Add `method_tool` Functionality

[nathanmargaglio]

I group my tools into classes, and they leverage attributes of the instantiated object to work properly. The function_tool decorator doesn't seem to support this. For example, the following works (as expected):

@function_tool
def get_random_number() -> int:
    """Get a random number."""
    return random.randint(0, 100)

agent = Agent(
    name="Assistant",
    instructions="You are a helpful assistant",
    tools=[get_random_number]
)
result = Runner.run_sync(agent, input="Get me a random number")
print(result.final_output)

But the following doesn't:

class Tools:
    @function_tool
    def get_random_number(self) -> int:
        """Get a random number."""
        return random.randint(0, 100)

tools = Tools()

agent = Agent(
    name="Assistant",
    instructions="You are a helpful assistant",
    tools=[tools.get_random_number]
)
result = Runner.run_sync(agent, input="Get me a random number")
print(result.final_output)

Running this produces an error like:

openai.BadRequestError: Error code: 400 - {'error': {'message': "Invalid schema for function 'get_random_number': In context=('properties', 'self'), schema must have a 'type' key.", 'type': 'invalid_request_error', 'param': 'tools[0].parameters', 'code': 'invalid_function_parameters'}}

i.e., the self parameter isn't accounted for.

There's probably a ton of different ways of approaching this, but I've been utilizing the class-based tool pattern quite extensively and I've found it works pretty well for keeping tooling clean and simple. I'm happy to consider alternative approaches, but if there was just a way to ignore the self parameter, then I think everything else would just work, no?

[rm-openai]

I'll take a look. For now there's a pretty simple workaround:

class Tools:
    def get_random_number(self) -> int:
        """Get a random number."""
        return random.randint(0, 100)

tools = Tools()

agent = Agent(
    name="Assistant",
    instructions="You are a helpful assistant",
    tools=[function_tool(tools.get_random_number)]
)
result = Runner.run_sync(agent, input="Get me a random number")
print(result.final_output)

[itsmesherry]

Hi @nathanmargaglio and @rm-openai, I've created pull request #137, which addresses the issue and includes tests to verify the changes. Please take a look and let me know if you have any feedback. Thanks!

[WSQsGithub]

I have the similar problem. I need to pass chat_id to some of the tool functions for searching purpose.

Currently I am doing this and it fails. I wonder if there's any better way that the tool functions can get chat_id

class Tools:
    def __init__(self, chat_uuid: str):
        self.chat_uuid = chat_uuid

    @function_tool(name_override="tell_time_tool", description_override="A tool that tells the current time.")
    async def tell_time_tool(self) -> str:
        return tell_time()


    @function_tool(
        name_override="ask_for_clarification_tool",
        description_override="A tool that asks for clarification if user's request doesn't make sense.",
    )
    async def ask_for_clarification_tool(self) -> str:
        return ask_for_clarification()

And this is how I implement the agent.

class TriageAgent(Agent):
    def __init__(self, chat_uuid: str, input: any) -> None: 
        self.chat_uuid = chat_uuid
        self.input = input
        self.tool_collections = Tools(chat_uuid=chat_uuid)

        self.tools = [self.tool_collections.tell_time_tool, self.tool_collections.ask_for_clarification_tool]
        self.handoffs = []

        self.agent = Agent(
            name="Triage Agent",
            instructions=INSTRUCTIONS,
            tools=self.tools,
            handoffs=self.handoffs,
        )


    async def run(self) -> None:
        with trace(workflow_name="Triage Handling"):
            # the chat pipeline starts with the triage agent
            result = await Runner.run(starting_agent=self.triage_agent, input=self.input)
            return result.final_output

[itsmesherry]

@WSQsGithub could you please share the error that you're getting?

[WSQsGithub]

@WSQsGithub could you please share the error that you're getting?
Here is what I'm getting. I guess it would be fixed once @Shehryar718's pr is approved.

openai.BadRequestError: Error code: 400 - {'error': {'message': "Invalid schema for function 'tell_time_tool': In context=('properties', 'self'), schema must have a 'type' key.", 'type': 'invalid_request_error', 'param': 'tools[0].parameters', 'code': 'invalid_function_parameters'}}

[itsmesherry]

@WSQsGithub could you please share the error that you're getting?
Here is what I'm getting. I guess it would be fixed once @Shehryar718's pr is approved.

openai.BadRequestError: Error code: 400 - {'error': {'message': "Invalid schema for function 'tell_time_tool': In context=('properties', 'self'), schema must have a 'type' key.", 'type': 'invalid_request_error', 'param': 'tools[0].parameters', 'code': 'invalid_function_parameters'}}

Yeah my pull request fixes it.

[amit-timalsina]

@WSQsGithub

Here is how you can fix this for now. Given that you require self in some of the tools to access chat_uuid, you can create a closure. See below for how:

class Tools:
    def __init__(self, chat_uuid: str):
        self.chat_uuid = chat_uuid

    async def tell_time_tool(self) -> str:
        @function_tool(name_override="tell_time_tool", description_override="A tool that tells the current time.")
        def tell_time():
            """Tell time"""

        return tell_time

    async def ask_for_clarification_tool(self) -> str:
        @function_tool(
            name_override="ask_for_clarification_tool",
            description_override="A tool that asks for clarification if user's request doesn't make sense.",
        )
        def ask_for_clarification()
            """Ask for Clarification"""

        return ask_for_clarification

This way tell_time and ask_for_clarification have access to self.

Then this is how you will implement agent:

class TriageAgent(Agent):
    def __init__(self, chat_uuid: str, input: any) -> None: 
        self.chat_uuid = chat_uuid
        self.input = input
        self.tool_collections = Tools(chat_uuid=chat_uuid)

        self.tools = [self.tool_collections.tell_time_tool(), self.tool_collections.ask_for_clarification_tool()]
        self.handoffs = []

        self.agent = Agent(
            name="Triage Agent",
            instructions=INSTRUCTIONS,
            tools=self.tools,
            handoffs=self.handoffs,
        )


    async def run(self) -> None:
        with trace(workflow_name="Triage Handling"):
            # the chat pipeline starts with the triage agent
            result = await Runner.run(starting_agent=self.triage_agent, input=self.input)
            return result.final_output

[mvanhorn]

I've submitted a PR to fix this: #2734

The fix skips the first parameter when it has no type annotation and is named self or cls in function_schema(). It also handles the case where RunContextWrapper/ToolContext follows self/cls as the second parameter. Includes 4 new test cases covering all the scenarios mentioned in this issue.