Update app.py

app.py

@@ -6,6 +6,7 @@ from langchain_community.utilities import WikipediaAPIWrapper
 from langchain.agents.agent_types import AgentType
 from langchain.agents import Tool, initialize_agent
 from langchain.callbacks import StreamlitCallbackHandler
+import datetime # Import datetime to add current date context
 
 # Set up Streamlit page configuration
 st.set_page_config(page_title="General Knowledge Assistant", page_icon="🧭")
@@ -18,49 +19,66 @@ if not groq_api_key:
     st.info("Please add your Groq API key to continue")
     st.stop()
 
-# Initialize the LLM (Groq API - llama-3.1-70b)
+# Initialize the LLM (Using the model from your original code)
+# You might consider trying 'llama3-70b-8192' if Maverick struggles with tool selection
 llm = ChatGroq(model="meta-llama/llama-4-maverick-17b-128e-instruct", groq_api_key=groq_api_key)
 
 # Initialize Wikipedia tool for information retrieval
-wikipedia_wrapper = WikipediaAPIWrapper()
+wikipedia_wrapper = WikipediaAPIWrapper(top_k_results=2, doc_content_chars_max=2000) # Limit results slightly
 wikipedia_tool = Tool(
-    name="Wikipedia",
+    name="Wikipedia Search", # Renamed slightly for clarity
     func=wikipedia_wrapper.run,
-    description="A tool for searching the Internet to find information on various topics, including general knowledge."
+    description=(
+        "Use this tool to find specific information, facts, or details about people, places, events, or topics. "
+        "It is especially useful for getting CURRENT and UP-TO-DATE information or checking facts that might change over time. "
+        "Input should be a clear search query."
+    )
 )
 
-# Prompt template for general knowledge questions
-prompt = """
-You are a knowledgeable assistant. Your task is to answer the user's questions accurately, using your general knowledge.
-If the answer is not readily available in your knowledge base, search Wikipedia for relevant information.
-Your information should be accurate and up to date.Whenever I tell you to write essay give a title also to the essay.
-Question: {question}
+# Prompt template for general knowledge questions (used by the Reasoning Tool)
+# Added a note about the current date to potentially help the LLM contextualize recency
+current_date = datetime.datetime.now().strftime("%Y-%m-%d")
+prompt_text = f"""
+You are a knowledgeable assistant. The current date is {current_date}.
+Your task is to answer the user's questions accurately using your general knowledge.
+If you are asked to write an essay, please provide a title for the essay.
+Your information should be accurate and up-to-date based on your internal knowledge cutoff.
+If you suspect your internal knowledge might be outdated for the question, mention that the information might not be the absolute latest.
+
+Question: {{question}}
 Answer:
 """
 
 # Initialize the prompt template
 prompt_template = PromptTemplate(
     input_variables=["question"],
-    template=prompt
+    template=prompt_text
 )
 
-# Combine all the tools into a chain for answering general knowledge questions
+# Create the LLMChain for the Reasoning tool
 chain = LLMChain(llm=llm, prompt=prompt_template)
 
 # Reasoning tool for logic-based or factual questions
 reasoning_tool = Tool(
-    name="Reasoning tool",
+    name="General Knowledge and Reasoning", # Renamed slightly for clarity
     func=chain.run,
-    description="A tool for answering general knowledge questions using logical reasoning and factual information.Try to use the latest information"
+    description=(
+        "Use this tool to answer general knowledge questions, perform reasoning tasks, or explain concepts based on the AI's internal knowledge base. "
+        "This tool relies on the AI's trained data, which might have a knowledge cut-off date. Do NOT use this tool if the question likely requires very recent information (use Wikipedia Search instead)."
+    )
 )
 
 # Initialize the agent with the tools and LLM
+# Ensure verbose=False and handle_parsing_errors=True as per your original code
 assistant_agent = initialize_agent(
     tools=[wikipedia_tool, reasoning_tool],
     llm=llm,
     agent=AgentType.ZERO_SHOT_REACT_DESCRIPTION,
-    verbose=False,
-    handle_parsing_errors=True
+    verbose=False, # Set to True temporarily if you need to debug the agent's thought process
+    handle_parsing_errors=True,
+    # Add max_iterations to prevent potential infinite loops if the agent gets stuck
+    max_iterations=5,
+    early_stopping_method="generate" # Stop generating if it thinks it's done
 )
 
 # Initialize session state for message history if it doesn't exist
@@ -73,23 +91,45 @@ if "messages" not in st.session_state:
 for msg in st.session_state.messages:
     st.chat_message(msg["role"]).write(msg['content'])
 
-# Get the user's question
-question = st.text_area("Enter your question:", "Please enter your general knowledge question here")
-
-# Handle the button click to process the question
-if st.button("find my answer"):
-    if question:
-        with st.spinner("Generate response.."):
-            st.session_state.messages.append({"role":"user","content":question})
-            st.chat_message("user").write(question)
-
-            st_cb=StreamlitCallbackHandler(st.container(),expand_new_thoughts=False)
-            response=assistant_agent.run(st.session_state.messages,callbacks=[st_cb]
-                                         )
-            st.session_state.messages.append({'role':'assistant',"content":response})
-            st.write('### Response:')
-            st.success(response)
-
-    else:
-        st.warning("Please enter the question")
-
+# Get the user's question (using st.chat_input for better UI)
+if user_query := st.chat_input("Please enter your general knowledge question here"):
+    # Add user message to state and display it
+    st.session_state.messages.append({"role": "user", "content": user_query})
+    st.chat_message("user").write(user_query)
+
+    # Format the input for the agent - pass only the latest user query
+    # Include conversation history *if* the agent type supports it well,
+    # but ZERO_SHOT_REACT_DESCRIPTION primarily focuses on the latest input.
+    # We will pass only the user_query for simplicity and correctness with .run()
+    agent_input = user_query
+
+    # Add context about the conversation history potentially (optional, advanced)
+    # agent_input = f"Previous conversation:\n{st.session_state.messages}\n\n Current question: {user_query}"
+
+    # Generate and display response
+    with st.chat_message("assistant"):
+        st_cb = StreamlitCallbackHandler(st.container(), expand_new_thoughts=False)
+        # *** FIX: Pass only the user_query string to the agent's run method ***
+        response = assistant_agent.run(agent_input, callbacks=[st_cb])
+        st.session_state.messages.append({"role": "assistant", "content": response})
+        st.write(response) # Display the final response
+
+# Note: Removed the text_area + button combo in favor of st.chat_input for a cleaner chat interface.
+# If you prefer the text_area and button:
+# user_question = st.text_area("Enter your question:", "Please enter your general knowledge question here", key="user_q_input")
+# if st.button("Find my answer", key="submit_q"):
+#     if user_question and user_question != "Please enter your general knowledge question here":
+#         # Add user message to state and display it
+#         st.session_state.messages.append({"role": "user", "content": user_question})
+#         st.chat_message("user").write(user_question)
+
+#         agent_input = user_question # Use the text_area content
+
+#         # Generate and display response
+#         with st.chat_message("assistant"):
+#             st_cb = StreamlitCallbackHandler(st.container(), expand_new_thoughts=False)
+#             response = assistant_agent.run(agent_input, callbacks=[st_cb])
+#             st.session_state.messages.append({"role": "assistant", "content": response})
+#             st.write(response)
+#     else:
+#         st.warning("Please enter a question.")