AgentExecutor ignores max_iteration

[blafab-hg]

Checked other resources

• I added a very descriptive title to this question.
• I searched the LangChain documentation with the integrated search.
• I used the GitHub search to find a similar question and didn't find it.

Commit to Help

• I commit to help with one of those options 👆

Example Code

from langchain import hub
from langchain.agents import create_tool_calling_agent, AgentExecutor
from langchain.tools import BingSearchResults
import os

bing_search = BingSearchResults(api_key=os.getenv("BING_KEY"))

tools = [bing_search]

from langchain_openai import ChatOpenAI
model = ChatOpenAI(model="gpt-4")

prompt = """
You are an agent that finds the country for a list of 50 random cities. 
You will run a separate Bing search query to find the country for each city in the list.
Here is the list of cities:
Tokyo, New York, Paris, London, Berlin, Sydney, Moscow, Beijing, Rio de Janeiro, Cairo,
Mumbai, Los Angeles, Buenos Aires, Rome, Istanbul, Toronto, Mexico City, Jakarta, Seoul, Lagos,
Bangkok, Madrid, Kuala Lumpur, Nairobi, Lima, Tehran, Baghdad, Dhaka, Karachi, Santiago,
Singapore, Hong Kong, Dubai, Riyadh, Cape Town, Hanoi, Athens, Vienna, Warsaw, Budapest,
Prague, Brussels, Zurich, Stockholm, Oslo, Helsinki, Copenhagen, Dublin, Lisbon.
"""

agent = create_tool_calling_agent(model, tools, prompt)

agent_executor = AgentExecutor(agent=agent, tools=tools, max_iterations=5)

response = agent_executor.invoke({"input": "Find the country for the list of cities provided."})
print(response["output"])

Description

We observed a similar query to the one above and saw that despite the max_iterations on the AgentExecutor we saw 50 invocations of the BingSearchResults tool. This causes huge load spikes on related services. We want to limit the amount of times a tool can be called by the Executor.

System Info

langchain==0.3.4
langchain-aws==0.2.2
langchain-community==0.3.3
langchain-core==0.3.12
langchain-openai==0.2.3
langchain-text-splitters==0.3.0