e))
+            return False
+
     def _simulate_paste_keystroke(self):
         """Send Ctrl+V using the best available tool."""
         if IS_WAYLAND and self._paste_tool:
             try:
                 if self._paste_tool == 'wtype':
-                    subprocess.run(['wtype', '-M', 'ctrl', '-P', 'v', '-m', 'ctrl'],
+                    subprocess.run(['wtype', '-M', 'ctrl', '-k', 'v', '-m', 'ctrl'],
                                    check=True, timeout=5)
                 elif self._paste_tool == 'dotool':
                     subprocess.run(['dotool'], input=b'key ctrl+v\n',
@@ -93,19 +113,26 @@ def paste_at_cursor(self, text: str):
         """Copy text to clipboard and paste at current cursor position."""
         log.debug("Paste at cursor called", text_length=len(text))
 
-        # Copy our text to clipboard
+        # Always copy to clipboard so user can re-paste manually if needed
         self.copy_to_clipboard(text)
         log.debug("Text copied to clipboard")
 
-        # Small delay to ensure clipboard is updated
+        # On Wayland, type text directly — works in terminals (Ctrl+V doesn't)
+        # and all other apps, regardless of their paste shortcut.
+        if IS_WAYLAND and self._paste_tool:
+            if self._type_text_directly(text):
+                log.debug("Paste complete via direct type")
+                return
+            # Direct type failed — fall through to Ctrl+V
+            log.warning("Direct type failed, falling back to Ctrl+V")
+
+        # Small delay to ensure clipboard is ready
         time.sleep(0.1)
 
-        # Simulate Ctrl+V
         log.debug("Simulating Ctrl+V")
         self._simulate_paste_keystroke()
         log.debug("Paste command sent")
 
-        # Small delay after paste
         time.sleep(0.1)
 
     def get_clipboard(self) -> str:

src-pyloid/services/gpu.py

@@ -76,9 +76,18 @@ def _check_cudnn_available() -> tuple[bool, Optional[str]]:
         Tuple of (is_available, error_message)
     """
     if sys.platform != "win32":
-        # On Linux, cuDNN is either bundled into ctranslate2 or not required.
-        # The nvidia cublas pip packages (preloaded at startup) provide what's needed.
-        return True, None
+        # On Linux, verify that libcublas is actually loadable at runtime.
+        # ctranslate2 can detect CUDA device presence without libcublas, but
+        # inference will fail if the library isn't available.
+        import ctypes
+        for lib_name in ("libcublas.so.12", "libcublas.so.11", "libcublas.so"):
+            try:
+                ctypes.CDLL(lib_name)
+                return True, None
+            except OSError:
+                continue
+        log.warning("libcublas not loadable on Linux, CUDA disabled")
+        return False, "CUDA libraries not found. Install nvidia-cuda-toolkit or use CPU mode."
 
     # First, add local cuDNN to PATH if it exists
     _add_local_cudnn_to_path()

src-pyloid/services/settings.py

@@ -55,6 +55,8 @@ class Settings:
     hold_hotkey_enabled: bool = True
     toggle_hotkey: str = "ctrl+shift+win"
     toggle_hotkey_enabled: bool = False
+    # Transcription settings
+    prepend_space: bool = False  # Add leading space before pasted text
 
 
 class SettingsService:
@@ -83,6 +85,8 @@ def get_settings(self) -> Settings:
             hold_hotkey_enabled=self.db.get_setting("hold_hotkey_enabled", "true") == "true",
             toggle_hotkey=self.db.get_setting("toggle_hotkey", "ctrl+shift+win"),
             toggle_hotkey_enabled=self.db.get_setting("toggle_hotkey_enabled", "false") == "true",
+            # Transcription settings
+            prepend_space=self.db.get_setting("prepend_space", "false") == "true",
         )
         self._cache = settings
         return settings
@@ -104,6 +108,7 @@ def update_settings(
         toggle_hotkey: Optional[str] = None,
         toggle_hotkey_enabled: Optional[bool] = None,
         show_popup: Optional[bool] = None,
+        prepend_space: Optional[bool] = None,
     ) -> Settings:
         if language is not None:
             self.db.set_setting("language", language)
@@ -125,6 +130,8 @@ def update_settings(
             self.db.set_setting("save_audio_to_history", "true" if save_audio_to_history else "false")
         if show_popup is not None:
             self.db.set_setting("show_popup", "true" if show_popup else "false")
+        if prepend_space is not None:
+            self.db.set_setting("prepend_space", "true" if prepend_space else "false")
         # Hotkey settings - normalize before storing for consistent format
         if hold_hotkey is not None:
             self.db.set_setting("hold_hotkey", normalize_hotkey(hold_hotkey))

src-pyloid/services/transcription.py

@@ -118,16 +118,38 @@ def transcribe(
 
         log.debug("Audio stats", length=len(audio), max_amplitude=float(np.abs(audio).max()), mean_amplitude=float(np.abs(audio).mean()))
 
-        segments, info = self._model.transcribe(
-            audio,
-            language=language_arg,
-            beam_size=5,
-            vad_filter=True,
-            vad_parameters=dict(
-                min_silence_duration_ms=500,  # Less aggressive silence detection
-                speech_pad_ms=400,  # More padding around speech
-            ),
-        )
+        try:
+            segments, info = self._model.transcribe(
+                audio,
+                language=language_arg,
+                beam_size=5,
+                vad_filter=True,
+                vad_parameters=dict(
+                    min_silence_duration_ms=500,  # Less aggressive silence detection
+                    speech_pad_ms=400,  # More padding around speech
+                ),
+            )
+        except RuntimeError as e:
+            if "not found or cannot be loaded" in str(e) and self._current_device == "cuda":
+                log.warning("CUDA runtime error during transcription, reloading on CPU", error=str(e))
+                model_name = self._current_model_name
+                repo_id = _get_repo_id(model_name)
+                self._model = WhisperModel(repo_id, device="cpu", compute_type="int8")
+                self._current_device = "cpu"
+                self._current_compute_type = "int8"
+                log.info("Model reloaded on CPU fallback")
+                segments, info = self._model.transcribe(
+                    audio,
+                    language=language_arg,
+                    beam_size=5,
+                    vad_filter=True,
+                    vad_parameters=dict(
+                        min_silence_duration_ms=500,
+                        speech_pad_ms=400,
+                    ),
+                )
+            else:
+                raise
 
         # Combine all segments
         segments_list = list(segments)

src/App.tsx

@@ -12,7 +12,9 @@ import type { Settings } from "@/lib/types";
 // Qt WebEngine on Linux often falls back to software rendering,
 // making backdrop-filter/blur extremely expensive
 // Skip for popup route - it needs transparent background
-if (navigator.platform.startsWith("Linux") && window.location.hash !== "#/popup") {
+const isLinux =
+  (navigator.userAgentData?.platform ?? navigator.userAgent).includes("Linux");
+if (isLinux && window.location.hash !== "#/popup") {
   document.documentElement.classList.add("reduced-effects");
 }
 
@@ -99,7 +101,7 @@ function AppRouter() {
   if (loading) {
     return (
       <div className="min-h-screen flex items-center justify-center">
-        <div className="text-muted-foreground">Loading...</div>
+        <div role="status" aria-live="polite" className="text-muted-foreground">Loading...</div>
       </div>
     );
   }

src/components/HistoryPage.tsx

@@ -59,7 +59,7 @@ export function HistoryPage() {
     try {
       await api.copyToClipboard(text);
       toast.success("Copied to clipboard");
-    } catch (error) {
+    } catch {
       try {
         await navigator.clipboard.writeText(text);
         toast.success("Copied to clipboard");
@@ -124,7 +124,7 @@ export function HistoryPage() {
         <div className="flex flex-col md:flex-row justify-between items-start md:items-end gap-6">
           <div>
             <h1 className="text-4xl md:text-5xl font-bold tracking-tighter text-foreground mb-2">
-              Full <span className="headline-serif text-primary">History</span>
+              Full History
             </h1>
             <p className="text-lg text-muted-foreground/80 font-light max-w-2xl">
               A complete archive of your voice notes and dictations.
@@ -200,12 +200,13 @@ export function HistoryPage() {
                                 </Badge>
                               )}
                             </div>
-                            <div className="opacity-0 group-hover:opacity-100 transition-opacity flex gap-1">
+                            <div className="opacity-0 group-hover:opacity-100 group-focus-within:opacity-100 transition-opacity flex gap-1">
                               <Button
                                 variant="ghost"
                                 size="icon-sm"
                                 className="h-7 w-7 text-muted-foreground hover:text-primary hover:bg-primary/10"
                                 onClick={() => handleCopy(entry.text)}
+                                aria-label="Copy transcription"
                               >
                                 <Copy className="h-3.5 w-3.5" />
                               </Button>
@@ -216,6 +217,7 @@ export function HistoryPage() {
                                   className="h-7 w-7 text-muted-foreground hover:text-primary hover:bg-primary/10"
                                   onClick={() => handlePlayAudio(entry.id)}
                                   disabled={loadingAudioFor === entry.id}
+                                  aria-label="Play audio recording"
                                 >
                                   <FileAudio className="h-3.5 w-3.5" />
                                 </Button>
@@ -225,6 +227,7 @@ export function HistoryPage() {
                                 size="icon-sm"
                                 className="h-7 w-7 text-muted-foreground hover:text-destructive hover:bg-destructive/10"
                                 onClick={() => handleDelete(entry.id)}
+                                aria-label="Delete transcription"
                               >
                                 <Trash2 className="h-3.5 w-3.5" />
                               </Button>

src/components/HistoryTab.tsx

@@ -41,7 +41,7 @@ export function HistoryTab() {
     try {
       await api.copyToClipboard(text);
       toast.success("Copied to clipboard");
-    } catch (error) {
+    } catch {
       // Fallback
       try {
         await navigator.clipboard.writeText(text);