Struggling to pass multiple arguments to Openai function agent in Langchain

Other Questions OpenAI
1

Hey guys, I’m stuck trying to create an agent with Langchain that can place orders for me. I’ve set up a function/Tool to handle the order placement, but it needs several inputs.

When I run my code, I keep getting this error: TypeError: __init__() takes 1 positional argument but 7 were given

Here’s a simplified version of what I’m working with:

class OrderPlacement(BaseModel):
    customer_name: str
    contact_number: str
    item_name: str
    birth_date: str
    item_count: str

class PlaceOrderTool(BaseTool):
    name = "submit_order"
    description = "Use this to place an order. Provide name, contact, item, birthdate, and quantity."

    def _run(self, customer_name: str, contact_number: str, item_name: str, birth_date: str, item_count: str):
        order_result = submit_order(customer_name, contact_number, birth_date, item_name, item_count)
        return order_result

    args_schema: Optional[Type[BaseModel]] = OrderPlacement

tools = [PlaceOrderTool()]
ai_agent = initialize_agent(tools, llm, agent=AgentType.OPENAI_FUNCTIONS, verbose=True)

a_agent.run("Place an order for Jane Doe, phone 1234567890, born 01/01/1990, for 3 blue widgets")

Can anyone spot what I’m doing wrong here? Thanks!

2

I’ve been using Langchain for a while now, and I’ve run into this exact problem before. The issue is with how the _run method is defined in your PlaceOrderTool class. Instead of specifying each argument separately, you need to use **kwargs to capture all the arguments at once.

Here’s what worked for me:

def _run(self, **kwargs):
    return submit_order(**kwargs)

This change allows the tool to accept any number of arguments defined in your OrderPlacement schema. It should resolve that TypeError you’re seeing.

Also, make sure your submit_order function can handle these keyword arguments correctly. You might need to unpack them within the function if it’s expecting individual parameters.

One more thing - double-check that your OrderPlacement schema matches exactly with what your submit_order function expects. Any mismatch there could cause issues down the line.

3

hey john, looks like ur issue might be with the _run method. try changing it to:

def _run(self, **kwargs):
    return submit_order(**kwargs)

this should handle multiple args better. lmk if that works!

4

yo john, i had the same problem last week. try this:

def _run(self, **kwargs):
    return submit_order(**kwargs)

it fixed it for me. also, make sure ur submit_order function can handle those kwargs. good luck man!

5

I encountered a similar issue when working with Langchain’s function agents. The problem lies in how the tool’s _run method is defined. Instead of listing all parameters individually, you should use **kwargs to capture all arguments passed to the function. Here’s how you can modify your PlaceOrderTool class:

class PlaceOrderTool(BaseTool):
    name = "submit_order"
    description = "Use this to place an order. Provide name, contact, item, birthdate, and quantity."

    def _run(self, **kwargs):
        return submit_order(**kwargs)

    args_schema: Optional[Type[BaseModel]] = OrderPlacement

This modification allows the tool to accept any number of arguments defined in your OrderPlacement schema. It should resolve the TypeError you’re encountering. Remember to ensure that your submit_order function can handle these keyword arguments correctly.