Update app.py

app.py

@@ -130,13 +130,13 @@ You have access to:
 When users ask about stocks or indices, fetch live data. Provide clear explanations with numbers."""
 
 # ========== Core conversational function ==========
-def process_query(query: str, history: List[Dict], exchange: str) -> Tuple[str, List[Dict], pd.DataFrame, str]:
+def process_query(query: str, history: List[List], exchange: str) -> Tuple[str, List[List], pd.DataFrame, str]:
     """
     Process user query with real-time data awareness.
-    history is a list of dicts: [{"role": "user", "content": "..."}, {"role": "assistant", "content": "..."}, ...]
+    history is a list of lists: [[user_msg1, assistant_msg1], [user_msg2, assistant_msg2], ...]
     Returns:
         - response_text: str (the assistant's reply)
-        - updated_history: list of dicts (new chat history)
+        - updated_history: list of lists (new chat history)
         - data_preview: pd.DataFrame
         - status: str (unused)
     """
@@ -210,13 +210,11 @@ def process_query(query: str, history: List[Dict], exchange: str) -> Tuple[str,
             llm = get_llm()
             from langchain_core.messages import HumanMessage, AIMessage, SystemMessage
             
-            # Convert Gradio history (list of dicts) to flat list for LangChain
+            # Convert Gradio history (list of lists) to flat list for LangChain
             flat_history = []
-            for turn in history:
-                if turn["role"] == "user":
-                    flat_history.append(turn["content"])
-                else:  # assistant
-                    flat_history.append(turn["content"])
+            for user_msg, assistant_msg in history:
+                flat_history.append(user_msg)
+                flat_history.append(assistant_msg)
             
             messages = [SystemMessage(content=SYSTEM_PROMPT)]
             for i, msg in enumerate(flat_history):
@@ -231,14 +229,11 @@ def process_query(query: str, history: List[Dict], exchange: str) -> Tuple[str,
         except Exception as e:
             response_text = f"LLM error: {str(e)}"
     
-    # Update history in Gradio's new format (list of dicts)
-    updated_history = history + [
-        {"role": "user", "content": query},
-        {"role": "assistant", "content": response_text}
-    ]
+    # Update history in Gradio's classic format (list of lists)
+    updated_history = history + [[query, response_text]]
     # Keep only last 10 exchanges (20 messages)
-    if len(updated_history) > 20:
-        updated_history = updated_history[-20:]
+    if len(updated_history) > 10:
+        updated_history = updated_history[-10:]
     
     return response_text, updated_history, data_preview, ""
 
@@ -366,7 +361,8 @@ with gr.Blocks(title="AI-Powered Index Analyst with Real-Time Data") as demo:
     """)
     
     with gr.Tab("💬 Conversational Assistant"):
-        chat_history = gr.Chatbot(label="Conversation", height=400, type="messages")
+        # Remove type="messages" – use classic list-of-lists format
+        chat_history = gr.Chatbot(label="Conversation", height=400)
         msg = gr.Textbox(
             label="Ask about stocks or indices",
             placeholder="E.g., 'What's the price of RELIANCE?' or 'Compare Nifty 50 index using Laspeyres and Fisher'"