fix(zerogpu): lower @spaces.GPU duration from 300s to 120s to stay under PRO cap

app.py

@@ -139,7 +139,7 @@ def retrieve_context(source_code: str, n_results: int = 3) -> str:
 # --- Inference ---
 
 
-@spaces.GPU(duration=300)
+@spaces.GPU(duration=120)
 def generate_spec(
     source_code: str,
     temperature: float = 0.7,