Merge pull request #169 from harmony-one/fix-tap-to-speak

Fix the logic when a user resets the speech recognition
harmony-one · Nov 13, 2023 · 5ad9ca0 · 5ad9ca0
2 parents 9932bcb + e15b9d5
commit 5ad9ca0
Show file tree

Hide file tree

Showing 3 changed files with 18 additions and 8 deletions.
diff --git a/voice/voice-ai/x/Actions/ActionHandler.swift b/voice/voice-ai/x/Actions/ActionHandler.swift
@@ -65,13 +65,16 @@ class ActionHandler: ObservableObject {
         let resetTapSpeakState = isTapToSpeakActive && !isItTapToSpeakAction
         if resetTapSpeakState {
             isTapToSpeakActive = false
-            speechRecognition.cancelSpeak()
+            lastRecordingStateChangeTime = 0
+            stopRecording(cancel: true)
         }
     }
 
     func handle(actionType: ActionType) {
         syncTapToSpeakState(actionType)
 
+        print("Run action \(actionType)")
+
         switch actionType {
         case .reset:
             resetThrottler.send()
@@ -126,11 +129,11 @@ class ActionHandler: ObservableObject {
         SpeechRecognition.shared.speak()
     }
 
-    func stopRecording() {
+    func stopRecording(cancel: Bool = false) {
         if isRecording {
             isRecording = false
             print("Stopped Recording")
-            speechRecognition.stopSpeak()
+            speechRecognition.stopSpeak(cancel: cancel)
             // Simulating delay before triggering a synthesizing state change
             DispatchQueue.main.asyncAfter(deadline: .now() + 0.2) {
                 self.isSynthesizing = true

diff --git a/voice/voice-ai/x/Actions/ActionsView.swift b/voice/voice-ai/x/Actions/ActionsView.swift
@@ -224,8 +224,10 @@ struct ActionsView: View {
                 GridButton(currentTheme: currentTheme, button: button, foregroundColor: .black, active: isSpeakButtonPressed, isPressed: isPressed) {}.simultaneousGesture(
                     DragGesture(minimumDistance: 0)
                         .onChanged { _ in
+                            if (self.isSpeakButtonPressed == false) {
+                                actionHandler.handle(actionType: ActionType.speak)
+                            }
                             self.isSpeakButtonPressed = true
-                            actionHandler.handle(actionType: ActionType.speak)
                         }
                         .onEnded { _ in
                             self.isSpeakButtonPressed = false

diff --git a/voice/voice-ai/x/SpeechRecognition/SpeechRecognition.swift b/voice/voice-ai/x/SpeechRecognition/SpeechRecognition.swift
@@ -12,7 +12,7 @@ protocol SpeechRecognitionProtocol {
     func pause(feedback: Bool?)
     func repeate()
     func speak()
-    func stopSpeak()
+    func stopSpeak(cancel: Bool?)
     func sayMore()
     func cancelSpeak()
 }
@@ -25,6 +25,10 @@ extension SpeechRecognitionProtocol {
     func pause() {
         pause(feedback: true)
     }
+
+    func stopSpeak() {
+        stopSpeak(cancel: false)
+    }
 }
 
 class SpeechRecognition: NSObject, ObservableObject, SpeechRecognitionProtocol {
@@ -400,7 +404,7 @@ class SpeechRecognition: NSObject, ObservableObject, SpeechRecognitionProtocol {
         handleQuery(retryCount: maxRetry)
     }
 
-    func pauseCapturing(cancel: Bool = false) {
+    func pauseCapturing(cancel: Bool? = false) {
         guard isCapturing else {
             return
         }
@@ -418,6 +422,7 @@ class SpeechRecognition: NSObject, ObservableObject, SpeechRecognitionProtocol {
         recognitionRequest?.endAudio()
         audioEngine.inputNode.removeTap(onBus: 0)
         audioEngine.stop()
+        print("stop")
     }
 
     func resumeCapturing() {
@@ -627,7 +632,7 @@ class SpeechRecognition: NSObject, ObservableObject, SpeechRecognitionProtocol {
     }
 
     // Refactored 'stopSpeak' function to finalize speech recognition
-    func stopSpeak() {
+    func stopSpeak(cancel: Bool? = false) {
         DispatchQueue.global(qos: .userInitiated).async {
             print("[SpeechRecognition][stopSpeak]")
 
@@ -654,7 +659,7 @@ class SpeechRecognition: NSObject, ObservableObject, SpeechRecognitionProtocol {
 
             // Pause capturing after handling the recognized text
             DispatchQueue.main.async {
-                self.pauseCapturing()
+                self.pauseCapturing(cancel: cancel)
             }
         }
     }