Update app.py

app.py

@@ -104,13 +104,18 @@ def generate_ai_response(user_input):
     if not user_input or len(user_input.strip()) < 2:
         return "I'm listening. Please say more."
         
-    # Generate response
-    raw_response = llm(user_input)[0]["generated_text"]
-    
-    # Process to get clean, short response
-    final_response = process_response(user_input, raw_response)
-    
-    return final_response
+    try:
+        # Generate response
+        raw_response = llm(user_input)[0]["generated_text"]
+        
+        # Process to get clean, short response
+        final_response = process_response(user_input, raw_response)
+        
+        return final_response
+    except Exception as e:
+        print(f"Error generating AI response: {str(e)}")
+        # Return a default response if anything goes wrong
+        return "I heard you, but I'm having trouble forming a response right now."
 
 @app.route("/talk", methods=["POST"])
 def talk():
@@ -119,38 +124,73 @@ def talk():
         
     # Save audio
     audio_file = request.files["audio"]
-    with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as tmp:
-        audio_path = tmp.name
-        audio_file.save(audio_path)
     
     try:
-        # Transcribe
-        segments, _ = whisper_model.transcribe(audio_path)
-        transcription = "".join([seg.text for seg in segments])
-        
-        print(f"Transcription: {transcription}")  # Debugging
+        with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as tmp:
+            audio_path = tmp.name
+            audio_file.save(audio_path)
         
-        if not transcription.strip():
-            # Handle empty transcription
-            final_response = "I didn't catch that. Could you please speak again?"
-        else:
-            # Use the centralized function to generate a response
-            final_response = generate_ai_response(transcription)
+        # Transcribe
+        try:
+            segments, _ = whisper_model.transcribe(audio_path)
+            transcription = "".join([seg.text for seg in segments])
             
-        print(f"Voice response: {final_response}")  # Debugging
+            print(f"Transcription: {transcription}")  # Debugging
+            
+            if not transcription.strip():
+                # Handle empty transcription
+                final_response = "I didn't catch that. Could you please speak again?"
+            else:
+                # Use the centralized function to generate a response
+                final_response = generate_ai_response(transcription)
+                
+            print(f"Voice response: {final_response}")  # Debugging
+        except Exception as e:
+            print(f"Transcription error: {str(e)}")
+            final_response = "I had trouble understanding that. Could you try again?"
         
-        # Synthesize speech
+        # Prepare TTS output path
         tts_audio_path = audio_path.replace(".wav", "_reply.wav")
-        tts.tts_to_file(text=final_response, file_path=tts_audio_path)
         
-        # Return both the audio file and the text response
-        response = send_file(tts_audio_path, mimetype="audio/wav")
-        response.headers["X-Response-Text"] = final_response
+        try:
+            # Synthesize speech
+            tts.tts_to_file(text=final_response, file_path=tts_audio_path)
+            
+            if not os.path.exists(tts_audio_path) or os.path.getsize(tts_audio_path) == 0:
+                raise Exception("TTS failed to generate audio file")
+                
+        except Exception as e:
+            print(f"TTS error: {str(e)}")
+            # If TTS fails, generate a simple audio file with a message
+            # In a production app, you might want to have a pre-recorded fallback audio
+            tts_audio_path = audio_path  # Just reuse the input path for now
+            final_response = "Sorry, I couldn't generate audio right now."
         
-        return response
+        # Return both the audio file and the text response
+        try:
+            response = send_file(tts_audio_path, mimetype="audio/wav")
+            response.headers["X-Response-Text"] = final_response
+            response.headers["Access-Control-Expose-Headers"] = "X-Response-Text"
+            return response
+        except Exception as e:
+            print(f"Error sending file: {str(e)}")
+            return jsonify({
+                "error": "Could not send audio response",
+                "text_response": final_response
+            }), 500
+            
     except Exception as e:
         print(f"Error in talk endpoint: {str(e)}")
         return jsonify({"error": str(e)}), 500
+    finally:
+        # Clean up temporary files
+        try:
+            if 'audio_path' in locals() and os.path.exists(audio_path):
+                os.unlink(audio_path)
+            if 'tts_audio_path' in locals() and os.path.exists(tts_audio_path) and tts_audio_path != audio_path:
+                os.unlink(tts_audio_path)
+        except Exception as cleanup_error:
+            print(f"Error cleaning up files: {str(cleanup_error)}")
 
 @app.route("/chat", methods=["POST"])
 def chat():