Using Shell as a tool with RequestCompletion

To allow for usage of shell as a tool, we can create a simple tool using from_fuction. The function could be modified to suit your needs, such as adding error handling or specific command restrictions. Below is a basic example of how to create a shell command execution tool using subprocess in Python.

import subprocess

def run_shell(command: str) -> str:
    """Run a bash command and return its output or error."""
    try:
        result = subprocess.run(command, shell=True, capture_output=True, text=True)
        if result.returncode == 0:
            return result.stdout.strip()
        else:
            return f"Error: {result.stderr.strip()}"
    except Exception as e:
        return f"Exception: {str(e)}"


bash_tool = rt.function_node(run_shell)

At this point, the tool can be used the same as any other RT tool. See the following code as a simple example.

import platform
import railtracks as rt
import asyncio

BashAgent = rt.agent_node(
    tool_nodes={bash_tool},
    system_message=f"You are a useful helper that can run local shell commands. "
                   f"You are on a {platform.system()} machine. Use appropriate shell commands to answer the user's questions.",
    llm=rt.llm.OpenAILLM("gpt-4o"),
)

user_prompt = """What directories are in the current directory?"""
message_history = rt.llm.MessageHistory()
message_history.append(rt.llm.UserMessage(user_prompt))

async def call_bash_agent():
    with rt.Session(logging_setting="VERBOSE"):
        result = await rt.call(BashAgent, message_history)

    print(result.content)

# asyncio.run(call_node())