Update app.py

app.py

@@ -187,6 +187,7 @@ def initialize_llmchain(llm_choice, temperature, max_tokens, top_k, vector_db, p
     return qa_chain, "LLM initialized successfully!"
 
 def conversation(qa_chain, message, history):
+    """Fixed conversation function returning all required outputs"""
     response = qa_chain.invoke({
         "question": message,
         "chat_history": [(hist[0], hist[1]) for hist in history]
@@ -196,16 +197,33 @@ def conversation(qa_chain, message, history):
     if "Helpful Answer:" in response_answer:
         response_answer = response_answer.split("Helpful Answer:")[-1]
     
+    # Get source documents, ensure we have exactly 3
     sources = response["source_documents"][:3]
-    source_contents = [s.page_content.strip() for s in sources]
-    source_pages = [s.metadata.get("page", 0) + 1 for s in sources]
+    source_contents = []
+    source_pages = []
     
-    while len(sources) < 3:
+    # Process available sources
+    for source in sources:
+        source_contents.append(source.page_content.strip())
+        source_pages.append(source.metadata.get("page", 0) + 1)
+    
+    # Pad with empty values if we have fewer than 3 sources
+    while len(source_contents) < 3:
         source_contents.append("")
         source_pages.append(0)
     
-    return (qa_chain, gr.update(value=""), history + [(message, response_answer)] + 
-            source_contents + source_pages)
+    # Return all required outputs in correct order
+    return (
+        qa_chain,  # State
+        gr.update(value=""),  # Clear message box
+        history + [(message, response_answer)],  # Updated chat history
+        source_contents[0],  # First source
+        source_pages[0],    # First page
+        source_contents[1],  # Second source
+        source_pages[1],    # Second page
+        source_contents[2],  # Third source
+        source_pages[2]     # Third page
+    )
 
 def demo():
     evaluator = RAGEvaluator()