Update app.py

app.py

@@ -4,6 +4,17 @@ from transformers import AutoTokenizer, AutoModelForCausalLM, BitsAndBytesConfig
 from threading import Thread
 from transformers import TextIteratorStreamer
 
+# ZeroGPU support for Hugging Face Spaces (effect-free if not in Spaces)
+try:
+    import spaces
+    GPU_DECORATOR = spaces.GPU(duration=30)
+    print("[SUCCESS] ZeroGPU support enabled")
+except ImportError:
+    # Fallback for local execution
+    def GPU_DECORATOR(func):
+        return func
+    print("[WARNING] Running without ZeroGPU (local mode)")
+
 # Language-specific system prompts (extracted from training set)
 LANGUAGE_SYSTEM_PROMPTS = {
     # Germanic languages
@@ -71,10 +82,15 @@ model = AutoModelForCausalLM.from_pretrained(
     device_map="auto",
 )
 
-print("Model loaded!")
+print("[SUCCESS] Model loaded!")
 
+@GPU_DECORATOR
 def chat(message, history, temperature, language):
-    """Generate response with configurable temperature and language."""
+    """Generate response with configurable temperature and language.
+    
+    This function is decorated with @spaces.GPU for dynamic GPU allocation
+    in Hugging Face Spaces (30 second duration). Effect-free locally.
+    """
     # Gradio ChatInterface passes history with content as string or list[dict]
     # Extract text content for apply_chat_template
     def extract_text(content):