fix(core): fix strict schema generation for functions with optional args

[keenborder786]

• When we use strict mode in OpenAI Function, we need to make sure additionalProperties set to False, otherwise we get 400 error. This is also stated in the Official Doc: https://platform.openai.com/docs/guides/function-calling#strict-mode
• That's why the user get's an error in the following issue: OpenAI.BadRequestError when using create_agent with structured output and tools with optional input parameters #34596

[codspeed-hq]

Merging this PR will not alter performance

⚠️ Unknown Walltime execution environment detected

Using the Walltime instrument on standard Hosted Runners will lead to inconsistent data.

For the most accurate results, we recommend using CodSpeed Macro Runners: bare-metal machines fine-tuned for performance measurement consistency.

Summary

✅ 13 untouched benchmarks
⏩ 21 skipped benchmarks¹

---

_{Comparing keenborder786:fix/additional_parameters (cb2faee) with master (25bb36d)}

Footnotes

1. 21 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩

[ccurme]

Can you add a test that fails on master and passes here?

On line 452 above we already apply _recursive_set_additional_properties_false

[keenborder786]

@ccurme just run the following example

import aiohttp
import asyncio
from langchain_core.messages import HumanMessage
from langchain_core.tools import StructuredTool
from langchain.agents import create_agent
from langchain.agents.structured_output import ProviderStrategy
from langchain_openai import ChatOpenAI
from pydantic import BaseModel, Field

class StructuredResult(BaseModel):
    answer: str

class SearchInput(BaseModel):
    query: str = Field(default="", description="The search query string to find relevant information.")
    limit: int = Field(default=3, description="The number of results to return.")

    model_config = {
        "extra": "forbid"
    }


def search(aiohttp_session) -> StructuredTool:
    """Search Tool"""

    async def _search(
        query: str = "",
        limit: int = 3,
    ) -> str:
        # Just echo the query
        answer = f"Echo: {query} | Limit: {limit}"
        return answer

    return StructuredTool.from_function(
        coroutine=_search,
        name="search-tool",
        description="Use this tool to search for information based on a query and an optional limit.",
        args_schema=SearchInput,
        return_direct=False,)


async def main():
    model = ChatOpenAI(
        model="gpt-4.1"
    )

    async with aiohttp.ClientSession() as session:
        # session shared across tools
        tools = [search(aiohttp_session=session)]

        agent = create_agent(
            model=model,
            tools=tools,
            system_prompt="You are a helpful assistant. Use the tools to answer user queries.",
            response_format=ProviderStrategy(schema=StructuredResult,strict=True)
        )

        input_state = {"messages": [HumanMessage(content="Search for information about AI research.")]}

        out = await agent.ainvoke(input_state)

        print(out,'out')


if __name__ == "__main__":
    asyncio.run(main())

Fails on master.