Update app.py

app.py

@@ -1,24 +1,23 @@
 import os
 import gradio as gr
 
-# Optional: only if torch is installed (it is in your requirements)
+# Optional CPU hygiene
+os.environ["TOKENIZERS_PARALLELISM"] = "false"
 try:
     import torch
-    os.environ["TOKENIZERS_PARALLELISM"] = "false"
     try:
-        torch.set_num_threads(2)  # nicer on 2 vCPUs
+        torch.set_num_threads(2)  # nice on 2 vCPUs
     except Exception:
         pass
 except Exception:
-    # If torch import ever fails, don't crash the app
     pass
 
 from transformers import pipeline
 
-# ---------------- Model config ----------------
+# -------- Model config --------
 GEN_MODEL_NAME = "MBZUAI/LaMini-Flan-T5-248M"  # CPU-friendly text2text model
 
-# Lazy-load the generator once
+# Lazy-loaded generator
 _t2t = None
 def get_t2t():
     global _t2t
@@ -26,9 +25,9 @@ def get_t2t():
         _t2t = pipeline(
             "text2text-generation",
             model=GEN_MODEL_NAME,
-            tokenizer=GEN_MODEL_NAME,  # explicit to avoid mismatch
-            device_map="cpu"           # force CPU on free tier
+            tokenizer=GEN_MODEL_NAME
         )
+        print(f"[startup] Loaded model: {GEN_MODEL_NAME}")
     return _t2t
 
 def ai_fallback(prompt: str) -> str:
@@ -41,7 +40,7 @@ def ai_fallback(prompt: str) -> str:
         )[0]["generated_text"]
         return (out or "").strip()
     except Exception as e:
-        # Log real error to console/Space logs to help debug if anything else fails
+        # Print the real error to logs so you can see it in the console
         print("AI fallback error:", repr(e))
         return "AI fallback had an issue. Please try a simpler question or use the topics in 'help'."