Doubao-powered voice input for Linux, Wayland and Omarchy
"""Main GtkApplication: orchestrates trigger, ASR, overlay, injector, login.
Lifecycle:
1. On first activate: build all components. Show login window if no
saved credentials; otherwise just start the PTT listener and
show the control window.
2. On subsequent activate (single-instance re-launch via the desktop
entry or `python -m doubao_input` again): just present the control
window.
3. Quit: stop everything cleanly.
"""
from __future__ import annotations
import logging
import signal
import shutil
from threading import Thread
from dataclasses import replace
import gi
gi.require_version("Gtk", "4.0")
from gi.repository import Gio, GLib, Gtk # type: ignore
from doubao_input.doubao.app_state import AppState, LoginStatus, RecordingState
from doubao_input.doubao.audio_capture import AudioCapture
from doubao_input.inject.delivery import Delivery
from doubao_input.inject.target import focused_target
from doubao_input.result import RecentResult
from doubao_input.doubao.params_store import ParamsStore
from doubao_input.doubao.transcription import TranscriptionManager
from doubao_input.inject.injector import Injector
from doubao_input.trigger.reader import TriggerReader
from doubao_input.trigger.controller import TriggerController
from doubao_input.setup_session import SetupSession
from doubao_input.inject.worker import InputWorker
from doubao_input.preferences import apply_preferences
from doubao_input.settings import (Settings, canonical_shortcut,
key_codes_match, trigger_shortcut_display)
from doubao_input.i18n import set_language, tr
from doubao_input.ui.control_window import ControlWindow
from doubao_input.ui.setup_actions import SetupActions
from doubao_input.ui.overlay import Overlay
from doubao_input.polish import ApiKeyStore, PolishManager
from doubao_input.updates import UpdateChecker
from doubao_input.update_install import prepare as prepare_update, launch as launch_update, consume_result as consume_update_result
from doubao_input.polish_preview import PolishPreview
from doubao_input.trigger.escape_guard import EscapeGuard
from doubao_input.diagnostics import DiagnosticTrace, report as diagnostic_report
from doubao_input.recognition_providers import recognition_provider
logger = logging.getLogger(__name__)
from doubao_input.app_voice import VoiceMixin
from doubao_input.app_runtime import RuntimeMixin
class DoubaoInputApp(VoiceMixin, RuntimeMixin, Gtk.Application):
def __init__(self, background: bool = False, trigger_debug: bool = False) -> None:
super().__init__(
application_id="md.lifeos.DoubaoSay",
flags=Gio.ApplicationFlags.HANDLES_COMMAND_LINE,
)
# PyGObject does NOT dispatch do_activate() automatically in all
# versions — connect the signal explicitly so the first run
# actually shows the UI.
self.connect("activate", self._on_activate)
self.connect("shutdown", self._on_shutdown)
self.connect("command-line", self._on_command_line)
open_control = Gio.SimpleAction.new("open-control", None)
open_control.connect("activate", lambda *_: self._control.show() if self._control else None)
self.add_action(open_control)
for name in ("background", "trigger-debug"):
self.add_main_option(name, 0, GLib.OptionFlags.NONE, GLib.OptionArg.NONE, name, None)
GLib.unix_signal_add(GLib.PRIORITY_DEFAULT, signal.SIGTERM, self._terminate)
self._setup_done = False
self._escape_guard = None
self._escape_timer = None
self._background = background
self._trigger_debug_requested = trigger_debug
self._debug_session = None
self.app_state = AppState()
self._overlay: Overlay | None = None
self._control: ControlWindow | None = None
self._settings_window = None
self._login_window = None
self._login_attempt = None
self._tray = None
self._triggers = None
self._setup_session = None
self._tm: TranscriptionManager | None = None
self._injector: Injector | None = None
self._enter_after_paste = False
self.app_state.connect("recording-state-changed", self._finish_preview_state)
self.recent = RecentResult()
from doubao_input.personal import PersonalStore
self._personal_store = PersonalStore()
self._history_id = None
self._personal_window = None
self._target = None
self._recovery_timer = None
self._delivery = None
self._polisher = None
self._update_checker = None
self._update_installing = False
self._update_timer = None
self._polish_context = None
self._polish_mode = None
self._preview_polish = PolishPreview(GLib.timeout_add, GLib.source_remove)
self._rms_speaking = False
self._asr_probe = None
self._diagnostics = DiagnosticTrace()
from doubao_input.input_latency import InputLatency
self._input_latency = InputLatency()
self.app_state.connect("transcription-text-changed", self._prepolish_text_changed)
try:
self.settings = Settings.load()
except (ValueError, TypeError, OSError):
logger.exception("Invalid settings; using defaults without overwriting the file")
self.settings = Settings.desktop_defaults()
set_language(self.settings.language)
@property
def _paste_pending(self):
return bool(self._delivery and self._delivery.busy)
@property
def _preview_testing(self):
return bool(self._setup_session and self._setup_session.voice_active)
@property
def _mic_test_running(self):
return bool(self._setup_session and self._setup_session.microphone_active)
# ---- Gtk.Application hooks (signal handlers, not do_*) ----
def _on_command_line(self, _app, command_line):
options = command_line.get_options_dict()
background = options.contains("background") or options.contains("trigger-debug")
if self._setup_done and background:
return 0
if not self._setup_done:
self._background = self._background or background
self._trigger_debug_requested = self._trigger_debug_requested or options.contains("trigger-debug")
self.activate()
return 0
def _terminate(self):
self._quit()
return GLib.SOURCE_REMOVE
def _on_activate(self, _app) -> None:
logger.info("on_activate: setup_done=%s", self._setup_done)
first_activation = not self._setup_done
if first_activation:
self._setup_done = True
self._build()
if self._background:
# No visible ApplicationWindow exists in daemon mode, so keep
# GApplication alive for evdev/ASR callbacks explicitly.
self.hold()
if self._control and (not first_activation or not self._background):
self._control.show()
# The tray indicates idle readiness; overlays are reserved for capture.
# ---- build ----
# ---- Voice commands (GTK main thread) ----
# ---- Paste ----
def _summary(self):
key = trigger_shortcut_display(self.settings.doubao_key, self.settings.doubao_modifiers)
setup = getattr(self, "_setup_session", None)
return {"key": key, "key_code": self.settings.doubao_key,
"key_modifiers": self.settings.doubao_modifiers,
"microphone": self.settings.microphone or tr("System default", "系统默认"),
"microphone_id": self.settings.microphone,
"microphone_ok": bool(setup and setup.microphone_ok),
"asr_provider": self.settings.asr_provider,
"asr_provider_name": recognition_provider(
self.settings.asr_provider).name,
"voice_test_ok": bool(setup and setup.voice_ok),
"onboarding_complete": self.settings.onboarding_complete,
"result": self.recent.text, "status": self.recent.status}
def _complete_setup(self):
if self.settings.onboarding_complete:
self._control.hide()
return
if self.app_state.login_status != LoginStatus.LOGGED_IN:
self._control.set_feedback(tr("Configure recognition credentials first, or return later.", "请先配置语音识别凭证,或稍后继续设置。"))
return
missing = []
if not self._setup_session.microphone_ok:
missing.append(tr("microphone check", "麦克风检查"))
if not self._setup_session.voice_ok:
missing.append(tr("voice test", "语音测试"))
if not self.settings.doubao_key:
missing.append(tr("enabled trigger key", "已启用的触发键"))
if missing:
self._control.set_feedback(
tr("Complete these steps first: ", "请先完成以下步骤:")
+ tr(", ", "、").join(missing)
+ tr(". You can return later.", "。也可以稍后继续。"))
return
from dataclasses import replace
self.apply_settings(replace(self.settings, onboarding_complete=True))
self._control.hide()
def _begin_key_capture(self, callback, preview=None):
if self._busy():
raise ValueError(tr("Finish the current operation first.", "请先结束当前操作。"))
if preview is None:
self._triggers.begin_capture(callback)
else:
self._triggers.begin_capture(callback, preview)
def _end_key_capture(self):
if self._triggers:
self._triggers.end_capture()
def _sign_out(self):
if self._busy():
raise ValueError(tr("Finish the current operation first.", "请先结束当前操作。"))
self._login_attempt = None
if self._login_window:
self._login_window.destroy()
self._login_window = None
provider = recognition_provider(
getattr(getattr(self, "settings", None), "asr_provider", "doubao"))
store = provider.credential_store
try:
store.clear()
except OSError as error:
raise ValueError(tr("Could not clear saved credentials. Check configuration folder permissions.",
"无法清除保存的凭证,请检查配置目录权限。")) from error
self.app_state.login_status = LoginStatus.NOT_LOGGED_IN
message = (tr("API key cleared.", "已清除 API Key。")
if provider.uses_api_key else
tr("Saved credentials cleared. Website sessions may require signing out separately.",
"已清除保存的凭证;网站会话可能还需单独退出登录。"))
self._control.set_feedback(message)
def _restart(self):
if self._busy() or self._login_window:
raise ValueError(tr("Finish recording and close sign-in before restarting.", "请结束录音并关闭登录窗口后重启。"))
# Rebuild translated windows in-process, preserving the singleton and tray.
set_language(self.settings.language)
current = getattr(self, "_settings_window", None)
if self._tray:
self._tray.menu.refresh()
if current:
current.window.destroy()
self._settings_window = None
self._control.destroy()
self._control.show()
if current:
self._show_settings()
def _show_settings(self):
from doubao_input.ui.settings_window import SettingsWindow
current = getattr(self, "_settings_window", None)
if current is not None and current.window.get_visible():
current.show()
return
self._settings_window = SettingsWindow(self._control.window, self.settings, self.apply_settings,
capture_key=self._begin_key_capture, cancel_capture=self._end_key_capture,
sign_out=self._sign_out, restart=self._restart, login=self._show_login,
preview=self._preview_overlay, apply_key=self._apply_trigger_key,
asr_has_key=self._api_key_has_saved, save_asr=self._save_asr_key,
clear_asr=self._clear_asr_key, test_asr=self._test_asr_key,
diagnostic_report=lambda: diagnostic_report(
self.settings, recording=self.app_state.is_recording,
trace=self._diagnostics))
self._settings_window.show()
def _preview_overlay(self):
if self._busy():
raise ValueError(tr("Finish the current operation first.", "请先结束当前操作。"))
self._setup_session.show_appearance()
def _cancel_voice_test(self):
self._setup_session.cancel_voice()
def _test_voice(self):
if self._preview_testing:
self._voice_stop()
return
if self._busy():
self._control.set_feedback(tr("Finish the current recording or microphone check first.", "请先结束当前录音或麦克风检查。"))
return
if self.app_state.login_status != LoginStatus.LOGGED_IN:
self._control.set_feedback(tr("Configure your recognition service in step 1 first.", "请先在第一步配置语音识别服务。"))
return
self._setup_session.begin_voice()
self._voice_start()
def _finish_preview_state(self, _state, value):
self._sync_escape()
if value == "idle":
if self._setup_session:
self._setup_session.recording_idle()
if not self._paste_pending:
self._enter_after_paste = False
# ---- Login ----
def _sync_escape(self):
if self._escape_guard:
self._escape_guard.sync(
self.app_state.recording_state != RecordingState.IDLE
or self._paste_pending or bool(self._polisher and self._polisher.busy)
or bool(self._recovery_timer))
return True
def _show_login(self) -> None:
if self._busy():
self._control.set_feedback(tr("Finish the current operation before opening sign-in.", "请先结束当前操作,再打开登录窗口。"))
return
from doubao_input.doubao.login_window import LoginWindow
if self._login_window is None:
lw = LoginWindow(self.app_state)
lw._on_login_status_change = lambda status, nickname: (
self._on_login_detected(status, nickname) if self._login_window is lw else None)
lw.on_close = self._cancel_login_attempt
self._login_window = lw
self._login_window.show()
def _connect_recognition(self) -> None:
provider = recognition_provider(self.settings.asr_provider)
if provider.uses_api_key:
self._control.set_feedback(tr(
"Add or test your speech API key in Settings.",
"请在设置中填写或测试语音 API Key。"))
self._show_settings()
else:
self._show_login()
def _show_recognition_issue(self, title: str, detail: str) -> None:
self._control.set_feedback(title + tr(": ", ":") + detail)
self._overlay.show_error(title, detail, self._show_settings)
def _prompt_recognition_setup(self) -> None:
provider = recognition_provider(self.settings.asr_provider)
detail = (tr("Open settings to check your account, service access and API key.",
"请打开设置检查账户、服务权限和 API Key。")
if provider.uses_api_key else
tr("Open settings to sign in again.", "请打开设置重新登录。"))
self._show_recognition_issue(
tr("Voice recognition is unavailable", "语音识别暂不可用"), detail)
def _on_service_error(self, reason: str) -> None:
if reason == "credentials":
self._prompt_recognition_setup()
elif reason == "timeout":
self._show_recognition_issue(
tr("Recognition timed out", "语音识别超时"),
tr("Any partial text was saved in the app. Check the service and try again.",
"部分文字已保留在应用中,请检查服务后重试。"))
else:
self._show_recognition_issue(
tr("Voice recognition failed", "语音识别失败"),
tr("Check your network or service status, then try again.",
"请检查网络或服务状态后重试。"))
def _api_key_has_saved(self):
provider = recognition_provider(self.settings.asr_provider)
return provider.uses_api_key and provider.credential_store.has_saved()
def _save_asr_key(self, key):
if self._busy():
raise ValueError(tr("Finish the current operation first.", "请先结束当前操作。"))
provider = recognition_provider(self.settings.asr_provider)
if not provider.uses_api_key:
raise ValueError(tr(
"The selected service does not use an API key.",
"当前服务不使用 API Key。"))
if key:
provider.credential_store.save(
provider.credentials_from_secret(key.strip()))
self._sync_recognition_status()
if self._control:
self._control.refresh()
def _clear_asr_key(self):
if self._busy():
raise ValueError(tr("Finish the current operation first.", "请先结束当前操作。"))
provider = recognition_provider(self.settings.asr_provider)
if provider.uses_api_key:
provider.credential_store.clear()
self._sync_recognition_status()
if self._control:
self._control.refresh()
def _test_asr_key(self, key, completed):
if self._busy():
raise ValueError(tr("Finish the current operation first.", "请先结束当前操作。"))
provider = recognition_provider(self.settings.asr_provider)
if not provider.uses_api_key:
raise ValueError(tr(
"The selected service does not use an API key.",
"当前服务不使用 API Key。"))
credentials = (provider.credentials_from_secret(key.strip()) if key else
provider.credential_store.load())
if credentials is None:
raise ValueError(tr("Enter an API key first.", "请先填写 API Key。"))
credentials.validate()
probe = self._asr_probe = provider.new_client()
finished = False
def complete(result, error):
nonlocal finished
if finished or self._asr_probe is not probe:
return GLib.SOURCE_REMOVE
finished = True
probe.disconnect()
self._asr_probe = None
if not error:
self._sync_recognition_status()
if self._control:
self._control.refresh()
completed(result, error)
return GLib.SOURCE_REMOVE
def deliver(result, error):
GLib.idle_add(complete, result, error)
# A successful WebSocket handshake validates the provider credential.
# No microphone data is needed for this settings test.
probe.on_open = lambda: deliver(tr("API key accepted", "API Key 可用"), "")
probe.on_auth_error = lambda: deliver(None, tr(
"API key rejected; check service activation and project access",
"API Key 被拒绝,请检查服务开通状态和项目权限"))
probe.on_error = lambda _error: deliver(None, tr(
"Connection failed; check the network and service status",
"连接失败,请检查网络和服务状态"))
try:
probe.prepare()
probe.connect(credentials)
except Exception:
self._asr_probe = None
probe.disconnect()
raise
def _cancel_login_attempt(self):
self._login_attempt = None
def _on_login_detected(self, status: str, nickname: str | None) -> None:
if status == "loggedIn":
logger.info("Web sign-in detected")
# Extract params after a short delay, then close the WebView.
owner = self._login_window
token = self._login_attempt = object()
GLib.timeout_add(800, self._extract_and_close_login, owner, token)
def _extract_and_close_login(self, owner=None, token=None) -> bool:
if token is None:
owner = self._login_window
token = self._login_attempt = object()
def current():
return (self._login_attempt is token and self._login_window is owner
and owner is not None and owner.is_active)
if not current():
return False
def on_params(params):
if not current():
return
self._login_attempt = None
if params:
try:
ParamsStore.save(params)
except (OSError, ValueError):
logger.warning("Could not persist sign-in credentials")
self._control.set_feedback(tr(
"Could not save sign-in. Check configuration folder permissions and disk space, then retry.",
"无法保存登录信息,请检查配置目录权限和磁盘空间后重试。"))
return
self.app_state.login_status = LoginStatus.LOGGED_IN
self._control.set_feedback(tr("Sign-in saved. Continue to the microphone check.", "登录信息已保存,请继续检查麦克风。"))
else:
self._control.set_feedback(tr("Sign-in is not ready yet. Complete the web sign-in, then try again.", "尚未完成登录,请在网页中完成登录后重试。"))
return
if self._login_window:
try:
self._login_window.hide()
except Exception:
pass
try:
self._login_window.destroy()
except Exception:
pass
self._login_window = None
self._control.advance_after_login()
owner.extract_params_async(on_params)
return False
def _provide_params(self, callback) -> None:
if self._login_window and self._login_window.is_active:
self._login_window.extract_params_async(callback)
else:
self._show_login()
callback(None)
def _on_auth_expired(self) -> None:
self._login_attempt = None
self.app_state.login_status = LoginStatus.NOT_LOGGED_IN
provider = recognition_provider(self.settings.asr_provider)
if provider.uses_api_key:
detail = (tr("Check your balance, model access and API key in Settings.",
"请在设置中检查余额、模型权限和 API Key。")
if provider.id == "volcengine" else
tr("Check your account balance and API key in Settings.",
"请在设置中检查账户余额和 API Key。"))
self._show_recognition_issue(
tr("Recognition service rejected the request", "语音识别服务拒绝请求"),
detail)
else:
self._show_recognition_issue(
tr("Doubao sign-in expired", "豆包登录已失效"),
tr("Open settings to sign in again.", "请打开设置重新登录。"))
def _check_mic(self) -> None:
if self._busy():
self._control.set_feedback(tr("Finish the current recording before checking the microphone.", "请先结束当前录音,再检查麦克风。"))
return
self._setup_session.check_microphone()
def _quit(self) -> None:
if self._personal_window:
self._personal_window.close()
self._login_attempt = None
self._cancel_input()
if self._triggers:
self._triggers.close()
if self._delivery:
self._delivery.close(self._injector.close)
self.quit()