make sure that @tool decorator can be applied to a method (regression?)

[eyurtsev]

Privileged issue

• [X] I am a LangChain maintainer, or was asked directly by a LangChain maintainer to create an issue here.

Issue Content

from langchain_core.tools import tool

class MyAPI:
    def __init__(self):
        self.base_url = 'https://en.wikipedia.org/wiki/'

    @tool
    def lookup(self, page) -> str:
        """Look up a page on wikipedia. Include just the name of the page. Rest of the URL will be filled in."""
        url = self.base_url + item
        return request.get(url)        

api = MyAPI()
api.lookup.invoke('hello')

[GlassesNoGlasses]

Hey @eyurtsev! I'm part of a group of UofT developers interested in this issue. Can we tackle it? If so, what specific requirements and output are you looking for?

[eyurtsev]

Yes, you're welcome to try to tackle it. What kind of timeline do you have in mind?

---

We have a set of unit tests for tools: https://github.com/langchain-ai/langchain/blob/master/libs/core/tests/unit_tests/test_tools.py

We want to make sure that the @tool decorator has the correct behavior whether it's a standalone function or a method.

@tool
def foo(x: int, y: float) -> str:
  """bar"""

@tool
async def foo(x: int, y: float) -> str:
  """bar"""

class A:
  @tool
  def foo(self, x: int, y: float) -> str:
    """bar"""

class A:
  @tool
  async def foo(self, x: int, y: float) -> str:
    """bar"""

All behave the same.

• The code should be executable with .invoke or .ainvoke
• foo.args should yield the same schema. And the schema should not include self

[GlassesNoGlasses]

Thank you for the succinct requirements, that will really help us get started. Our timeline to get this finished is around 2 or 3 weeks from now, though the sooner the better. Does that work for you?

[eyurtsev]

Could you check in before getting started. This is in core so we may end up prioritizing this fix