@chain decorator on class functions

[kim-borgen]

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_core.runnables import chain

class SomeModule:
    def __init__(self):
        self.value = 5
    
    @chain
    def add_one(self, input):
        return input + 1

    @chain
    def router(self, input):
        if self.value == 5:
            return self.add_one
        else:
            return input

    def get_chain(self):
        return self.router

m = SomeModule()
m.get_chain().invoke(5)

Description

Hi, we have a similar pattern to the toy example above. We wish to use the @chain decorator for simplicity to wrap python functions to RunnableLambdas. However, when I try to add this @chain decorator to our class function, it doesn`t recognize the self variable I assume and thus throws this error:

TypeError: SomeModule.router() missing 1 required positional argument: 'input'

System Info

Truncated package Information

langchain_core: 0.2.9
langchain: 0.2.5

[dosubot[bot]]

Hey @kim-borgen! I'm here to help you with any bugs, questions, or contribution-related matters. Let's take a look at the issue you've encountered together.

To resolve the issue where the @chain decorator does not recognize the self variable in class methods, you can modify your methods to be static or class methods. Here is how you can adjust your code:

Using Static Methods

from langchain_core.runnables import chain

class SomeModule:
    def __init__(self):
        self.value = 5
    
    @staticmethod
    @chain
    def add_one(input):
        return input + 1

    @staticmethod
    @chain
    def router(input):
        if SomeModule().value == 5:
            return SomeModule.add_one
        else:
            return input

    def get_chain(self):
        return self.router

m = SomeModule()
m.get_chain().invoke(5)

Using Class Methods

from langchain_core.runnables import chain

class SomeModule:
    def __init__(self):
        self.value = 5
    
    @classmethod
    @chain
    def add_one(cls, input):
        return input + 1

    @classmethod
    @chain
    def router(cls, input):
        if cls().value == 5:
            return cls.add_one
        else:
            return input

    def get_chain(self):
        return self.router

m = SomeModule()
m.get_chain().invoke(5)

By using static or class methods, you avoid the issue of the self variable not being recognized by the @chain decorator ^[1][2].

To continue talking to Dosu, mention @dosu.

[kim-borgen]

@dosu I need access to the dynamic variables in the class to execute our functionality, thus I cannot use the classmethod or staticmethod decorators.

[dosubot[bot]]

To use the @chain decorator on class methods while still having access to instance variables, you can define the method normally and then apply the @chain decorator. Here is an example:

from langchain_core.runnables import chain

class SomeModule:
    def __init__(self):
        self.value = 5
    
    @chain
    def add_one(self, input):
        return input + 1

    @chain
    def router(self, input):
        if self.value == 5:
            return self.add_one
        else:
            return input

    def get_chain(self):
        return self.router

m = SomeModule()
result = m.get_chain().invoke(5)
print(result)  # Output will be 6

In this example, add_one and router are decorated with @chain and can still access the instance variable self.value ^[1].

To continue talking to Dosu, mention @dosu.

[kim-borgen]

@dosu, as in my initial question, this throws the error TypeError: SomeModule.router() missing 1 required positional argument: 'input'

[dosubot[bot]]

To use the @chain decorator on class methods while still having access to instance variables and avoid the TypeError: SomeModule.router() missing 1 required positional argument: 'input' error, you need to ensure that the method signature includes self and properly handle the instance context. Here is an example of how to do this:

from langchain_core.runnables import chain

class SomeModule:
    def __init__(self):
        self.value = 5
    
    @chain
    def add_one(self, input):
        return input + 1

    @chain
    def router(self, input):
        if self.value == 5:
            return self.add_one(input)
        else:
            return input

    def get_chain(self):
        return self.router

m = SomeModule()
result = m.get_chain().invoke(5)
print(result)  # Output will be 6

In this example, the router method is correctly calling self.add_one(input) to ensure the input is passed correctly, and the invoke method is used to call the runnable with the required input ^[1].

To continue talking to Dosu, mention @dosu.

[avalatea]

I am having this problem as well. Should I open a ticket for this?