fix linux

.claude/settings.local.json

@@ -63,7 +63,11 @@
       "Bash(echo \"Desktop: $XDG_CURRENT_DESKTOP\")",
       "Bash(nvidia-smi)",
       "Bash(lspci)",
-      "Bash(pacman -Q)"
+      "Bash(pacman -Q)",
+      "Bash(/run/media/chk/Ventoy/projects/chrka/whisper-dictation/.venv-linux/bin/pip install:*)",
+      "Bash(.venv-linux/bin/python build.py)",
+      "Bash(.venv-linux/bin/pip list:*)",
+      "Bash(.venv-linux/bin/python -c \":*)"
     ]
   }
 }

README.md

@@ -41,6 +41,16 @@ sudo apt install python3-tk libayatana-appindicator3-1 wl-clipboard xdotool
 | `wl-clipboard` | Text injection on Wayland (`wl-copy`) |
 | `xdotool` | Simulates Ctrl+V paste on Wayland, text typing on X11 |
 
+**Low-latency hotkey (recommended):**
+
+For fast hotkey response via evdev (instead of the slower XWayland fallback), add your user to the `input` group:
+
+```bash
+sudo usermod -aG input $USER
+```
+
+Log out and back in for the change to take effect.
+
 **Optional (for GPU acceleration):**
 
 Arch/CachyOS:
@@ -48,7 +58,9 @@ Arch/CachyOS:
 sudo pacman -S nvidia cuda
 ```
 
-Without CUDA, the app runs on CPU. Use `int8` compute type and a smaller model (`small` or `base`) for acceptable speed on CPU.
+Note: The system CUDA package may install a newer version (e.g. CUDA 13) than what faster-whisper/ctranslate2 requires (CUDA 12). The CUDA 12 runtime libraries (`nvidia-cublas-cu12`, `nvidia-cudnn-cu12`) are installed via pip in the virtual environment and bundled into the PyInstaller build, so the system CUDA version does not matter for the app itself. The system `nvidia` + `cuda` packages are only needed for the GPU driver and kernel module.
+
+Without an NVIDIA GPU, the app runs on CPU. Use `int8` compute type and a smaller model (`small` or `base`) for acceptable speed on CPU.
 
 **Python:**
 - Python 3.10+

install.sh

@@ -8,8 +8,8 @@ python3 -m venv --system-site-packages .venv-linux
 echo "Installing dependencies..."
 .venv-linux/bin/pip install --upgrade pip
 .venv-linux/bin/pip install -r requirements.txt
+.venv-linux/bin/pip install -r requirements-cuda.txt
 .venv-linux/bin/pip install pyinstaller
-# CUDA: uses system-installed CUDA libs (no pip CUDA packages needed)
 
 echo ""
 echo "Done. Run ./start.sh to launch, or ./build-linux.sh to create a standalone build."

shared_data/vocabulary.json

@@ -1,63 +1,67 @@
-{
-  "words": [
-    "test"
-  ],
-  "replacements": [
-    {
-      "from": "KRA",
-      "to": "KRAH"
-    },
-    {
-      "from": "Atos",
-      "to": "ATHOS"
-    },
-    {
-      "from": "Resistec",
-      "to": "RESISTEC"
-    },
-    {
-      "from": "Resistek",
-      "to": "RESISTEC"
-    },
-    {
-      "from": "HES",
-      "to": "HEES"
-    },
-    {
-      "from": "Ackerschot",
-      "to": "Ackerschott"
-    },
-    {
-      "from": "Carrois",
-      "to": "Kauer"
-    },
-    {
-      "from": "Jouer fixe",
-      "to": "Jour-Fixe"
-    },
-    {
-      "from": "Docuware",
-      "to": "DocuWare"
-    },
-    {
-      "from": "Nates",
-      "to": "Nejc"
-    },
-    {
-      "from": "Bittzeit",
-      "to": "BitSight"
-    },
-    {
-      "from": "Kalmikow",
-      "to": "Kalmykov"
-    },
-    {
-      "from": "Leifert",
-      "to": "Leifer"
-    },
-    {
-      "from": "Kiyosa",
-      "to": "Key-User"
-    }
-  ]
+{
+  "words": [
+    "test"
+  ],
+  "replacements": [
+    {
+      "from": "KRA",
+      "to": "KRAH"
+    },
+    {
+      "from": "Atos",
+      "to": "ATHOS"
+    },
+    {
+      "from": "Resistec",
+      "to": "RESISTEC"
+    },
+    {
+      "from": "Resistek",
+      "to": "RESISTEC"
+    },
+    {
+      "from": "HES",
+      "to": "HEES"
+    },
+    {
+      "from": "Ackerschot",
+      "to": "Ackerschott"
+    },
+    {
+      "from": "Carrois",
+      "to": "Kauer"
+    },
+    {
+      "from": "Jouer fixe",
+      "to": "Jour-Fixe"
+    },
+    {
+      "from": "Docuware",
+      "to": "DocuWare"
+    },
+    {
+      "from": "Nates",
+      "to": "Nejc"
+    },
+    {
+      "from": "Bittzeit",
+      "to": "BitSight"
+    },
+    {
+      "from": "Kalmikow",
+      "to": "Kalmykov"
+    },
+    {
+      "from": "Leifert",
+      "to": "Leifer"
+    },
+    {
+      "from": "Kiyosa",
+      "to": "Key-User"
+    },
+    {
+      "from": "Kashi",
+      "to": "Cachy"
+    }
+  ]
 }

whisper-dictation.spec

@@ -16,13 +16,18 @@ _platform_tag = 'windows' if _is_windows else 'linux'
 
 # ── Platform-specific binaries ────────────────────────────────────────────────
 _binaries = []
+_sp = next(p for p in site.getsitepackages() if p.endswith('site-packages'))
+_nvidia = os.path.join(_sp, 'nvidia')
 if _is_windows:
-    _sp = next(p for p in site.getsitepackages() if p.endswith('site-packages'))
-    _nvidia = os.path.join(_sp, 'nvidia')
     _binaries = [
         (os.path.join(_nvidia, 'cublas', 'bin', '*.dll'), '.'),
         (os.path.join(_nvidia, 'cudnn', 'bin', '*.dll'), '.'),
     ]
+else:
+    _binaries = [
+        (os.path.join(_nvidia, 'cublas', 'lib', 'libcublas.so*'), '.'),
+        (os.path.join(_nvidia, 'cudnn', 'lib', 'libcudnn.so*'), '.'),
+    ]
 
 # ── Platform-specific hidden imports ──────────────────────────────────────────
 _hiddenimports = [

whisper_app/hotkey.py

@@ -1,91 +1,227 @@
-_pynput_loaded = False
-Key = KeyCode = KeyboardListener = None
-
-def _ensure_pynput():
-    global _pynput_loaded, Key, KeyCode, KeyboardListener, _MODIFIER_MAP, _KEY_MAP
-    if _pynput_loaded:
-        return
-    from pynput.keyboard import Listener as _Listener, Key as _Key, KeyCode as _KeyCode
-    Key = _Key
-    KeyCode = _KeyCode
-    KeyboardListener = _Listener
-    _MODIFIER_MAP.update({
-        "ctrl": {Key.ctrl_l, Key.ctrl_r},
-        "ctrl_l": {Key.ctrl_l}, "ctrl_r": {Key.ctrl_r},
-        "shift": {Key.shift_l, Key.shift_r},
-        "shift_l": {Key.shift_l}, "shift_r": {Key.shift_r},
-        "alt": {Key.alt_l, Key.alt_r},
-        "alt_l": {Key.alt_l}, "alt_r": {Key.alt_r},
-    })
-    _KEY_MAP.update({
-        "space": Key.space, "tab": Key.tab, "enter": Key.enter,
-        "esc": Key.esc, "escape": Key.esc,
-        "up": Key.up, "down": Key.down, "left": Key.left, "right": Key.right,
-        "home": Key.home, "end": Key.end, "page_up": Key.page_up, "page_down": Key.page_down,
-        "insert": Key.insert, "delete": Key.delete, "backspace": Key.backspace,
-    })
-    for i in range(1, 13):
-        _KEY_MAP[f"f{i}"] = getattr(Key, f"f{i}")
-    _pynput_loaded = True
-
-_MODIFIER_MAP = {}
-_KEY_MAP = {}
-
-
-def _parse_hotkey(hotkey_str):
-    """Parse hotkey string into (modifier_sets, trigger_key).
-    Returns: (list of sets-of-pynput-keys for each modifier, pynput key for trigger)
-    """
-    _ensure_pynput()
-    parts = [p.strip().lower() for p in hotkey_str.split("+")]
-    modifiers = []
-    for p in parts[:-1]:
-        if p in _MODIFIER_MAP:
-            modifiers.append(_MODIFIER_MAP[p])
-        elif p in _KEY_MAP:
-            modifiers.append({_KEY_MAP[p]})
-        else:
-            modifiers.append({KeyCode.from_char(p)})
-    trigger_part = parts[-1]
-    if trigger_part in _KEY_MAP:
-        trigger = _KEY_MAP[trigger_part]
-    elif trigger_part in _MODIFIER_MAP:
-        trigger = next(iter(_MODIFIER_MAP[trigger_part]))
-    else:
-        trigger = KeyCode.from_char(trigger_part)
-    return modifiers, trigger
-
-
-class HotkeyListener:
-    """Hold-to-record hotkey using pynput. No root required on X11."""
-
-    def __init__(self, hotkey_str, on_press, on_release):
-        self._modifiers, self._trigger = _parse_hotkey(hotkey_str)
-        self._on_press = on_press
-        self._on_release = on_release
-        self._pressed = set()
-        self._active = False
-        self._listener = KeyboardListener(on_press=self._key_down, on_release=self._key_up)
-        self._listener.daemon = True
-        self._listener.start()
-
-    def _matches_trigger(self, key):
-        return key == self._trigger
-
-    def _modifiers_held(self):
-        return all(any(k in self._pressed for k in mod_set) for mod_set in self._modifiers)
-
-    def _key_down(self, key):
-        self._pressed.add(key)
-        if not self._active and self._matches_trigger(key) and self._modifiers_held():
-            self._active = True
-            self._on_press()
-
-    def _key_up(self, key):
-        self._pressed.discard(key)
-        if self._active and self._matches_trigger(key):
-            self._active = False
-            self._on_release()
-
-    def stop(self):
-        self._listener.stop()
+import os
+import threading
+
+# ── evdev key name → hotkey string mapping ───────────────────────────────────
+_EVDEV_MODIFIER_MAP = {
+    "ctrl": {"KEY_LEFTCTRL", "KEY_RIGHTCTRL"},
+    "ctrl_l": {"KEY_LEFTCTRL"}, "ctrl_r": {"KEY_RIGHTCTRL"},
+    "shift": {"KEY_LEFTSHIFT", "KEY_RIGHTSHIFT"},
+    "shift_l": {"KEY_LEFTSHIFT"}, "shift_r": {"KEY_RIGHTSHIFT"},
+    "alt": {"KEY_LEFTALT", "KEY_RIGHTALT"},
+    "alt_l": {"KEY_LEFTALT"}, "alt_r": {"KEY_RIGHTALT"},
+}
+
+_EVDEV_KEY_MAP = {
+    "space": "KEY_SPACE", "tab": "KEY_TAB", "enter": "KEY_ENTER",
+    "esc": "KEY_ESC", "escape": "KEY_ESC",
+    "up": "KEY_UP", "down": "KEY_DOWN", "left": "KEY_LEFT", "right": "KEY_RIGHT",
+    "home": "KEY_HOME", "end": "KEY_END", "page_up": "KEY_PAGEUP", "page_down": "KEY_PAGEDOWN",
+    "insert": "KEY_INSERT", "delete": "KEY_DELETE", "backspace": "KEY_BACKSPACE",
+}
+for _i in range(1, 13):
+    _EVDEV_KEY_MAP[f"f{_i}"] = f"KEY_F{_i}"
+
+
+def _parse_hotkey_evdev(hotkey_str):
+    """Parse hotkey string into evdev key code sets."""
+    parts = [p.strip().lower() for p in hotkey_str.split("+")]
+    modifiers = []
+    for p in parts[:-1]:
+        if p in _EVDEV_MODIFIER_MAP:
+            modifiers.append(_EVDEV_MODIFIER_MAP[p])
+        elif p in _EVDEV_KEY_MAP:
+            modifiers.append({_EVDEV_KEY_MAP[p]})
+        else:
+            modifiers.append({f"KEY_{p.upper()}"})
+    trigger_part = parts[-1]
+    if trigger_part in _EVDEV_KEY_MAP:
+        trigger = _EVDEV_KEY_MAP[trigger_part]
+    elif trigger_part in _EVDEV_MODIFIER_MAP:
+        trigger = next(iter(_EVDEV_MODIFIER_MAP[trigger_part]))
+    else:
+        trigger = f"KEY_{trigger_part.upper()}"
+    return modifiers, trigger
+
+
+def _can_use_evdev():
+    """Check if evdev is available and we have access to input devices."""
+    if os.name == "nt":
+        return False
+    try:
+        import evdev
+        devices = [evdev.InputDevice(p) for p in evdev.list_devices()]
+        keyboards = [d for d in devices if 1 in d.capabilities()]  # EV_KEY
+        for d in devices:
+            d.close()
+        return len(keyboards) > 0
+    except Exception:
+        return False
+
+
+class _EvdevListener:
+    """Low-latency hotkey listener using evdev (Linux kernel input)."""
+
+    def __init__(self, hotkey_str, on_press, on_release):
+        self._modifiers, self._trigger = _parse_hotkey_evdev(hotkey_str)
+        self._on_press = on_press
+        self._on_release = on_release
+        self._pressed = set()
+        self._active = False
+        self._stop_event = threading.Event()
+        self._thread = threading.Thread(target=self._run, daemon=True)
+        self._thread.start()
+
+    def _run(self):
+        import evdev
+        import select
+
+        devices = [evdev.InputDevice(p) for p in evdev.list_devices()]
+        keyboards = [d for d in devices if 1 in d.capabilities()]
+        # close non-keyboard devices
+        for d in devices:
+            if d not in keyboards:
+                d.close()
+
+        if not keyboards:
+            return
+
+        while not self._stop_event.is_set():
+            r, _, _ = select.select(keyboards, [], [], 0.5)
+            for dev in r:
+                try:
+                    for event in dev.read():
+                        if event.type != 1:  # EV_KEY
+                            continue
+                        code_name = evdev.ecodes.KEY.get(event.code)
+                        if isinstance(code_name, list):
+                            code_name = code_name[0]
+                        if code_name is None:
+                            continue
+                        if event.value == 1:  # key down
+                            self._key_down(code_name)
+                        elif event.value == 0:  # key up
+                            self._key_up(code_name)
+                except OSError:
+                    pass
+
+        for d in keyboards:
+            d.close()
+
+    def _key_down(self, code_name):
+        self._pressed.add(code_name)
+        if not self._active and code_name == self._trigger and self._modifiers_held():
+            self._active = True
+            self._on_press()
+
+    def _key_up(self, code_name):
+        self._pressed.discard(code_name)
+        if self._active and code_name == self._trigger:
+            self._active = False
+            self._on_release()
+
+    def _modifiers_held(self):
+        return all(
+            any(k in self._pressed for k in mod_set)
+            for mod_set in self._modifiers
+        )
+
+    def stop(self):
+        self._stop_event.set()
+
+
+# ── pynput fallback (Windows, or Linux without evdev/input group) ────────────
+
+_pynput_loaded = False
+Key = KeyCode = KeyboardListener = None
+_PYNPUT_MODIFIER_MAP = {}
+_PYNPUT_KEY_MAP = {}
+
+
+def _ensure_pynput():
+    global _pynput_loaded, Key, KeyCode, KeyboardListener
+    if _pynput_loaded:
+        return
+    from pynput.keyboard import Listener as _Listener, Key as _Key, KeyCode as _KeyCode
+    Key = _Key
+    KeyCode = _KeyCode
+    KeyboardListener = _Listener
+    _PYNPUT_MODIFIER_MAP.update({
+        "ctrl": {Key.ctrl_l, Key.ctrl_r},
+        "ctrl_l": {Key.ctrl_l}, "ctrl_r": {Key.ctrl_r},
+        "shift": {Key.shift_l, Key.shift_r},
+        "shift_l": {Key.shift_l}, "shift_r": {Key.shift_r},
+        "alt": {Key.alt_l, Key.alt_r},
+        "alt_l": {Key.alt_l}, "alt_r": {Key.alt_r},
+    })
+    _PYNPUT_KEY_MAP.update({
+        "space": Key.space, "tab": Key.tab, "enter": Key.enter,
+        "esc": Key.esc, "escape": Key.esc,
+        "up": Key.up, "down": Key.down, "left": Key.left, "right": Key.right,
+        "home": Key.home, "end": Key.end, "page_up": Key.page_up, "page_down": Key.page_down,
+        "insert": Key.insert, "delete": Key.delete, "backspace": Key.backspace,
+    })
+    for i in range(1, 13):
+        _PYNPUT_KEY_MAP[f"f{i}"] = getattr(Key, f"f{i}")
+    _pynput_loaded = True
+
+
+def _parse_hotkey_pynput(hotkey_str):
+    _ensure_pynput()
+    parts = [p.strip().lower() for p in hotkey_str.split("+")]
+    modifiers = []
+    for p in parts[:-1]:
+        if p in _PYNPUT_MODIFIER_MAP:
+            modifiers.append(_PYNPUT_MODIFIER_MAP[p])
+        elif p in _PYNPUT_KEY_MAP:
+            modifiers.append({_PYNPUT_KEY_MAP[p]})
+        else:
+            modifiers.append({KeyCode.from_char(p)})
+    trigger_part = parts[-1]
+    if trigger_part in _PYNPUT_KEY_MAP:
+        trigger = _PYNPUT_KEY_MAP[trigger_part]
+    elif trigger_part in _PYNPUT_MODIFIER_MAP:
+        trigger = next(iter(_PYNPUT_MODIFIER_MAP[trigger_part]))
+    else:
+        trigger = KeyCode.from_char(trigger_part)
+    return modifiers, trigger
+
+
+class _PynputListener:
+    """Hotkey listener using pynput (Windows / fallback)."""
+
+    def __init__(self, hotkey_str, on_press, on_release):
+        self._modifiers, self._trigger = _parse_hotkey_pynput(hotkey_str)
+        self._on_press = on_press
+        self._on_release = on_release
+        self._pressed = set()
+        self._active = False
+        self._listener = KeyboardListener(on_press=self._key_down, on_release=self._key_up)
+        self._listener.daemon = True
+        self._listener.start()
+
+    def _key_down(self, key):
+        self._pressed.add(key)
+        if not self._active and key == self._trigger and self._modifiers_held():
+            self._active = True
+            self._on_press()
+
+    def _key_up(self, key):
+        self._pressed.discard(key)
+        if self._active and key == self._trigger:
+            self._active = False
+            self._on_release()
+
+    def _modifiers_held(self):
+        return all(any(k in self._pressed for k in mod_set) for mod_set in self._modifiers)
+
+    def stop(self):
+        self._listener.stop()
+
+
+# ── Public API ───────────────────────────────────────────────────────────────
+
+def HotkeyListener(hotkey_str, on_press, on_release):
+    """Create the best available hotkey listener for the current platform."""
+    if _can_use_evdev():
+        return _EvdevListener(hotkey_str, on_press, on_release)
+    return _PynputListener(hotkey_str, on_press, on_release)