Skip to main content

How to add a human-in-the-loop for tools

There are certain tools that we don't trust a model to execute on its own. One thing we can do in such situations is require human approval before the tool is invoked.

info

This how-to guide shows a simple way to add human-in-the-loop for code running in a jupyter notebook or in a terminal.

To build a production application, you will need to do more work to keep track of application state appropriately.

We recommend using langgraph for powering such a capability. For more details, please see this guide.

Setup​

We'll need to install the following packages:

%pip install --upgrade --quiet langchain

And set these environment variables:

import getpass
import os

# If you'd like to use LangSmith, uncomment the below:
# os.environ["LANGCHAIN_TRACING_V2"] = "true"
# os.environ["LANGCHAIN_API_KEY"] = getpass.getpass()

Chain​

Let's create a few simple (dummy) tools and a tool-calling chain:

pip install -qU langchain-openai
import getpass
import os

os.environ["OPENAI_API_KEY"] = getpass.getpass()

from langchain_openai import ChatOpenAI

llm = ChatOpenAI(model="gpt-3.5-turbo-0125")
from typing import Dict, List

from langchain_core.messages import AIMessage
from langchain_core.runnables import Runnable, RunnablePassthrough
from langchain_core.tools import tool


@tool
def count_emails(last_n_days: int) -> int:
"""Multiply two integers together."""
return last_n_days * 2


@tool
def send_email(message: str, recipient: str) -> str:
"Add two integers."
return f"Successfully sent email to {recipient}."


tools = [count_emails, send_email]
llm_with_tools = llm.bind_tools(tools)


def call_tools(msg: AIMessage) -> List[Dict]:
"""Simple sequential tool calling helper."""
tool_map = {tool.name: tool for tool in tools}
tool_calls = msg.tool_calls.copy()
for tool_call in tool_calls:
tool_call["output"] = tool_map[tool_call["name"]].invoke(tool_call["args"])
return tool_calls


chain = llm_with_tools | call_tools
chain.invoke("how many emails did i get in the last 5 days?")
[{'name': 'count_emails',
'args': {'last_n_days': 5},
'id': 'toolu_01QYZdJ4yPiqsdeENWHqioFW',
'output': 10}]

Adding human approval​

Let's add a step in the chain that will ask a person to approve or reject the tall call request.

On rejection, the step will raise an exception which will stop execution of the rest of the chain.

import json


class NotApproved(Exception):
"""Custom exception."""


def human_approval(msg: AIMessage) -> AIMessage:
"""Responsible for passing through its input or raising an exception.

Args:
msg: output from the chat model

Returns:
msg: original output from the msg
"""
tool_strs = "\n\n".join(
json.dumps(tool_call, indent=2) for tool_call in msg.tool_calls
)
input_msg = (
f"Do you approve of the following tool invocations\n\n{tool_strs}\n\n"
"Anything except 'Y'/'Yes' (case-insensitive) will be treated as a no.\n >>>"
)
resp = input(input_msg)
if resp.lower() not in ("yes", "y"):
raise NotApproved(f"Tool invocations not approved:\n\n{tool_strs}")
return msg
chain = llm_with_tools | human_approval | call_tools
chain.invoke("how many emails did i get in the last 5 days?")
Do you approve of the following tool invocations

{
"name": "count_emails",
"args": {
"last_n_days": 5
},
"id": "toolu_01WbD8XeMoQaRFtsZezfsHor"
}

Anything except 'Y'/'Yes' (case-insensitive) will be treated as a no.
>>> yes
[{'name': 'count_emails',
'args': {'last_n_days': 5},
'id': 'toolu_01WbD8XeMoQaRFtsZezfsHor',
'output': 10}]
try:
chain.invoke("Send sally@gmail.com an email saying 'What's up homie'")
except NotApproved as e:
print()
print(e)
Do you approve of the following tool invocations

{
"name": "send_email",
"args": {
"recipient": "sally@gmail.com",
"message": "What's up homie"
},
"id": "toolu_014XccHFzBiVcc9GV1harV9U"
}

Anything except 'Y'/'Yes' (case-insensitive) will be treated as a no.
>>> no
``````output

Tool invocations not approved:

{
"name": "send_email",
"args": {
"recipient": "sally@gmail.com",
"message": "What's up homie"
},
"id": "toolu_014XccHFzBiVcc9GV1harV9U"
}

Was this page helpful?


You can also leave detailed feedback on GitHub.