Update app.py

app.py

@@ -503,7 +503,7 @@ def chat_with_llm(df,question):
 def bind_llm(llm, tools,prompt_template):
     llm = llm.bind()
     agent = create_tool_calling_agent(llm, tools, ChatPromptTemplate.from_template(prompt_template))
-    agent_executor = AgentExecutor(agent=agent, tools=tools, verbose=True)
+    agent_executor = AgentExecutor(agent=agent, tools=tools, verbose=True,return_intermediate_steps=True ) # Enable intermediate steps tracking)
     return agent_executor
 
 # Define input and output models using Pydantic
@@ -697,10 +697,10 @@ def answer_question_thread(user_question, chatbot,audio=None):
         except sr.RequestError:
             user_question = "Could not request results from Google Speech Recognition service."
         """
-
+    intermediate_steps = []
     while iterations < max_iterations:
 
-        response = agent_executor.invoke({"input": user_question}, config={"callbacks": [langfuse_handler]}, early_stopping_method="generate")
+        response, intermediate_steps = agent_executor.invoke({"input": user_question}, config={"callbacks": [langfuse_handler]}, early_stopping_method="generate")
         #create_file_HF()
         if isinstance(response, dict):
             response_text = response.get("output", "")
@@ -754,13 +754,13 @@ def answer_question_thread(user_question, chatbot,audio=None):
         return user_question, response_text
     else:
         if ("max iterations" in response_text):
-            print("11111::")
-            print(agent_executor.agent.return_values)
-            print("1122::")
-            for response_values in agent_executor.agent.return_values:
-                print(response_values)
-            print("222:::")
-            print(agent_executor.agent.return_stopped_response)
+            # Loop through and print all intermediate steps
+            for i, (action, observation) in enumerate(intermediate_steps):
+                print(f"\nIteration {i + 1}:")
+                print(f"  Action: {action.tool}")
+                print(f"  Tool Input: {action.tool_input}")
+                print(f"  Observation: {observation}")
+           
         return user_question, response_text
     # response_text = response_text.replace('\n', ' ').replace('  ', ' ').strip()
     # return response_text