fix: add llm to use instead match

app.py

@@ -11,6 +11,9 @@ from inference import proc_folder_direct
 from pathlib import Path
 import spaces
 from pydub.exceptions import CouldntEncodeError
+from transformers import pipeline
+
+model = pipeline('text-generation', model='EleutherAI/gpt-neo-125M')
 
 OUTPUT_FOLDER = "separation_results/"
 INPUT_FOLDER = "input"
@@ -25,23 +28,27 @@ def delete_input_files(input_dir):
         wav_file.unlink()
         print(f"Deleted {wav_file}")
 
+def analyze_filename_with_llm(filename):
+    prompt = f"Extract the artist and song title from the following filename and format it as 'Artist - Title':\n\n{filename}"
+    
+    # Generate a response using the local model
+    response = model(prompt, max_length=50, do_sample=False)[0]['generated_text']
+    
+    # Extract the first line of the response, which should be the "Artist - Title"
+    artist_title = response.strip().split('\n')[0]
+    
+    return artist_title
+
 def handle_file_upload(file):
     if file is None:
         return None, "No file uploaded"
 
     filename = os.path.basename(file.name)
+    
+    # Use LLM to analyze the filename and return the formatted title
+    formatted_title = analyze_filename_with_llm(filename)
 
-    # This regex captures both "Artist - Title" and "Title - Artist" formats
-    match = re.match(
-        r'^(.*? - .*?)\s*(?:[\(\[].*?[\)\]])?$', filename
-    )
-
-    if match:
-        # This directly captures the "Artist - Title" part
-        formatted_title = match.group(1).strip()
-    else:
-        # If no match, fallback to the original filename
-        formatted_title = sanitize_filename(filename.strip())
+    formatted_title = sanitize_filename(formatted_title.strip())
 
     input_path = os.path.join(INPUT_FOLDER, "wav", f"{formatted_title}.wav")
     os.makedirs(os.path.dirname(input_path), exist_ok=True)
@@ -51,6 +58,7 @@ def handle_file_upload(file):
 
     return input_path, formatted_title
 
+
 def run_inference(model_type, config_path, start_check_point, input_dir, output_dir, device_ids="0"):
     command = [
         "python", "inference.py",