Update app.py

app.py

@@ -1,54 +1,67 @@
+import os
+import torch
 import gradio as gr
 from transformers import pipeline
 
-import gradio as gr
-from transformers import pipeline
+# Tidy tokenizer threading + CPU thread count for smoother free-tier performance
+os.environ["TOKENIZERS_PARALLELISM"] = "false"
+try:
+    torch.set_num_threads(2)
+except Exception:
+    pass  # safe fallback if torch is missing threads control on some envs
 
-t2t = None
+# --- Model  ---
+GEN_MODEL_NAME = "MBZUAI/LaMini-Flan-T5-248M"  # text2text model that runs on CPU
+
+# Lazy-loaded text2text generator
+_t2t = None
 def get_t2t():
-    global t2t
-    if t2t is None:
-        t2t = pipeline(
+    global _t2t
+    if _t2t is None:
+        _t2t = pipeline(
             "text2text-generation",
-            model="google/flan-t5-small"   # instruction-tuned T5
+            model=GEN_MODEL_NAME,
+            tokenizer=GEN_MODEL_NAME,  # explicit to avoid tokenizer mismatch
+            device_map="cpu"           # ensure CPU-only on free tier
         )
-    return t2t
+    return _t2t
 
-def ai_fallback(prompt):
+def ai_fallback(prompt: str) -> str:
     try:
         gen = get_t2t()
-        # FLAN-T5 follows instructions well; keep outputs short & tame
+        # Keep answers short & deterministic (good for CPU + chatbot UX)
         out = gen(
-            f"Answer briefly: {prompt}",
+            f"Answer briefly and concisely: {prompt}",
             max_new_tokens=48,
-            do_sample=False   # deterministic, less rambling
+            do_sample=False
         )[0]["generated_text"]
-        return out.strip()
+        return (out or "").strip()
     except Exception:
         return "AI fallback had an issue. Please try a simpler question or use the topics in 'help'."
 
-
 HELP_TEXT = (
     "Ask me about: Bastet • cat mummies • daily life • worship\n"
     "Type anything else to try the AI fallback."
 )
 
-
-
 def reply(message, history):
     msg = (message or "").strip().lower()
     if msg in {"hi", "hello", "hey"} or "help" in msg:
         return "Hi! I share facts about cats in ancient Egypt.\n\n" + HELP_TEXT
     if "bastet" in msg or "bast" in msg:
-        return ("Bastet (later cat-headed) … major cult center at Bubastis in the Nile Delta.")
+        return "Bastet (later cat-headed) … major cult center at Bubastis in the Nile Delta."
     if any(w in msg for w in ["mummy", "mummies", "mummified", "offering"]):
-        return ("Millions of animal mummies (cats common), esp. Late Period (664–332 BCE).")
+        return "Millions of animal mummies (cats common), esp. Late Period (664–332 BCE)."
     if any(w in msg for w in ["daily", "life", "pest", "mouse", "rat", "snake"]):
-        return ("Cats protected grain stores; art shows them under chairs/on leashes with owners.")
+        return "Cats protected grain stores; art shows them under chairs/on leashes with owners."
     if any(w in msg for w in ["worship", "god", "goddess", "taboo"]):
-        return ("People didn’t worship pet cats as gods; they revered cats via Bastet and votive offerings.")
+        return "People didn’t worship pet cats as gods; they revered cats via Bastet and votive offerings."
     return ai_fallback(message)
 
-demo = gr.ChatInterface(fn=reply, title="Cats of Ancient Egypt — Hybrid (Rules + Tiny Transformer)")
+demo = gr.ChatInterface(
+    fn=reply,
+    title="Cats of Ancient Egypt — Hybrid (Rules + LaMini-Flan-T5-248M)"
+)
+
 if __name__ == "__main__":
     demo.launch()