Update app.py

app.py

@@ -12,6 +12,9 @@ import time
 import asyncio
 from concurrent.futures import ThreadPoolExecutor
 
+from transformers import AutoModelForSeq2SeqLM  # ✅ Needed for T5 and FLAN models
+
+
 # Suppress warnings for cleaner output
 warnings.filterwarnings("ignore")
 
@@ -24,7 +27,7 @@ import faiss
 import torch
 from transformers import (
     AutoTokenizer, 
-    AutoModelForCausalLM, 
+    AutoModelForSeq2SeqLM,  
     BitsAndBytesConfig,
     pipeline
 )
@@ -311,18 +314,22 @@ class OptimizedGazaRAGSystem:
 
     
     def _initialize_llm(self):
-        """Initialize FLAN-T5 model (CPU-friendly)"""
+        """Load flan-t5-base for CPU fallback"""
         model_name = "google/flan-t5-base"
         try:
             logger.info(f"🔄 Loading fallback CPU model: {model_name}")
             self.tokenizer = AutoTokenizer.from_pretrained(model_name)
-            self.llm = AutoModelForCausalLM.from_pretrained(model_name)
-            self.generation_pipeline = pipeline("text2text-generation", model=self.llm,tokenizer=self.tokenizer,return_full_text=False)
+            self.llm = AutoModelForSeq2SeqLM.from_pretrained(model_name)
+            self.generation_pipeline = pipeline(
+                "text2text-generation",  # ✅ correct pipeline for T5
+                model=self.llm,
+                tokenizer=self.tokenizer,
+                return_full_text=False
+            )
             logger.info("✅ FLAN-T5 model loaded successfully")
         except Exception as e:
             logger.error(f"❌ Error loading FLAN-T5 model: {e}")
             self.llm = None
-            self.generation_pipeline = None