How unittest a tool in the latest versions?

[marcoaleixo]

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

e

from langchain_core.runnables.config import RunnableConfig

@tool()
def search_knowledge_base_tool(question: str, config: RunnableConfig):
    """longdescription
    """

    user_id = config['configurable']['user_id']

My testcase

    def test_search_knowledge_base_tool(self, mock_get_conversation):

        config = RunnableConfig(
            recursion_limit=5,
            configurable={
                "folder_ids": [str(self.folder.external_id)],
                "conversation_id": str(self.conversation.conversation_id),
                "user_id": str(self.user.external_id),
                "message_id": str(uuid.uuid4()),
            }
        )

        question = "O que o documento sobre dipirona diz?"

        input_data = {"question": question, "config": config}

        input_data = {
            "question": "What is AI?",
            "config": {
                "tags": [],
                "metadata": {},
                "callbacks": None,
                "recursion_limit": 25,
                "configurable": {"user_id": "12345"}
            }
        }
        response = search_knowledge_base_tool.invoke(input_data)

Description

I'm trying to write a unittest for this tool.
When I execute this, the parameter config on my function tool is a default RunnableConfig and does not contain the expected fields.
That means on my tool I cant access config['configurable']['user_id'].
I tried different things ( as you can see on my code ) and everytime my configurable is replaced with the default value.

Ah, this tool works fine when executing it in our normal flow ( agent setup etc )

System Info

root@59b2fbb97124:/mrturing# pip freeze | grep langchain
langchain==0.3.9
langchain-community==0.3.8
langchain-core==0.3.21
langchain-experimental==0.3.3
langchain-google-vertexai==2.0.8
langchain-openai==0.2.10
langchain-text-splitters==0.3.2

[feijoes]

@marcoaleixo You need to pass the config as a second parameter in the invocation:

search_knowledge_base_tool.invoke(input_data, config)

[marcoaleixo]

Yeaah, that's it!
Thanks.

[marcoaleixo]

Ah, what if the tool accepts three arguments?

@tool()
def message_extraction_tool(question: str, config: RunnableConfig, state: Annotated[dict, InjectedState]):

result = message_extraction_tool.invoke(input={"question": question, "config": config, "state": state})

didn't did the trick

[marcoaleixo]

For future reference

result = message_extraction_tool.invoke(input={"question": question, "state": state}, config=config)

I need to pas config as separate argument.