From cd78b0fb93bd7e43f7fbfa79c10e862daeb54412 Mon Sep 17 00:00:00 2001
From: "info@pi-farm.de" <pi-farm@noreply.git.pi-farm.de>
Date: Wed, 11 Mar 2026 12:13:05 +0000
Subject: [PATCH] app/src/main/java/com/example/jarvis_stts/MainActivity.kt
 aktualisiert

---
 .../com/example/jarvis_stts/MainActivity.kt   | 250 ++++++------------
 1 file changed, 77 insertions(+), 173 deletions(-)

diff --git a/app/src/main/java/com/example/jarvis_stts/MainActivity.kt b/app/src/main/java/com/example/jarvis_stts/MainActivity.kt
index 963b514..486e71a 100644
--- a/app/src/main/java/com/example/jarvis_stts/MainActivity.kt
+++ b/app/src/main/java/com/example/jarvis_stts/MainActivity.kt
@@ -15,42 +15,38 @@ import androidx.appcompat.app.AppCompatActivity
 import androidx.core.app.ActivityCompat
 import androidx.core.content.ContextCompat
 import okhttp3.*
-import java.util.Locale
-//import ai.picovoice.porcupine.PorcupineManager
-//import ai.picovoice.porcupine.PorcupineManagerCallback
-//import ai.picovoice.porcupine.Porcupine
 import org.vosk.Model
 import org.vosk.Recognizer
+import org.vosk.android.RecognitionListener
 import org.vosk.android.SpeechService
 import org.vosk.android.StorageService
-import org.vosk.android.RecognitionLi
+import java.io.IOException
+import java.util.Locale
 
-class MainActivity : AppCompatActivity(), RecognitionListener {
+class MainActivity : AppCompatActivity(), RecognitionListener, TextToSpeech.OnInitListener {
 
+    // UI Elemente
+    private lateinit var tvStatus: TextView
+    private lateinit var etUrl: EditText
+    private lateinit var spinnerVoices: Spinner
+    private lateinit var tts: TextToSpeech
+
+    // Vosk & Netzwerk
     private var voskService: SpeechService? = null
     private var voskModel: Model? = null
-
-    override fun onCreate(savedInstanceState: Bundle?) {
-        super.onCreate(savedInstanceState)
-        initVosk()
-    }
-
     private val client = OkHttpClient()
     private var webSocket: WebSocket? = null
 
-    // Liste für die Stimmen-Objekte und deren Namen
+    // TTS Stimmen
     private var availableVoices = mutableListOf<Voice>()
     private var voiceNames = mutableListOf<String>()
 
+    // Launcher für Google Spracherkennung
     private val speechRecognizerLauncher = registerForActivityResult(
         ActivityResultContracts.StartActivityForResult()
     ) { result ->
-        // Wenn Google fertig ist, starten wir das Wake-word wieder
-        try {
-            porcupineManager?.start()
-        } catch (e: Exception) {
-            Log.e("JARVIS", "Neustart nach Spracheingabe fehlgeschlagen: ${e.message}")
-        }
+        // Nach der Google-Eingabe starten wir Vosk wieder
+        startVosk() 
 
         if (result.resultCode == RESULT_OK && result.data != null) {
             val spokenText = result.data!!.getStringArrayListExtra(RecognizerIntent.EXTRA_RESULTS)?.get(0) ?: ""
@@ -59,7 +55,42 @@ class MainActivity : AppCompatActivity(), RecognitionListener {
         }
     }
 
-    private fun initVosk() {
+    override fun onCreate(savedInstanceState: Bundle?) {
+        super.onCreate(savedInstanceState)
+        setContentView(R.layout.activity_main)
+
+        // 1. UI initialisieren
+        tvStatus = findViewById(R.id.tvStatus)
+        etUrl = findViewById(R.id.etUrl)
+        spinnerVoices = findViewById(R.id.spinnerVoices)
+        val btnConnect = findViewById<Button>(R.id.btnConnect)
+        val btnSpeak = findViewById<Button>(R.id.btnSpeak)
+
+        tts = TextToSpeech(this, this)
+
+        // 2. SharedPreferences
+        val prefs = getSharedPreferences("JarvisPrefs", MODE_PRIVATE)
+        etUrl.setText(prefs.getString("server_url", ""))
+
+        btnConnect.setOnClickListener {
+            val url = etUrl.text.toString()
+            if (url.isNotEmpty()) {
+                prefs.edit().putString("server_url", url).apply()
+                connectToServer(url)
+            }
+        }
+
+        btnSpeak.setOnClickListener { startVoiceInput() }
+
+        // 3. Vosk Modell laden & Berechtigungen
+        if (ContextCompat.checkSelfPermission(this, Manifest.permission.RECORD_AUDIO) != PackageManager.PERMISSION_GRANTED) {
+            ActivityCompat.requestPermissions(this, arrayOf(Manifest.permission.RECORD_AUDIO), 1)
+        } else {
+            initVoskModel()
+        }
+    }
+
+    private fun initVoskModel() {
         StorageService.unpack(this, "model-de", "model",
             { model: Model ->
                 voskModel = model
@@ -73,125 +104,51 @@ class MainActivity : AppCompatActivity(), RecognitionListener {
 
     private fun startVosk() {
         try {
-            // "jarvis" als Hotword im Recognizer definieren
+            if (voskModel == null) return
+            // Wir horchen auf "jarvis". [unk] lässt unbekannte Wörter zu.
             val rec = Recognizer(voskModel, 16000.0f, "[\"jarvis\", \"[unk]\"]")
             voskService = SpeechService(rec, 16000.0f)
             voskService?.startListening(this)
-            tvStatus.text = "Bereit (Vosk)"
+            runOnUiThread { tvStatus.text = "Bereit (Warte auf 'Jarvis')" }
         } catch (e: Exception) {
             Log.e("JARVIS", "Vosk Start Fehler: ${e.message}")
         }
     }
 
-    // --- RecognitionListener Methoden ---
-
+    // --- Vosk RecognitionListener ---
     override fun onPartialResult(hypothesis: String) {
-        // Hier prüfen wir, ob "jarvis" im Teil-Ergebnis steht
-        if (hypothesis.contains("jarvis")) {
-            voskService?.stop() // Kurz Pause machen
-            startSpeechRecognition() // Dein Google STT starten
+        // Vosk liefert JSON, z.B. {"partial" : "jarvis"}
+        if (hypothesis.contains("jarvis", ignoreCase = true)) {
+            voskService?.stop() // Mikrofon für Google frei machen
+            startVoiceInput()
         }
     }
 
-    override fun onResult(hypothesis: String) { }
-    override fun onFinalResult(hypothesis: String) { }
-    override fun onError(exception: Exception) { }
-    override fun onTimeout() { }
+    override fun onResult(hypothesis: String) {}
+    override fun onFinalResult(hypothesis: String) {}
+    override fun onError(e: Exception) { Log.e("JARVIS", "Vosk Error: ${e.message}") }
+    override fun onTimeout() {}
 
-
-    override fun onCreate(savedInstanceState: Bundle?) {
-        super.onCreate(savedInstanceState)
-        setContentView(R.layout.activity_main)
-
-        // 1. UI-Elemente initialisieren
-        tvStatus = findViewById(R.id.tvStatus)
-        etUrl = findViewById(R.id.etUrl)
-        spinnerVoices = findViewById(R.id.spinnerVoices)
-
-        // Hier fehlten wahrscheinlich diese Definitionen:
-        val btnConnect = findViewById<Button>(R.id.btnConnect)
-        val btnSpeak = findViewById<Button>(R.id.btnSpeak)
-
-        tts = TextToSpeech(this, this)
-
-        // 2. URL aus dem Speicher laden (SharedPreferences)
-        val prefs = getSharedPreferences("JarvisPrefs", MODE_PRIVATE)
-        val savedUrl = prefs.getString("server_url", "")
-        if (!savedUrl.isNullOrEmpty()) {
-            etUrl.setText(savedUrl)
+    // --- Google STT & TTS ---
+    private fun startVoiceInput() {
+        val intent = Intent(RecognizerIntent.ACTION_RECOGNIZE_SPEECH).apply {
+            putExtra(RecognizerIntent.EXTRA_LANGUAGE_MODEL, RecognizerIntent.LANGUAGE_MODEL_FREE_FORM)
+            putExtra(RecognizerIntent.EXTRA_LANGUAGE, "de-DE")
         }
-
-        // 3. Click-Listener für den Verbinden-Button
-        btnConnect.setOnClickListener {
-            val url = etUrl.text.toString()
-            if (url.isNotEmpty()) {
-                // URL speichern
-                prefs.edit().putString("server_url", url).apply()
-                connectToServer(url)
-            } else {
-                Toast.makeText(this, "Bitte URL eingeben!", Toast.LENGTH_SHORT).show()
-            }
-        }
-
-        // 4. Click-Listener für den Sprechen-Button
-        btnSpeak.setOnClickListener { startVoiceInput() }
-
-        // 5. Mikrofon-Berechtigung prüfen
-        if (ContextCompat.checkSelfPermission(this, Manifest.permission.RECORD_AUDIO) != PackageManager.PERMISSION_GRANTED) {
-            // App hat noch keine Erlaubnis -> Wir fragen den Nutzer
-            ActivityCompat.requestPermissions(this, arrayOf(Manifest.permission.RECORD_AUDIO), 1)
-        } else {
-            // App hat die Erlaubnis schon (z.B. beim zweiten App-Start) -> Direkt starten!
-            
-        }
-    }
-
-    override fun onRequestPermissionsResult(requestCode: Int, permissions: Array<out String>, grantResults: IntArray) {
-        super.onRequestPermissionsResult(requestCode, permissions, grantResults)
-        
-        if (requestCode == 1) {
-            if (grantResults.isNotEmpty() && grantResults[0] == PackageManager.PERMISSION_GRANTED) {
-                // Juhu, Mikrofon ist freigegeben! Jetzt darf Porcupine starten.
-                initWakeWord()
-            } else {
-                Toast.makeText(this, "Ohne Mikrofon-Erlaubnis funktioniert J.A.R.V.I.S. nicht!", Toast.LENGTH_LONG).show()
-            }
-        }
-    }
-
-    override fun onResume() {
-        super.onResume()
-        // Nur starten, wenn wir die Erlaubnis haben
-        if (ContextCompat.checkSelfPermission(this, Manifest.permission.RECORD_AUDIO) == PackageManager.PERMISSION_GRANTED) {
-            // Falls der Manager schon läuft, stoppen wir ihn kurz, um ihn sauber neu zu starten
-            porcupineManager?.stop()
-            initWakeWord()
-        }
-    }
-
-    override fun onPause() {
-        super.onPause()
-        // Mikrofon pausieren, wenn die App nicht im Fokus ist, um den Fehler zu vermeiden
-        porcupineManager?.stop()
+        speechRecognizerLauncher.launch(intent)
     }
 
     private fun connectToServer(url: String) {
         webSocket?.close(1000, "Neuverbindung")
         val request = Request.Builder().url(url).build()
-
         webSocket = client.newWebSocket(request, object : WebSocketListener() {
             override fun onOpen(webSocket: WebSocket, response: Response) {
                 runOnUiThread { tvStatus.text = "Verbunden!" }
             }
-
             override fun onMessage(webSocket: WebSocket, text: String) {
                 runOnUiThread { tvStatus.text = "J.A.R.V.I.S.: $text" }
                 speakOut(text)
             }
-
-            override fun onFailure(webSocket: WebSocket, t: Throwable, response: Response?) {
-                runOnUiThread { tvStatus.text = "Fehler: ${t.message}" }
-            }
         })
     }
 
@@ -205,84 +162,31 @@ class MainActivity : AppCompatActivity(), RecognitionListener {
     private fun setupVoiceSpinner() {
         availableVoices.clear()
         voiceNames.clear()
-
-        // Alle verfügbaren deutschen Stimmen finden
-        val allVoices = tts.voices
-        for (v in allVoices) {
+        for (v in tts.voices) {
             if (v.locale.language == "de") {
                 availableVoices.add(v)
-                // Ein lesbarer Name für den Spinner
                 voiceNames.add("${v.name} (${if(v.isNetworkConnectionRequired) "Online" else "Lokal"})")
             }
         }
-
         val adapter = ArrayAdapter(this, android.R.layout.simple_spinner_item, voiceNames)
         adapter.setDropDownViewResource(android.R.layout.simple_spinner_dropdown_item)
-
-        runOnUiThread {
-            spinnerVoices.adapter = adapter
-            spinnerVoices.onItemSelectedListener = object : AdapterView.OnItemSelectedListener {
-                override fun onItemSelected(parent: AdapterView<*>?, view: View?, position: Int, id: Long) {
-                    tts.voice = availableVoices[position]
-                    Toast.makeText(this@MainActivity, "Stimme geändert", Toast.LENGTH_SHORT).show()
-                }
-                override fun onNothingSelected(parent: AdapterView<*>?) {}
+        spinnerVoices.adapter = adapter
+        spinnerVoices.onItemSelectedListener = object : AdapterView.OnItemSelectedListener {
+            override fun onItemSelected(p0: AdapterView<*>?, p1: View?, pos: Int, p3: Long) {
+                tts.voice = availableVoices[pos]
             }
+            override fun onNothingSelected(p0: AdapterView<*>?) {}
         }
     }
 
-    private fun startVoiceInput() {
-        val intent = Intent(RecognizerIntent.ACTION_RECOGNIZE_SPEECH).apply {
-            putExtra(RecognizerIntent.EXTRA_LANGUAGE_MODEL, RecognizerIntent.LANGUAGE_MODEL_FREE_FORM)
-            putExtra(RecognizerIntent.EXTRA_LANGUAGE, "de-DE")
-        }
-        speechRecognizerLauncher.launch(intent)
-    }
-
     private fun speakOut(text: String) {
         tts.speak(text, TextToSpeech.QUEUE_FLUSH, null, "")
     }
 
-    private fun initWakeWord() {
-        try {
-            val accessKey = "DEIN_KEY_HIER" // Bitte nochmal prüfen!
-            
-            porcupineManager = PorcupineManager.Builder()
-                .setAccessKey(accessKey)
-                .setKeyword(Porcupine.BuiltInKeyword.JARVIS)
-                .build(applicationContext, PorcupineManagerCallback { keywordIndex ->
-                    runOnUiThread {
-                        // 1. Wake-word stoppen, damit Google das Mikrofon kriegt
-                        porcupineManager?.stop() 
-                        // 2. Google-Eingabe starten
-                        startVoiceInput() 
-                    }
-                })
-                
-            porcupineManager?.start()
-            Log.d("JARVIS", "Porcupine erfolgreich gestartet")
-
-        } catch (e: Exception) {
-            val errorMessage = when {
-                // Wir prüfen den Klassennamen als String, falls die Klasse beim Kompilieren nicht gefunden wird
-                e.javaClass.simpleName.contains("PorcupineActivationNetworkException") -> "Keine Internetverbindung zur Key-Prüfung"
-                e.javaClass.simpleName.contains("PorcupineActivationException") -> "Key ungültig oder Limit erreicht"
-                else -> e.message ?: "Initialisierungsfehler"
-            }
-            
-            Log.e("JARVIS", "Detail-Fehler: $errorMessage")
-            runOnUiThread {
-                tvStatus.text = "Fehler: $errorMessage"
-            }
-        }
-    }
-
     override fun onDestroy() {
-        // Wichtig: Beim Schließen der App aufräumen!
-        porcupineManager?.stop()
-        porcupineManager?.delete()
+        voskService?.stop()
+        voskService?.shutdown()
         webSocket?.close(1000, "App Ende")
-        tts.stop()
         tts.shutdown()
         super.onDestroy()
     }