Update app.py

app.py

@@ -406,15 +406,17 @@ Provide a complete, detailed answer (150-250 words):"""
         logger.error(f"  ✗ Generation error: {e}")
         return None
 
-def synthesize_direct_answer(
-# Removed synthetic fallback - only use LLM
+# ============================================================================
+# MAIN RAG FUNCTION
+# ============================================================================
+
 def generate_answer_langchain(
     query: str,
     vectorstore,
     llm_client
 ) -> str:
     """
-    Main RAG pipeline: Retrieve → Generate → Fallback
+    Main RAG pipeline: Retrieve → Generate (no fallback)
     """
     logger.info(f"\n{'='*80}")
     logger.info(f"Processing query: '{query}'")
@@ -442,10 +444,10 @@ def generate_answer_langchain(
         else:
             logger.warning(f"  → Attempt {attempt}/4 failed, retrying...")
     
-    # Step 3: Fallback if all attempts fail
+    # Step 3: If all attempts fail, return error
     if not llm_answer:
-        logger.error(f"  ✗ All 4 LLM attempts failed - using enhanced fallback")
-        llm_answer = synthesize_direct_answer(query, retrieved_docs)
+        logger.error(f"  ✗ All 4 LLM attempts failed")
+        return "I apologize, but I'm having trouble generating a response. Please try rephrasing your question or ask something else."
     
     return llm_answer
 
@@ -470,12 +472,12 @@ def fashion_chatbot(message: str, history: List[List[str]]):
             message.strip(),
             vectorstore,
             top_k=CONFIG["top_k"]
-    # Step 3: If all attempts fail, return error
-    if not llm_answer:
-        logger.error(f"  ✗ All 4 LLM attempts failed")
-        return "I apologize, but I'm having trouble generating a response. Please try rephrasing your question or ask something else."
-    
-    return llm_answer
+        )
+        
+        if not retrieved_docs:
+            yield "I couldn't find relevant information to answer your question."
+            return
+        
         # Show generating indicator
         yield f"💭 Generating answer ({len(retrieved_docs)} sources found)..."
         
@@ -488,12 +490,11 @@ def fashion_chatbot(message: str, history: List[List[str]]):
             if llm_answer:
                 break
         
-        # Fallback if needed
+        # If LLM fails, show error
         if not llm_answer:
-            logger.error(f"  ✗ All LLM attempts failed - using enhanced fallback")
-            llm_answer = synthesize_direct_answer(message.strip(), retrieved_docs)
-            # Notify user about the issue
-            yield "⚠️ LLM temporarily unavailable, providing knowledge-based answer...\n\n"
+            logger.error(f"  ✗ All LLM attempts failed")
+            yield "I apologize, but I'm having trouble generating a response. Please try rephrasing your question."
+            return
         
         # Stream the answer word by word for natural flow
         import time
@@ -515,11 +516,12 @@ def fashion_chatbot(message: str, history: List[List[str]]):
 # ============================================================================
 # INITIALIZE AND LAUNCH
 # ============================================================================
-        # If LLM fails, show error
-        if not llm_answer:
-            logger.error(f"  ✗ All LLM attempts failed")
-            yield "I apologize, but I'm having trouble generating a response. Please try rephrasing your question."
-            return
+
+# Global variables
+llm_client = None
+embeddings = None
+vectorstore = None
+
 def startup():
     """Initialize all models and load vector store"""
     global llm_client, embeddings, vectorstore