Refactor agent.py to remove the get_image_file tool and update app copy.py to process and print multiple questions from the API response.

agent.py

@@ -81,7 +81,7 @@ def web_search(keywords: str, max_results:int = 5) -> str:
 
 
 class AgentState(TypedDict):
-    messages: List[Union[HumanMessage, AIMessage]]
+    messages: List[Union[HumanMessage, AIMessage, SystemMessage]]
 
 
 def read_message(state: AgentState) -> AgentState:
@@ -107,7 +107,7 @@ def read_message(state: AgentState) -> AgentState:
 #     return {"messages": messages + [response]}
 
 # Augment the LLM with tools
-tools = [web_search,get_image_file]
+tools = [web_search]
 tools_by_name = {tool.name: tool for tool in tools}
 model_with_tools = model.bind_tools(tools)
 
@@ -124,34 +124,40 @@ def answer_message(state: AgentState) -> AgentState:
     Response: 24    
     """)]
     messages = prompt + messages
+
+    # First pass: let model decide whether to call web_search
     ai_msg = model_with_tools.invoke(messages)
     messages.append(ai_msg)
 
-    # Step 2: Execute tools and collect results
-    for tool_call in ai_msg.tool_calls:
-        # Execute the tool with the generated arguments
-        name = tool_call['name']
-        args = tool_call['args']
+    # If the model didn't request any tools, its content is already the answer
+    tool_calls = getattr(ai_msg, "tool_calls", None) or []
+    if not tool_calls:
+        print(f"Final response: {ai_msg}")
+        return {"messages": messages}
+
+    # Execute requested tools and append their text output into the conversation
+    for tool_call in tool_calls:
+        name = tool_call["name"]
+        args = tool_call["args"]
         tool = tools_by_name[name]
-        tool_result = tool.invoke(args)
-        messages.append(tool_result)
-    
+        tool_result = tool.invoke(args)  # this is a plain string from web_search
+        messages.append(HumanMessage(content=f"Tool result ({name}):\n{tool_result}"))
+
+    # Second pass: force a plain-text final answer (no tool calls expected)
     final_instruction = HumanMessage(
-    content=(
-            "Using the tool results above, provide the FINAL answer now. "
+        content=(
+            "Using the tool results above, provide the FINAL numeric/text answer now. "
             "Do not call any tools. Respond with only the answer."
         )
     )
     messages.append(final_instruction)
 
-    final_response = model_with_tools.invoke(messages)
-
-    # final_response = model_with_tools.invoke(messages)
+    final_response = model.invoke(messages)
     print(f"Final response: {final_response}")
-    final_response = final_response.content.split('</think>')[1].strip()
 
-    # Append the model's answer to the messages list
-    return {"messages": [final_response]}
+    # Return messages including the final AIMessage so BasicAgent reads .content
+    messages.append(final_response)
+    return {"messages": messages}
 
 
 

app copy.py

@@ -30,9 +30,14 @@ questions_url = f"{DEFAULT_API_URL}/questions"
 response = requests.get(questions_url, timeout=15)
 response.raise_for_status()
 questions_data = response.json()
-question_text = questions_data[3].get("question")
-output = agent(question_text)
-print(output)
+for item in questions_data[:5]:
+    question_text = item.get("question")
+    if question_text is None:
+        continue
+    output = agent(question_text)
+    print("Q:", question_text)
+    print("A:", output)
+    print("-" * 40)
 
 
 # def run_and_submit_all( profile: gr.OAuthProfile | None):