Update app.py

app.py

@@ -4,15 +4,16 @@ import json
 import torch
 from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline
 from ppt_parser import transfer_to_structure
+from functools import lru_cache
 
-# ✅ Hugging Face token for gated model access
+# ✅ Get Hugging Face token from Space Secrets
 hf_token = os.getenv("HF_TOKEN")
 
-# ✅ Load summarization pipeline
+# ✅ Load summarization model (BART)
 summarizer = pipeline("summarization", model="facebook/bart-large-cnn")
 
-# ✅ Load Mistral 7B Instruct model
-@gr.cache()
+# ✅ Load Mistral model (memoized to avoid reloading)
+@lru_cache(maxsize=1)
 def load_mistral():
     tokenizer = AutoTokenizer.from_pretrained("mistralai/Mistral-7B-Instruct-v0.1", token=hf_token)
     model = AutoModelForCausalLM.from_pretrained(
@@ -21,12 +22,11 @@ def load_mistral():
         device_map="auto",
         token=hf_token
     )
-    pipe = pipeline("text-generation", model=model, tokenizer=tokenizer, max_new_tokens=512)
-    return pipe
+    return pipeline("text-generation", model=model, tokenizer=tokenizer, max_new_tokens=512)
 
 mistral_pipe = load_mistral()
 
-# ✅ Global text buffer
+# ✅ Global variable to hold extracted content
 extracted_text = ""
 
 def extract_text_from_pptx_json(parsed_json: dict) -> str: