feat(settings): show Voxtype runtime controls

.github/workflows/midscene-omarchy-4.0.3.yml

@@ -50,6 +50,7 @@ on:
           - omarchy-shard-4
           - omarchy-shell
           - omarchy-providers
+          - omarchy-voxtype-settings
 
 permissions:
   contents: read

.github/workflows/midscene-ubuntu-22.04.yml

@@ -41,6 +41,7 @@ on:
           - ubuntu-shard-4
           - ubuntu-polishing
           - ubuntu-providers
+          - ubuntu-voxtype-settings
 
 permissions:
   contents: read

README.md

@@ -87,6 +87,11 @@ file-mode completion API. Install Voxtype separately, run
 `voxtype setup --download`, start its daemon, then use **Refresh Voxtype status**
 in onboarding. The engine, model, language, audio device, and acceleration remain
 configured in Voxtype; this integration does not rewrite its configuration.
+The Settings window reads Voxtype's versioned schema and extended status to show
+the running daemon version, state, engine, model, audio device, compute backend,
+schema version, and config path. **Open Voxtype configuration** launches
+`voxtype configure` in an available terminal, so Voxtype still validates and
+writes every setting itself.
 
 For each confirmed dictation, Doubao Say asks Voxtype to write one transcript in
 a private per-user runtime directory, waits for the `.done` completion signal,

README.zh-CN.md

@@ -66,6 +66,9 @@ API Key 保存在 `~/.config/doubao-say/deepgram_api_key`（或对应的
 Voxtype，运行 `voxtype setup --download`，启动 daemon，再在首次设置中点击
 **刷新 Voxtype 状态**。引擎、模型、语言、音频设备和加速方式仍在 Voxtype 中配置；
 该接入不会改写 Voxtype 配置。
+设置窗口会读取 Voxtype 的版本化 Schema 和扩展状态，显示 daemon 版本、状态、引擎、
+模型、音频设备、计算后端、Schema 版本和配置文件路径。点击**打开 Voxtype 配置**后，
+应用会在可用终端里启动 `voxtype configure`，所有配置仍由 Voxtype 自己校验和写入。
 
 每次听写手势确认后，豆包说会让 Voxtype 把一份转写写入当前用户的私有 runtime 目录，
 等待 `.done` 完成信号，读取原子写入的最终文字，然后删除这两个文件。若 Voxtype 已在

src/doubao_input/app.py

@@ -854,11 +854,26 @@ def _show_settings(self):
             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,
+            voxtype_details=self._voxtype_details,
+            configure_voxtype=self._configure_voxtype,
             diagnostic_report=lambda: diagnostic_report(
                 self.settings, recording=self.app_state.is_recording,
                 trace=self._diagnostics))
         self._settings_window.show()
 
+    @staticmethod
+    def _voxtype_details():
+        from doubao_input.voxtype.control import inspect_details
+        return inspect_details()
+
+    @staticmethod
+    def _configure_voxtype():
+        from doubao_input.voxtype.control import launch_configure
+        try:
+            launch_configure()
+        except RuntimeError as error:
+            raise ValueError(str(error)) from error
+
     def _preview_overlay(self):
         if self._busy():
             raise ValueError(tr("Finish the current operation first.", "请先结束当前操作。"))

src/doubao_input/ui/settings_window.py

@@ -1,5 +1,7 @@
 """Grouped preferences, safe hardware-key capture and allowlisted diagnostics."""
 from dataclasses import replace
+import subprocess
+import threading
 from gi.repository import GLib, Gtk, Gdk
 from doubao_input.ui.trigger_picker import TriggerPicker
 from doubao_input.i18n import LANGUAGES, tr
@@ -22,6 +24,8 @@ def __init__(self, parent, settings, apply, *, capture_key=None,
                  apply_key=None, asr_has_key=lambda: False,
                  save_asr=lambda key: None, clear_asr=lambda: None,
                  test_asr=lambda key, completed: None,
+                 voxtype_details=lambda: None,
+                 configure_voxtype=lambda: None,
                  diagnostic_report=None):
         self.window = Gtk.Window(title=tr("Doubao Say Settings", "豆包说设置"), transient_for=parent, modal=True)
         self.window.set_default_size(540, 580)
@@ -35,6 +39,9 @@ def __init__(self, parent, settings, apply, *, capture_key=None,
         self._save_asr = save_asr
         self._clear_asr = clear_asr
         self._test_asr = test_asr
+        self._voxtype_details = voxtype_details
+        self._configure_voxtype = configure_voxtype
+        self._voxtype_request_id = 0
         self._asr_save_source = 0
         self._asr_testing = False
         self._diagnostic_report = diagnostic_report or (lambda: report(self._settings))
@@ -141,6 +148,25 @@ def run(*_):
         self.asr_details.append(self.asr_status)
         self.asr_details.set_visible(selected_provider.uses_api_key)
         box.append(self.asr_details)
+        self.voxtype_details = Gtk.Box(
+            orientation=Gtk.Orientation.VERTICAL, spacing=8)
+        self.voxtype_summary = Gtk.Label(
+            xalign=0, wrap=True, selectable=True)
+        self.voxtype_details.append(self.voxtype_summary)
+        voxtype_actions = Gtk.Box(spacing=8, homogeneous=True)
+        self.voxtype_refresh = Gtk.Button(label=tr(
+            "Refresh Voxtype status", "刷新 Voxtype 状态"))
+        self.voxtype_refresh.connect(
+            "clicked", self._refresh_voxtype_details)
+        self.voxtype_configure = Gtk.Button(label=tr(
+            "Open Voxtype configuration", "打开 Voxtype 配置"))
+        self.voxtype_configure.connect(
+            "clicked", self._open_voxtype_configuration)
+        voxtype_actions.append(self.voxtype_refresh)
+        voxtype_actions.append(self.voxtype_configure)
+        self.voxtype_details.append(voxtype_actions)
+        self.voxtype_details.set_visible(selected_provider.is_local)
+        box.append(self.voxtype_details)
 
         section(tr("Input", "输入"))
         self.input_method = Gtk.DropDown.new_from_strings([
@@ -468,8 +494,66 @@ def _sync_provider_details(self):
         self.clear_credentials_button.set_visible(provider.interactive_auth)
         self.provider_help.set_text(provider.credential_help)
         self.provider_help.set_visible(bool(provider.credential_help))
+        self.voxtype_details.set_visible(provider.is_local)
+        if provider.is_local:
+            self._refresh_voxtype_details()
+        else:
+            self._voxtype_request_id += 1
+            self.voxtype_refresh.set_sensitive(True)
         self.privacy_copy.set_text(provider.privacy)
 
+    def _refresh_voxtype_details(self, *_):
+        self._voxtype_request_id += 1
+        request_id = self._voxtype_request_id
+        self.voxtype_refresh.set_sensitive(False)
+        self.voxtype_summary.set_text(tr(
+            "Reading Voxtype status…", "正在读取 Voxtype 状态…"))
+
+        def read_details():
+            try:
+                details = self._voxtype_details()
+                if details is None:
+                    raise RuntimeError("Voxtype details are unavailable")
+                error = None
+            except (OSError, RuntimeError, subprocess.SubprocessError) as exc:
+                details, error = None, str(exc)
+            GLib.idle_add(self._show_voxtype_details, request_id, details, error)
+
+        threading.Thread(target=read_details, name="voxtype-status",
+                         daemon=True).start()
+
+    def _show_voxtype_details(self, request_id, details, error):
+        if request_id != self._voxtype_request_id:
+            return GLib.SOURCE_REMOVE
+        self.voxtype_refresh.set_sensitive(True)
+        if error:
+            self.voxtype_summary.set_text(tr(
+                "Voxtype status unavailable: ",
+                "无法读取 Voxtype 状态：") + error)
+            return GLib.SOURCE_REMOVE
+        self.voxtype_summary.set_text("\n".join((
+            tr("CLI version: ", "CLI 版本：") + details.cli_version,
+            tr("Daemon version: ", "Daemon 版本：") + details.daemon_version,
+            tr("State: ", "状态：") + details.state,
+            tr("Engine: ", "引擎：") + details.engine,
+            tr("Model: ", "模型：") + details.model,
+            tr("Audio device: ", "音频设备：") + details.device,
+            tr("Compute backend: ", "计算后端：") + details.backend,
+            tr("Config schema: ", "配置 Schema：") + str(details.schema_version),
+            tr("Config file: ", "配置文件：") + details.config_path,
+        )))
+        return GLib.SOURCE_REMOVE
+
+    def _open_voxtype_configuration(self, *_):
+        try:
+            self._configure_voxtype()
+        except (OSError, ValueError) as error:
+            self.voxtype_summary.set_text(str(error))
+            return
+        self.voxtype_summary.set_text(tr(
+            "Voxtype configuration opened in a terminal.",
+            "已在终端中打开 Voxtype 配置。"))
+
     def _queue_asr_key_save(self, *_):
         if self._updating or not self.asr_key.get_text().strip():
             return

src/doubao_input/voxtype/control.py

@@ -0,0 +1,119 @@
+"""Read Voxtype's stable GUI-facing contracts and open its own configurator."""
+
+from __future__ import annotations
+
+from dataclasses import dataclass
+import json
+import os
+import shlex
+import shutil
+import subprocess
+
+from doubao_input.voxtype.runtime import VoxtypeRuntime, inspect_runtime
+
+
+def _run(command, *, timeout=3):
+    return subprocess.run(
+        command,
+        capture_output=True,
+        text=True,
+        timeout=timeout,
+        check=False,
+    )
+
+
+@dataclass(frozen=True)
+class VoxtypeDetails:
+    cli_version: str
+    daemon_version: str
+    state: str
+    engine: str
+    model: str
+    device: str
+    backend: str
+    schema_version: int | str
+    config_path: str
+
+
+def inspect_details(*, runtime=None, runner=_run) -> VoxtypeDetails:
+    runtime = runtime or inspect_runtime(runner=runner)
+    status = _json_command([
+        runtime.executable, "status", "--extended", "--format", "json",
+    ], runner)
+    state = status.get("alt") or status.get("class")
+    if state == "stopped":
+        raise RuntimeError("Voxtype daemon is not running")
+    schema = _json_command([
+        runtime.executable, "config", "schema", "--json",
+    ], runner)
+    return VoxtypeDetails(
+        cli_version=str(schema.get("voxtype_version") or runtime.version),
+        daemon_version=str(schema.get("daemon_version_label") or "unknown"),
+        state=str(state or "unknown"),
+        engine=str(schema.get("engine") or "unknown"),
+        model=str(status.get("model") or "unknown"),
+        device=str(status.get("device") or "system default"),
+        backend=str(status.get("backend") or "unknown"),
+        schema_version=schema.get("schema_version", "unknown"),
+        config_path=str(schema.get("config_path") or "unknown"),
+    )
+
+
+def _json_command(command, runner):
+    result = runner(command)
+    if result.returncode:
+        message = (result.stderr or result.stdout or "").strip()
+        raise RuntimeError(message[:500] or "Voxtype command failed")
+    try:
+        payload = json.loads(result.stdout)
+    except (TypeError, ValueError) as error:
+        raise RuntimeError("Voxtype returned invalid JSON") from error
+    if not isinstance(payload, dict):
+        raise RuntimeError("Voxtype returned invalid JSON")
+    return payload
+
+
+def launch_configure(*, runtime: VoxtypeRuntime | None = None,
+                     which=shutil.which, popen=subprocess.Popen,
+                     environment=None) -> None:
+    runtime = runtime or inspect_runtime()
+    if environment is None:
+        environment = os.environ
+    terminal = environment.get("TERMINAL", "").strip()
+    if terminal:
+        prefix = shlex.split(terminal)
+        executable = which(prefix[0]) if prefix else None
+        if executable:
+            _launch([
+                executable, *prefix[1:], "-e",
+                runtime.executable, "configure",
+            ], popen)
+            return
+    candidates = (
+        ("kitty", "-e"),
+        ("alacritty", "-e"),
+        ("foot", "-e"),
+        ("wezterm", "start", "--"),
+        ("gnome-terminal", "--"),
+        ("konsole", "-e"),
+        ("xfce4-terminal", "-e"),
+        ("xterm", "-e"),
+    )
+    for candidate in candidates:
+        executable = which(candidate[0])
+        if executable:
+            _launch([
+                executable, *candidate[1:], runtime.executable, "configure",
+            ], popen)
+            return
+    raise RuntimeError("No supported terminal was found for Voxtype configure")
+
+
+def _launch(command, popen):
+    popen(
+        command,
+        stdin=subprocess.DEVNULL,
+        stdout=subprocess.DEVNULL,
+        stderr=subprocess.DEVNULL,
+        start_new_session=True,
+    )

tests/contracts/test_pages_report_history.mjs

@@ -124,9 +124,9 @@ test('assigns every product case to exactly one balanced shard', async () => {
       caseCount += 1;
     }
   }
-  assert.equal(caseCount, 16);
+  assert.equal(caseCount, 17);
   assert.deepEqual(Object.fromEntries(shardCounts), {
-    'shard-1': 2,
+    'shard-1': 3,
     'shard-2': 3,
     'shard-3': 7,
     'shard-4': 4,

tests/e2e/cases/onboarding-regressions.yaml

@@ -213,6 +213,16 @@ cases:
       - aiTap: Finish & check result button in the Voice test content panel
       - wait: { duration: 1250 }
       - aiAssert: The voice test shows Synthetic Voxtype English transcript, the Listening overlay is gone, and feedback says the synthetic Voxtype CLI completed with transcript files removed and nothing pasted or sent.
+  - name: Inspect Voxtype configuration in Settings
+    tags: [shard-1, provider, voxtype-settings]
+    steps:
+      - fixture.prepare: { mode: voxtype-settings }
+      - aiAssert: The Doubao Say Settings window is open with the Voxtype status and action buttons visible.
+      - aiAssert: The Voxtype status shows CLI version 1.0.1, daemon version 1.0.1, idle state, whisper engine, synthetic-english-model, CI-only synthetic microphone, CPU backend, and config schema 1. The API Key field is hidden.
+      - aiTap: Refresh Voxtype status button in the Settings window
+      - aiAssert: The Voxtype status still shows idle and synthetic-english-model, and Open Voxtype configuration is available.
+      - aiTap: Open Voxtype configuration button in the Settings window
+      - aiAssert: The Settings window says Voxtype configuration opened in a terminal.
   - name: Changing microphone requires another check
     tags: [shard-2]
     steps:

tests/e2e/gtk_fixture.py

@@ -36,7 +36,7 @@ def main():
     Gtk.init()
     set_language("en")
     mode = fixture_mode()
-    if mode == "voxtype-live":
+    if mode in {"voxtype-live", "voxtype-settings"}:
         fake_cli = Path(__file__).parent / "fakes"
         os.environ["PATH"] = str(fake_cli) + os.pathsep + os.environ["PATH"]
     cleanup = (