Assistants API: Unexpected `tool_call` type in `on_tool_call_created`

[ekassos]

Confirm this is an issue with the Python library and not an underlying OpenAI API

• This is an issue with the Python library

Describe the bug

When calling our implementation of method on_tool_call_created of AsyncAssistantEventHandler, tool_call is of type Dict and not of type ToolCall. In our testing, this issue occurs only when the tool call created is File Search and not Code Interpreter (we do not use Function Calling). The issue presents itself only when using AsyncOpenAI and not AsyncAzureOpenAI.

See definition of on_tool_call_created:

openai-python/src/openai/lib/streaming/_assistants.py

Lines 186 to 187 in d6bb8c1

	def on_tool_call_created(self, tool_call: ToolCall) -> None:
	"""Callback that is fired when a tool call is created"""

To Reproduce

1. Implement on_tool_call_created as

async def on_tool_call_created(self, tool_call) -> None:
    self.enqueue(
        {
            "type": "tool_call_created",
            "tool_call": tool_call.model_dump(),
        }
    )

2. When a File Search tool call is created, on_tool_call_created will fail with error 'dict' object has no attribute 'model_dump'.

Code snippets

class BufferedStreamHandler(openai.AsyncAssistantEventHandler):
    def __init__(self, file_names: dict[str, str], *args, **kwargs):
        super().__init__(*args, **kwargs)
        self.__buffer = io.BytesIO()
        self.file_names = file_names

    def enqueue(self, data: Dict) -> None:
        self.__buffer.write(orjson.dumps(data))
        self.__buffer.write(b"\n")

    def flush(self) -> bytes:
        value = self.__buffer.getvalue()
        self.__buffer.truncate(0)
        self.__buffer.seek(0)
        return value

    async def on_image_file_done(self, image_file: ImageFile) -> None:
        self.enqueue(
            {
                "type": "image_file_done",
                "file_id": image_file.file_id,
            }
        )

    async def on_message_created(self, message) -> None:
        self.enqueue(
            {
                "type": "message_created",
                "role": "assistant",
                "message": message.model_dump(),
            }
        )

    async def on_message_delta(self, delta, snapshot) -> None:
        self.enqueue(
            {
                "type": "message_delta",
                "delta": delta.model_dump(),
            }
        )

    async def on_tool_call_created(self, tool_call) -> None:
        self.enqueue(
            {
                "type": "tool_call_created",
                "tool_call": tool_call.model_dump(),
            }
        )

    async def on_tool_call_delta(self, delta, snapshot) -> None:
        self.enqueue(
            {
                "type": "tool_call_delta",
                "delta": delta.model_dump(),
            }
        )

    async def on_timeout(self) -> None:
        self.enqueue(
            {
                "type": "error",
                "detail": "Stream timed out waiting for response",
            }
        )

    async def on_done(self, run) -> None:
        self.enqueue({"type": "done"})

    async def on_exception(self, exception) -> None:
        self.enqueue(
            {
                "type": "error",
                "detail": str(exception),
            }
        )

OS

macOS

Python version

Python v3.11.10

Library version

openai v1.65.3

[gmirc12]

Found the problem, had to set a custom query param, you can close this issue

[ekassos]

Found the problem, had to set a custom query param, you can close this issue

@gmirc12 I think you’re referencing another issue: #2161, this issue remains unresolved.