Merge pull request #304 from celskeggs/cskeggs-patch-5

Author: Stypox

app/src/main/kotlin/org/stypox/dicio/di/SttInputDeviceWrapper.kt

@@ -9,6 +9,7 @@ import dagger.Provides
 import dagger.hilt.InstallIn
 import dagger.hilt.android.qualifiers.ApplicationContext
 import dagger.hilt.components.SingletonComponent
+import javax.inject.Singleton
 import kotlinx.coroutines.CoroutineScope
 import kotlinx.coroutines.Dispatchers
 import kotlinx.coroutines.Job
@@ -24,15 +25,15 @@ import org.stypox.dicio.io.input.SttState
 import org.stypox.dicio.io.input.external_popup.ExternalPopupInputDevice
 import org.stypox.dicio.io.input.vosk.VoskInputDevice
 import org.stypox.dicio.settings.datastore.InputDevice
-import org.stypox.dicio.settings.datastore.InputDevice.INPUT_DEVICE_NOTHING
 import org.stypox.dicio.settings.datastore.InputDevice.INPUT_DEVICE_EXTERNAL_POPUP
+import org.stypox.dicio.settings.datastore.InputDevice.INPUT_DEVICE_NOTHING
 import org.stypox.dicio.settings.datastore.InputDevice.INPUT_DEVICE_UNSET
 import org.stypox.dicio.settings.datastore.InputDevice.INPUT_DEVICE_VOSK
 import org.stypox.dicio.settings.datastore.InputDevice.UNRECOGNIZED
 import org.stypox.dicio.settings.datastore.SttPlaySound
 import org.stypox.dicio.settings.datastore.UserSettings
 import org.stypox.dicio.util.distinctUntilChangedBlockingFirst
-import javax.inject.Singleton
+import org.stypox.dicio.util.toStateFlowDistinctBlockingFirst
 
 
 interface SttInputDeviceWrapper {
@@ -58,6 +59,7 @@ class SttInputDeviceWrapperImpl(
 
     private var inputDeviceSetting: InputDevice
     private var sttPlaySoundSetting: SttPlaySound
+    private val silencesBeforeStop: StateFlow<Int>
     private var sttInputDevice: SttInputDevice?
 
     // null means that the user has not enabled any STT input device
@@ -75,6 +77,8 @@ class SttInputDeviceWrapperImpl(
 
         inputDeviceSetting = firstSettings.first
         sttPlaySoundSetting = firstSettings.second
+        silencesBeforeStop = dataStore.data.map(SttInputDevice::getSttSilenceDurationOrDefault)
+            .toStateFlowDistinctBlockingFirst(scope)
         sttInputDevice = buildInputDevice(inputDeviceSetting)
         scope.launch {
             restartUiStateJob()
@@ -102,7 +106,7 @@ class SttInputDeviceWrapperImpl(
         return when (setting) {
             UNRECOGNIZED,
             INPUT_DEVICE_UNSET,
-            INPUT_DEVICE_VOSK -> VoskInputDevice(appContext, okHttpClient, localeManager)
+            INPUT_DEVICE_VOSK -> VoskInputDevice(appContext, okHttpClient, localeManager, silencesBeforeStop)
             INPUT_DEVICE_EXTERNAL_POPUP ->
                 ExternalPopupInputDevice(appContext, activityForResultManager, localeManager)
             INPUT_DEVICE_NOTHING -> null

app/src/main/kotlin/org/stypox/dicio/io/input/SttInputDevice.kt

@@ -1,6 +1,7 @@
 package org.stypox.dicio.io.input
 
 import kotlinx.coroutines.flow.StateFlow
+import org.stypox.dicio.settings.datastore.UserSettings
 
 interface SttInputDevice {
     val uiState: StateFlow<SttState>
@@ -12,4 +13,12 @@ interface SttInputDevice {
     fun onClick(eventListener: (InputEvent) -> Unit)
 
     suspend fun destroy()
+
+    companion object {
+        const val DEFAULT_STT_SILENCE_DURATION = 2
+        fun getSttSilenceDurationOrDefault(settings: UserSettings): Int {
+            // unfortunately there is no way to tell protobuf to use "2" as the default value
+            return settings.sttSilenceDuration.takeIf { it > 0 } ?: DEFAULT_STT_SILENCE_DURATION
+        }
+    }
 }

app/src/main/kotlin/org/stypox/dicio/io/input/vosk/VoskInputDevice.kt

@@ -49,6 +49,7 @@ import org.stypox.dicio.io.input.vosk.VoskState.NotDownloaded
 import org.stypox.dicio.io.input.vosk.VoskState.NotInitialized
 import org.stypox.dicio.io.input.vosk.VoskState.NotLoaded
 import org.stypox.dicio.io.input.vosk.VoskState.Unzipping
+import org.stypox.dicio.settings.datastore.UserSettings
 import org.stypox.dicio.ui.util.Progress
 import org.stypox.dicio.util.FileToDownload
 import org.stypox.dicio.util.LocaleUtils
@@ -68,6 +69,7 @@ class VoskInputDevice(
     @ApplicationContext appContext: Context,
     private val okHttpClient: OkHttpClient,
     localeManager: LocaleManager,
+    private val silencesBeforeStop: StateFlow<Int>
 ) : SttInputDevice {
 
     private val _state: MutableStateFlow<VoskState>
@@ -84,7 +86,6 @@ class VoskInputDevice(
     private val modelDirectory: File get() = File(filesDir, "vosk-model")
     private val modelExistFileCheck: File get() = File(modelDirectory, "ivector")
 
-
     init {
         // Run blocking, because the locale is always available right away since LocaleManager also
         // initializes in a blocking way. Moreover, if VoskInputDevice were not initialized straight
@@ -420,7 +421,7 @@ class VoskInputDevice(
         eventListener: (InputEvent) -> Unit,
     ) {
         _state.value = Listening(speechService, eventListener)
-        speechService.startListening(VoskListener(this, eventListener, speechService))
+        speechService.startListening(VoskListener(this, eventListener, silencesBeforeStop.value, speechService))
     }
 
     /**

app/src/main/kotlin/org/stypox/dicio/io/input/vosk/VoskListener.kt

@@ -31,10 +31,14 @@ import org.vosk.android.SpeechService
  * This class is an implementation of [RecognitionListener] that listens to events generated by a
  * Vosk [SpeechService], pushes them to [eventListener], and handles the listening state via
  * [VoskInputDevice].
+ *
+ * @param silencesBeforeStop how many times Vosk should report a "silence" before giving up trying
+ * to get speech; must be `>= 1`, and setting this to 1 is equivalent to disabling this feature
  */
 internal class VoskListener(
     private val voskInputDevice: VoskInputDevice,
     private val eventListener: (InputEvent) -> Unit,
+    private var silencesBeforeStop: Int,
     private val speechService: SpeechService,
 ) : RecognitionListener {
 
@@ -67,10 +71,7 @@ internal class VoskListener(
      */
     @Suppress("ktlint:Style:NestedBlockDepth")
     override fun onResult(s: String) {
-        Log.d(TAG, "onResult called with s = $s")
-
-        // we only want to listen one sentence at a time
-        voskInputDevice.stopListening(speechService, eventListener, false)
+        Log.d(TAG, "onResult called with s = $s; silencesBeforeStop=$silencesBeforeStop")
 
         // collect all alternative user inputs generated by the STT model
         val inputs = try {
@@ -79,9 +80,18 @@ internal class VoskListener(
         } catch (e: JSONException) {
             Log.e(TAG, "Can't obtain result from $s", e)
             eventListener(InputEvent.Error(e))
+            voskInputDevice.stopListening(speechService, eventListener, false)
+            return
+        }
+
+        if (inputs.isEmpty() && silencesBeforeStop > 1) {
+            silencesBeforeStop -= 1
             return
         }
 
+        // we only want to listen one sentence at a time
+        voskInputDevice.stopListening(speechService, eventListener, false)
+
         // emit the final event
         if (inputs.isEmpty()) {
             eventListener(InputEvent.None)

app/src/main/kotlin/org/stypox/dicio/settings/Definitions.kt

@@ -9,6 +9,7 @@ import androidx.compose.material.icons.filled.Cloud
 import androidx.compose.material.icons.filled.ColorLens
 import androidx.compose.material.icons.filled.DarkMode
 import androidx.compose.material.icons.filled.Hearing
+import androidx.compose.material.icons.filled.HourglassEmpty
 import androidx.compose.material.icons.filled.InvertColors
 import androidx.compose.material.icons.filled.KeyboardAlt
 import androidx.compose.material.icons.filled.Language
@@ -30,6 +31,7 @@ import org.stypox.dicio.settings.datastore.SttPlaySound
 import org.stypox.dicio.settings.datastore.Theme
 import org.stypox.dicio.settings.datastore.WakeDevice
 import org.stypox.dicio.settings.ui.BooleanSetting
+import org.stypox.dicio.settings.ui.IntSetting
 import org.stypox.dicio.settings.ui.ListSetting
 
 
@@ -171,6 +173,15 @@ fun speechOutputDevice() = ListSetting(
     ),
 )
 
+@Composable
+fun sttSilenceDuration() = IntSetting(
+    title = stringResource(R.string.pref_stt_silence_duration_title),
+    icon = Icons.Default.HourglassEmpty,
+    description = @Composable { stringResource(R.string.pref_stt_silence_duration_description, it) },
+    minimum = 1,
+    maximum = 7,
+)
+
 @Composable
 fun sttAutoFinish() = BooleanSetting(
     title = stringResource(R.string.pref_stt_auto_finish_title),

app/src/main/kotlin/org/stypox/dicio/settings/MainSettingsScreen.kt

@@ -32,6 +32,7 @@ import androidx.compose.ui.tooling.preview.Preview
 import androidx.compose.ui.unit.dp
 import androidx.hilt.navigation.compose.hiltViewModel
 import org.stypox.dicio.R
+import org.stypox.dicio.io.input.SttInputDevice
 import org.stypox.dicio.settings.datastore.InputDevice
 import org.stypox.dicio.settings.datastore.Language
 import org.stypox.dicio.settings.datastore.SpeechOutputDevice
@@ -187,6 +188,12 @@ private fun MainSettingsScreen(
                 viewModel::setSttPlaySound
             )
         }
+        item {
+            sttSilenceDuration().Render(
+                SttInputDevice.getSttSilenceDurationOrDefault(settings),
+                viewModel::setSttSilenceDuration
+            )
+        }
         item {
             sttAutoFinish().Render(
                 settings.autoFinishSttPopup,

app/src/main/kotlin/org/stypox/dicio/settings/MainSettingsViewModel.kt

@@ -71,6 +71,8 @@ class MainSettingsViewModel @Inject constructor(
         updateData { it.setSpeechOutputDevice(value) }
     fun setSttPlaySound(value: SttPlaySound) =
         updateData { it.setSttPlaySound(value) }
+    fun setSttSilenceDuration(value: Int) =
+        updateData { it.setSttSilenceDuration(value) }
     fun setAutoFinishSttPopup(value: Boolean) =
         updateData { it.setAutoFinishSttPopup(value) }
 }

app/src/main/kotlin/org/stypox/dicio/settings/ui/Setting.kt

@@ -15,9 +15,11 @@ import androidx.compose.material.icons.filled.BookmarkAdded
 import androidx.compose.material.icons.filled.BookmarkRemove
 import androidx.compose.material.icons.filled.Pets
 import androidx.compose.material3.Card
+import androidx.compose.material3.ExperimentalMaterial3Api
 import androidx.compose.material3.Icon
 import androidx.compose.material3.MaterialTheme
 import androidx.compose.material3.RadioButton
+import androidx.compose.material3.Slider
 import androidx.compose.material3.Switch
 import androidx.compose.material3.Text
 import androidx.compose.material3.TextButton
@@ -44,6 +46,7 @@ import androidx.compose.ui.unit.dp
 import androidx.compose.ui.window.Dialog
 import kotlinx.coroutines.job
 import org.stypox.dicio.R
+import kotlin.math.roundToInt
 
 interface SettingWithValue<T> {
     @Composable
@@ -79,6 +82,38 @@ class BooleanSetting(
     }
 }
 
+class IntSetting(
+    private val title: String,
+    private val minimum: Int,
+    private val maximum: Int,
+    private val icon: ImageVector? = null,
+    private val description: (@Composable (Int) -> String)? = null,
+) : SettingWithValue<Int> {
+    init {
+        assert(maximum > minimum)
+    }
+
+    @Composable
+    override fun Render(
+        value: Int,
+        onValueChange: (Int) -> Unit,
+    ) {
+        SettingsItem(
+            title = title,
+            icon = icon,
+            description = description?.invoke(value),
+            innerContent = {
+                Slider(
+                    value = value.toFloat(),
+                    onValueChange = { onValueChange(it.roundToInt()) },
+                    steps = maximum - minimum - 1,
+                    valueRange = minimum.toFloat()..maximum.toFloat()
+                )
+            },
+        )
+    }
+}
+
 class ListSetting<T>(
     private val title: String,
     private val icon: ImageVector? = null,

app/src/main/kotlin/org/stypox/dicio/settings/ui/SettingsItem.kt

@@ -37,6 +37,7 @@ fun SettingsItem(
     icon: ImageVector? = null,
     description: String? = null,
     content: (@Composable () -> Unit)? = null,
+    innerContent: (@Composable () -> Unit)? = null,
 ) {
     Row(
         verticalAlignment = Alignment.CenterVertically,
@@ -70,6 +71,10 @@ fun SettingsItem(
                     style = MaterialTheme.typography.bodyMedium,
                 )
             }
+            if (innerContent != null) {
+                Spacer(modifier = Modifier.height(3.dp))
+                innerContent()
+            }
         }
         if (content != null) {
             Spacer(modifier = Modifier.width(16.dp))

app/src/main/proto/user_settings.proto

@@ -20,4 +20,5 @@ message UserSettings {
     map<string, bool> enabled_skills = 7;
     WakeDevice wake_device = 8;
     SttPlaySound stt_play_sound = 9;
+    int32 stt_silence_duration = 10;
 }