How to make the Claude model not respond with content when calling a tool?

[gbaian10]

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 dotenv import load_dotenv
from langchain_anthropic import ChatAnthropic
from langchain_core.tools import tool

load_dotenv()


@tool
def add(a: int, b: int) -> int:
    """Add two numbers."""
    return a + b


model = ChatAnthropic(model="claude-3-5-sonnet-20240620").bind_tools([add])
print(model.invoke("What is 1+1?"))

Description

The Claude models always include content when calling tools, but I don't need it, and generating it slows down the response speed.

According to the official documentation, it seems possible to disable the content response. How can I achieve this in Langchain?

System Info

langchain-core==0.3.9
langchain-anthropic==0.2.3

platform==windows
python-version==3.12.6

[dosubot[bot]]

I couldn't find specific information on how to suppress the content field in the ChatAnthropic model's tool responses within the LangChain repository. However, you might consider checking the official documentation or experimenting with any available configuration options in the ChatAnthropic class that might allow you to control the response format.

If the official documentation suggests that it's possible to disable the content response, it might involve setting a specific parameter or using a method that isn't directly exposed in the LangChain framework. You could also explore the source code of the ChatAnthropic class to see if there are any undocumented features or parameters that could help achieve this.

If you have access to the official documentation or any additional resources, they might provide further insights into how this can be configured.

To continue talking to Dosu, mention @dosu.