Improve app.py: unified text input experience and performance optimizations

- Unify text and voice input processing to use conversational generation logic
- Optimize flash-attn compilation with dynamic CPU core allocation using min(4, cpu_count())
- Update performance notice to reflect realistic generation time (6-8 minutes)
- Add multiprocessing import for CPU count detection

app.py

@@ -4,6 +4,7 @@ MiloMusic - Hugging Face Spaces Version
 AI-powered music generation platform optimized for cloud deployment with high-performance configuration.
 """
 
+import multiprocessing
 import os
 import sys
 import subprocess
@@ -25,7 +26,7 @@ def setup_spaces_environment():
     os.environ["TRANSFORMERS_CACHE"] = "/tmp/transformers_cache"
     os.environ["HF_HUB_CACHE"] = "/tmp/hf_hub_cache"
     
-    # PyTorch CUDA memory optimization
+    # 1.PyTorch CUDA memory optimization 2.用PyTorch的可扩展内存段分配, 提高GPU内存使用效率, 减少内存碎片问题
     os.environ["PYTORCH_CUDA_ALLOC_CONF"] = "expandable_segments:True"
     
     # Set temp directory for audio files
@@ -34,7 +35,7 @@ def setup_spaces_environment():
     print("🚀 Environment setup complete for Spaces")
 
 # Install flash-attn if not already installed
-def install_flash_attn():
+def install_flash_attn() -> bool:
     """Install flash-attn from source with proper compilation flags"""
     try:
         import flash_attn
@@ -54,7 +55,8 @@ def install_flash_attn():
             
             # Use more parallel jobs for faster compilation in Spaces
             env = os.environ.copy()
-            env["MAX_JOBS"] = "4"  # Utilize more CPU cores
+            max_jobs = min(4, multiprocessing.cpu_count())  # Utilize more CPU cores
+            env["MAX_JOBS"] = str(max_jobs)
             env["NVCC_PREPEND_FLAGS"] = "-ccbin /usr/bin/gcc"
             
             result = subprocess.run(cmd, env=env, capture_output=True, text=True, timeout=1800)  # 30 min timeout
@@ -527,46 +529,28 @@ def generate_music_spaces(lyrics: str, genre: str, mood: str, progress=gr.Progre
         return f"Error during music generation: {str(e)}"
 
 def respond(message, state):
-    """Enhanced response function for lyrics generation"""
+    """Enhanced response function for conversational lyrics generation"""
     try:
         # Add user message to conversation
         state.conversation.append({"role": "user", "content": message})
         
-        # Generate response using your existing lyrics generation logic
-        song_structure = generate_structured_lyrics(
-            state.conversation,
-            state.genre, 
-            state.mood, 
-            state.theme
-        )
-        
-        # Format the structured lyrics for display
-        response = format_lyrics(song_structure)
+        # Use conversational generation logic (same as voice input)
+        response = generate_chat_completion(groq_client, state.conversation, state.genre, state.mood, state.theme)
         
         # Add assistant response
         state.conversation.append({"role": "assistant", "content": response})
         
-        # Update lyrics if this looks like final lyrics
-        if any(marker in response.lower() for marker in ["[verse", "[chorus", "[bridge"]):
+        # Update lyrics with improved format recognition
+        if any(marker in response.lower() for marker in ["[verse", "[chorus", "[bridge", "**verse", "**chorus", "sectiontype.verse", "verse:"]):
             state.lyrics = response
         
-        # Return conversation for display in messages format
-        conversation_display = []
-        for msg in state.conversation:
-            conversation_display.append({"role": msg["role"], "content": msg["content"]})
-        
-        return "", conversation_display, state
+        # Format conversation for display
+        return "", [{"role": msg["role"], "content": msg["content"]} for msg in state.conversation], state
         
     except Exception as e:
         error_response = f"Sorry, I encountered an error: {str(e)}"
         state.conversation.append({"role": "assistant", "content": error_response})
-        
-        # Format conversation for display in messages format
-        conversation_display = []
-        for msg in state.conversation:
-            conversation_display.append({"role": msg["role"], "content": msg["content"]})
-        
-        return "", conversation_display, state
+        return "", [{"role": msg["role"], "content": msg["content"]} for msg in state.conversation], state
 
 def build_interface():
     """Build the Gradio interface optimized for Spaces with high performance"""
@@ -591,7 +575,7 @@ def build_interface():
         gr.Markdown("""
         <div class="performance-notice">
         🚀 <strong>High-Performance Mode:</strong> Running on Spaces GPU with optimized settings for best quality.
-        Generation time: ~3-5 minutes for professional-grade music with vocals and instruments.
+        Generation time: ~6-8 minutes for professional-grade music with vocals and instruments.
         </div>
         """)