Different behaviour between tool usage and directly calling a function

[fede-bello]

Checked other resources

• This is a bug, not a usage question. For questions, please use GitHub Discussions.
• I added a clear and detailed title that summarizes the issue.
• I read what a minimal reproducible example is (https://stackoverflow.com/help/minimal-reproducible-example).
• I included a self-contained, minimal example that demonstrates the issue INCLUDING all the relevant imports. The code run AS IS to reproduce the issue.

I was doing a small personal project where the objective is to given a prompt with a country,get an agent to display a map of the country.

I am using langchain and langgraph to create the agent, earth engine to get the data and geemap to display the map.

Going to simplify a bit the code for the sake of the example. The function I was using was something like this:

import ee
import geemap
from langchain_cohere import ChatCohere
from langchain.tools import tool
from IPython.display import display
from langgraph.prebuilt import create_react_agent
import asyncio

ee.Authenticate()
ee.Initialize(project="unicef-geospatial")

def get_uruguay_map():
    """Displays a map of Uruguay"""
    country_boundries = ee.FeatureCollection("USDOS/LSIB_SIMPLE/2017")
    uruguay_boundries = country_boundries.filter(ee.Filter.eq("country_na", "Uruguay"))

    country_map = geemap.Map()
    country_map.center_object(uruguay_boundries)
    country_map.add_layer(uruguay_boundries, {}, "Uruguay Boundaries")
    display(country_map)
    return "success"

If I call the function directly, it works as expected.

When changing the function to a tool and hooking it up to the agent, it doesn't work as expected. The code was having at the start was:

@tool
def get_uruguay_map():
    """Displays a map of Uruguay"""
    country_boundries = ee.FeatureCollection("USDOS/LSIB_SIMPLE/2017")
    uruguay_boundries = country_boundries.filter(ee.Filter.eq("country_na", "Uruguay"))

    country_map = geemap.Map()
    country_map.center_object(uruguay_boundries)
    country_map.add_layer(uruguay_boundries, {}, "Uruguay Boundaries")
    display(country_map)
    return "success"

llm = ChatCohere(temperature=0.0)
tools = [get_uruguay_map]

graph = create_react_agent(
    tools=tools,
    model=llm,
)

# %%
inputs = {
    "messages": [
        {
            "role": "user",
            "content": "Show me the map of Uruguay",
        }
    ]
}
graph.invoke(inputs)

With that code, I was getting the following output:

ToolMessage(content='Error: RuntimeError("There is no current event loop in thread \'ThreadPoolExecutor-3_0\'.")\n Please fix your mistakes.', name='get_uruguay_map', id='4bba79a1-c59a-4fbb-b065-f7e51122db5c', tool_call_id='eec8f8104e7a44a4ba63044c6a06dfa5', status='error'),

Which I couldn't diagnose but I found that it could be fixed by setting the event loop.

@tool
def get_uruguay_map():
    """Displays a map of Uruguay"""
    loop = asyncio.new_event_loop()
    asyncio.set_event_loop(loop)
    country_boundries = ee.FeatureCollection("USDOS/LSIB_SIMPLE/2017")
    uruguay_boundries = country_boundries.filter(ee.Filter.eq("country_na", "Uruguay"))

    country_map = geemap.Map()
    country_map.center_object(uruguay_boundries)
    country_map.add_layer(uruguay_boundries, {}, "Uruguay Boundaries")
    display(country_map)
    return "success"

llm = ChatCohere(temperature=0.0)
tools = [get_uruguay_map]

graph = create_react_agent(
    tools=tools,
    model=llm,
)

# %%
inputs = {
    "messages": [
        {
            "role": "user",
            "content": "Show me the map of Uruguay",
        }
    ]
}
graph.invoke(inputs)

However, now instead of getting the map centered on Uruguay, I am getting the map centered on the world, even though the layer is properly displayed.

I am not sure why this is happening or how to properly diagnose/fix the issue.

Any thoughts?

System Info

System Information
------------------
> OS:  Darwin
> OS Version:  Darwin Kernel Version 24.1.0: Thu Oct 10 21:00:32 PDT 2024; root:xnu-11215.41.3~2/RELEASE_ARM64_T6030
> Python Version:  3.12.7 (main, Oct  1 2024, 02:05:46) [Clang 15.0.0 (clang-1500.3.9.4)]

Package Information
-------------------
> langchain_core: 0.3.21
> langchain: 0.3.9
> langchain_community: 0.3.9
> langsmith: 0.1.147
> langchain_cohere: 0.3.3
> langchain_experimental: 0.3.3
> langchain_openai: 0.2.11
> langchain_text_splitters: 0.3.2
> langgraph_sdk: 0.1.43

Optional packages not installed
-------------------------------
> langserve

Other Dependencies
------------------
> aiohttp: 3.11.10
> async-timeout: Installed. No version info available.
> cohere: 5.13.3
> dataclasses-json: 0.6.7
> httpx: 0.28.1
> httpx-sse: 0.4.0
> jsonpatch: 1.33
> langsmith-pyo3: Installed. No version info available.
> numpy: 1.26.4
> openai: 1.57.0
> orjson: 3.10.12
> packaging: 24.2
> pandas: 2.2.3
> pydantic: 2.10.3
> pydantic-settings: 2.6.1
> PyYAML: 6.0.2
> requests: 2.32.3
> requests-toolbelt: 1.0.0
> SQLAlchemy: 2.0.36
> tabulate: 0.9.0
> tenacity: 9.0.0
> tiktoken: 0.8.0
> typing-extensions: 4.12.2

[gbaian10]

It looks like you're running in a Jupyter environment, which already has an event loop running. I suspect this might be causing the issue.
If you're in Jupyter, can the same code run in a .py file?

Or try using nest_asyncio to see if it solves the problem.

import nest_asyncio
nest_asyncio.apply()

[fede-bello]

Running it with nest_asyncio solves the first issue, but still the map is not being centered when I use it as a tool

[fede-bello]

Seems to be a geemap bug that for some reason appears when using the tool.
Thanks

[fede-bello]

For anyone having a similar issue, it's being tracked here