Update app.py

app.py

@@ -7,14 +7,16 @@ import pretty_midi
 import subprocess
 import os
 from openai import OpenAI
-# Audio playback support
+# Audio support flag
 try:
     import pygame
     pygame.mixer.init()
-    PYGAME_AVAILABLE = True
+    AUDIO_AVAILABLE = True
 except Exception as e:
-    print(f"[WARNING] pygame mixer init failed: {e}")
-    PYGAME_AVAILABLE = False
+    print(f"[WARN] pygame audio init failed: {e}")
+    AUDIO_AVAILABLE = False
+
+midi_path_global = None
 
 # === LLM APIs ===
 def query_llm(prompt, model_name=None):
@@ -29,7 +31,7 @@ def query_llm(prompt, model_name=None):
     api_key=os.environ.get("NEBIUS_API_KEY")
 )
         response = client.chat.completions.create(
-    model="meta-llama/Llama-3.3-70B-Instruct",
+    model="Qwen/Qwen3-30B-A3B",
     max_tokens=512,
     temperature=0.6,
     top_p=0.9,
@@ -115,42 +117,46 @@ def midi_from_plan(melody, tempo):
     midi.instruments.append(instrument)
     return midi
 
-# === Generate audio preview from MIDI ===
-def midi_to_wav(midi_path):
-    try:
-        import tempfile
-        import subprocess
-        import os
-
-        # Convert MIDI to WAV using FluidSynth if installed, else fallback to empty
-        wav_path = tempfile.NamedTemporaryFile(delete=False, suffix=".wav").name
-
-        # Use fluidsynth if available, else skip audio preview
-        fluidsynth_cmd = ["fluidsynth", "-ni", "/usr/share/sounds/sf2/FluidR3_GM.sf2", midi_path, "-F", wav_path, "-r", "44100"]
-        result = subprocess.run(fluidsynth_cmd, capture_output=True)
-        if result.returncode != 0:
-            print("[WARNING] FluidSynth conversion failed or is not installed.")
-            return None
-        return wav_path
-    except Exception as e:
-        print(f"[ERROR] midi_to_wav failed: {e}")
-        return None
-
-# === Main function to generate MIDI and audio preview ===
-def generate_midi_and_audio(prompt, model_name):
+def generate_midi_from_prompt(prompt, model_name):
+    global midi_path_global
     intent = get_intent_from_prompt(prompt, model_name)
     melody = get_melody_from_intent(intent, model_name)
     midi = midi_from_plan(melody, intent.get("tempo", 120))
 
     with tempfile.NamedTemporaryFile(delete=False, suffix=".mid") as tmp:
         midi.write(tmp.name)
-        midi_path = tmp.name
-
-    audio_path = None
-    if PYGAME_AVAILABLE:
-        audio_path = midi_path  # We'll use pygame to play midi directly if possible
-
-    return midi_path, audio_path
+        midi_path_global = tmp.name
+    return midi_path_global
+
+def play_audio():
+    if not AUDIO_AVAILABLE:
+        return "Audio playback not available."
+    if midi_path_global is None:
+        return "No MIDI file loaded yet."
+    try:
+        pygame.mixer.music.load(midi_path_global)
+        pygame.mixer.music.play()
+        return "Playing audio..."
+    except Exception as e:
+        return f"Error playing audio: {e}"
+
+def pause_audio():
+    if not AUDIO_AVAILABLE:
+        return "Audio playback not available."
+    pygame.mixer.music.pause()
+    return "Paused."
+
+def unpause_audio():
+    if not AUDIO_AVAILABLE:
+        return "Audio playback not available."
+    pygame.mixer.music.unpause()
+    return "Unpaused."
+
+def stop_audio():
+    if not AUDIO_AVAILABLE:
+        return "Audio playback not available."
+    pygame.mixer.music.stop()
+    return "Stopped."
 
 # === Get Ollama models ===
 def get_ollama_models():
@@ -165,17 +171,28 @@ def get_ollama_models():
 models = get_ollama_models()
 
 demo = gr.Interface(
-    fn=generate_midi_and_audio,
+    fn=generate_midi_from_prompt,
     inputs=[
         gr.Textbox(label="Music Prompt"),
         gr.Dropdown(choices=models, label="LLM Model", value=models[0])
     ],
     outputs=[
-        gr.File(label="🎵 Download MIDI File"),
-        gr.Audio(label="🎧 Audio Preview (MIDI Playback, if supported)", type="filepath")
+        gr.File(label="🎵 Download MIDI File")
     ],
     title="🎼 Music Command Prompt (MCP Agent)",
     description="Describe your music idea and download a generated MIDI file. Choose from local or OpenAI LLMs."
 )
 
-demo.launch(mcp_server=True)
+with gr.Row():
+    with gr.Column(scale=1):
+        play_btn = gr.Button("▶ Play")
+        pause_btn = gr.Button("⏸ Pause")
+        unpause_btn = gr.Button("▶ Resume")
+        stop_btn = gr.Button("⏹ Stop")
+    with gr.Column(scale=3):
+        status = gr.Textbox(label="Audio Status", interactive=False)
+
+play_btn.click(play_audio, None, status)
+pause_btn.click(pause_audio, None, status)
+unpause_btn.click(unpause_audio, None, status)
+stop_btn.click(stop_audio, None, status)