How to set max_execution_time to agents with PythonREPLTool

[Troyanovsky]

I am building an agent that can execute code with the PythonREPLTool to provide answers to questions. But I found that the "max_execution_time" parameter is not applicable.

When the agent writes code and executes it with PythonREPLTool, the agent may include "input()" function in the code to ask for user input. But it does not time out when the max_execution_time is reached and no user input is provided.

How can I set max_execution_time to agents with PythonREPLTool?

(I guess another solution could be detecting if the agent has written code with "input()" and deny it from using the code. But I don't know if it's possible to do that.)

Here is my code to initialize the agent:

llm = OpenAI(temperature=0)

tools = []
tools.append(wolfram_alpha_tool)
tools.extend(load_tools(["python_repl"], llm=llm))

agent = initialize_agent(tools, llm, agent=AgentType.ZERO_SHOT_REACT_DESCRIPTION, return_intermediate_steps=True, max_execution_time=30,early_stopping_method="generate")