Update app.py

app.py

@@ -56,7 +56,7 @@ load_dotenv()
 
 # langfuse analytics
 from langfuse.callback import CallbackHandler
-
+from langchain.callbacks import CallbackManager, BaseCallbackHandler
 # Inventory API data table
 from tabulate import tabulate
 
@@ -121,6 +121,13 @@ branch = os.getenv("branch")
 langfuse_handler = CallbackHandler()
 langfuse_handler.auth_check()  # Optional: Checks if the authentication is successful
 
+# Create the callback manager and add both the response history and langfuse handler
+callback_handler = ResponseHistoryCallback()
+
+
+callback_manager = CallbackManager([callback_handler, langfuse_handler])
+
+
 nltk.download('punkt')
 
 open_api_key_token = os.getenv("OPENAI_API_KEY")
@@ -156,7 +163,18 @@ apis = [
 # LLM setup
 llm = ChatOpenAI(model="gpt-4o-mini", max_tokens=300, temperature=0.1)
 llm_chart = OpenAI(is_safe=False)
-
+# Define a custom callback to capture responses and track history
+class ResponseHistoryCallback(BaseCallbackHandler):
+    def __init__(self):
+        self.history = []  # Initialize history to store responses
+
+    def on_agent_finish(self, output):
+        # Capture the response when the agent finishes
+        self.history.append(output)
+        print(f"Agent's response: {output}")  # Optionally print
+
+    def get_history(self):
+        return self.history  # Return the captured history
 def get_schema(_):
     schema_info = db.get_table_info()  # This should be a string of your SQL schema
     return schema_info
@@ -705,9 +723,10 @@ def answer_question_thread(user_question, chatbot,audio=None):
 
     while iterations < max_iterations:
 
-        response = agent_executor.invoke({"input": user_question}, config={"callbacks": [langfuse_handler]}, early_stopping_method="generate")
+        #response = agent_executor.invoke({"input": user_question}, config={"callbacks": [langfuse_handler]}, early_stopping_method="generate")
+        response = agent_executor.invoke({"input": user_question}, config={"callbacks": [callback_manager]}, early_stopping_method="generate")
         # Track the response
-        agent_responses.append(agent_output)
+        agent_responses.append(response)
         
         #create_file_HF()
         if isinstance(response, dict):
@@ -771,9 +790,10 @@ def answer_question_thread(user_question, chatbot,audio=None):
     else:
         if("max iterations" in response_text):
             print("max iterations error in 2")
-            print(agent_responses[-1])
-            if agent_executor.history:
-                print(agent_executor.history[-1])
+            # Access the response history after execution
+            response_history = callback_handler.get_history()
+            print("Response History:", response_history)
+
         return user_question, response_text
     # response_text = response_text.replace('\n', ' ').replace('  ', ' ').strip()
     # return response_text