Title: Cache ParakeetModule across transcription calls to avoid repeated model loading

Summary: Currently, runParakeetOnWavFile() creates a new ParakeetModule and calls .close()
on every transcription request (L235-251). This means the model is loaded from disk on
every single inference call, and the reported latency includes model load time.
This change caches the ParakeetModule at the class level and only
recreates it when the user changes model settings.

Author: Github Executorch

parakeet/android/ParakeetApp/app/src/main/java/com/example/parakeetapp/MainActivity.kt

@@ -82,6 +82,12 @@ class MainActivity : ComponentActivity() {
     private lateinit var viewModel: ModelSettingsViewModel
     private lateinit var downloadViewModel: ModelDownloadViewModel
 
+    // Cached ParakeetModule: loaded once and reused across transcription calls.
+    private var parakeetModule: ParakeetModule? = null
+    private var loadedModelPath: String = ""
+    private var loadedTokenizerPath: String = ""
+    private var loadedDataPath: String? = null
+
     enum class Screen {
         DOWNLOAD,
         MAIN,
@@ -211,6 +217,35 @@ class MainActivity : ComponentActivity() {
         }.start()
     }
 
+    /**
+     * Returns the cached ParakeetModule, creating or recreating it only if
+     * the model settings have changed since the last load.
+     */
+    private fun getOrCreateModule(settings: ModelSettings): ParakeetModule {
+        val dataPath = settings.dataPath.ifBlank { null }
+        if (parakeetModule != null &&
+            loadedModelPath == settings.modelPath &&
+            loadedTokenizerPath == settings.tokenizerPath &&
+            loadedDataPath == dataPath
+        ) {
+            return parakeetModule!!
+        }
+
+        // Settings changed or first load — close the old module and create a new one.
+        parakeetModule?.close()
+        Log.v(TAG, "Loading model: ${settings.modelPath}")
+        val module = ParakeetModule(
+            modelPath = settings.modelPath,
+            tokenizerPath = settings.tokenizerPath,
+            dataPath = dataPath
+        )
+        parakeetModule = module
+        loadedModelPath = settings.modelPath
+        loadedTokenizerPath = settings.tokenizerPath
+        loadedDataPath = dataPath
+        return module
+    }
+
     /**
      * Common method to run Parakeet on a WAV file path.
      */
@@ -232,24 +267,18 @@ class MainActivity : ComponentActivity() {
             buttonEnabled = false
         }
 
-        val parakeetModule = ParakeetModule(
-            modelPath = settings.modelPath,
-            tokenizerPath = settings.tokenizerPath,
-            dataPath = settings.dataPath.ifBlank { null }
-        )
+        val module = getOrCreateModule(settings)
 
         Log.v(TAG, "Starting transcribe for: $wavFilePath")
         runOnUiThread {
             statusText = "Transcribing..."
         }
         val startTime = System.currentTimeMillis()
-        val result = parakeetModule.transcribe(wavFilePath)
+        val result = module.transcribe(wavFilePath)
         val elapsedTime = System.currentTimeMillis() - startTime
         val elapsedSeconds = elapsedTime / 1000.0
         Log.v(TAG, "Finished transcribe in ${elapsedSeconds}s")
 
-        parakeetModule.close()
-
         runOnUiThread {
             transcriptionOutput = result
             statusText = "Transcription complete (%.2fs)".format(elapsedSeconds)
@@ -468,6 +497,8 @@ class MainActivity : ComponentActivity() {
         if (isRecording) {
             stopRecording()
         }
+        parakeetModule?.close()
+        parakeetModule = null
     }
 
     private val requestPermissionLauncher = registerForActivityResult(