Update app.py

app.py

@@ -530,16 +530,16 @@ def generate_answer_langchain(
         llm_answer = generate_llm_answer(query, retrieved_docs, llm_client, attempt)
         
         if llm_answer:
+            logger.info(f"  ✅ LLM answer generated successfully")
+            break
+        else:
+            logger.warning(f"  → Attempt {attempt}/4 failed, retrying...")
+    
     # Step 3: Fallback if all attempts fail
     if not llm_answer:
         logger.error(f"  ✗ All 4 LLM attempts failed - using enhanced fallback")
         llm_answer = synthesize_direct_answer(query, retrieved_docs)
     
-    return llm_answerk if all attempts fail
-    if not llm_answer:
-        logger.error(f"  ✗ All 4 LLM attempts failed - using fallback")
-        llm_answer = synthesize_direct_answer(query, retrieved_docs)
-    
     return llm_answer
 
 # ============================================================================
@@ -576,6 +576,11 @@ def fashion_chatbot(message: str, history: List[List[str]]):
         llm_answer = None
         for attempt in range(1, 5):
             logger.info(f"\n  🤖 LLM Generation Attempt {attempt}/4")
+            llm_answer = generate_llm_answer(message.strip(), retrieved_docs, llm_client, attempt)
+            
+            if llm_answer:
+                break
+        
         # Fallback if needed
         if not llm_answer:
             logger.error(f"  ✗ All LLM attempts failed - using enhanced fallback")
@@ -583,11 +588,6 @@ def fashion_chatbot(message: str, history: List[List[str]]):
             # Notify user about the issue
             yield "⚠️ LLM temporarily unavailable, providing knowledge-based answer...\n\n"
         
-        # Fallback if needed
-        if not llm_answer:
-            logger.error(f"  ✗ All LLM attempts failed - using fallback")
-            llm_answer = synthesize_direct_answer(message.strip(), retrieved_docs)
-        
         # Stream the answer word by word for natural flow
         import time
         words = llm_answer.split()