Providing a pydantic model instead of docstring for tool parameters.

[HG2407]

Please read this first

• Have you read the docs?Agents SDK docs
• Have you searched for related issues? Others may have had similar requests

Question

@rm-openai What if I want some parameters to be required and others to be optional, I can easily do this via pydantic, where any field that I have not wrapped in an optional object will be required. Is there a way to replicate this here ? Because as far as I understand, if I do strict_mode = False, while low, there's a chance that llm might not pass anything, I want to avoid that.

Secondly, suppose my tool calls an api, which is linked to the backend, now I want to pass an auth token to the api, which I currently do by adding it as a parameter in each tool and asking agent to provide it through its system prompt, is there a better way to do it ?

Edit: I just checked there's something called function_schema that you use to extract the parameters and description from the docstring and then convert them to pydantic model. is there a way to directly provide a pydantic model to this ?

[pakrym-oai]

All parameters must be required for strict mode but the can be nullable.

Secondly, suppose my tool calls an api, which is linked to the backend, now I want to pass an auth token to the api, which I currently do by adding it as a parameter in each tool and asking agent to provide it through its system prompt, is there a better way to do it ?

Use agent context for this. Context is available to tools but never leaves the process.

openai-agents-python/examples/customer_service/main.py

Line 61 in f976349

context: RunContextWrapper[AirlineAgentContext], confirmation_number: str, new_seat: str

[HG2407]

hey @pakrym-oai thanks for the reply. I have another doubt.

 async def main():
    current_agent: Agent[AirlineAgentContext] = triage_agent
    input_items: list[TResponseInputItem] = []
    context = AirlineAgentContext()

    # Normally, each input from the user would be an API request to your app, and you can wrap the request in a trace()
    # Here, we'll just use a random UUID for the conversation ID
    conversation_id = uuid.uuid4().hex[:16]

    while True:
        user_input = input("Enter your message: ")
        with trace("Customer service", group_id=conversation_id):
            input_items.append({"content": user_input, "role": "user"})
            result = await Runner.run(current_agent, input_items, context=context)

            for new_item in result.new_items:
                agent_name = new_item.agent.name
                if isinstance(new_item, MessageOutputItem):
                    print(f"{agent_name}: {ItemHelpers.text_message_output(new_item)}")
                elif isinstance(new_item, HandoffOutputItem):
                    print(
                        f"Handed off from {new_item.source_agent.name} to {new_item.target_agent.name}"
                    )
                elif isinstance(new_item, ToolCallItem):
                    print(f"{agent_name}: Calling a tool")
                elif isinstance(new_item, ToolCallOutputItem):
                    print(f"{agent_name}: Tool call output: {new_item.output}")
                else:
                    print(f"{agent_name}: Skipping item: {new_item.__class__.__name__}")
            input_items = result.to_input_list()
            current_agent = result.last_agent

suppose here, when I get the chat history through input_items, I also want to see which particular agent sent a particular message. Is there a way to do that ? I just need it for debugging purposes.

P.S.: I also want to use handoff_filters.remove_all_tools to filter out the tool calls.

input_items = result.to_input_list()
filtered_input_items = handoff_filters.remove_all_tools(input_items)

But I get this error:
AttributeError: 'list' object has no attribute 'input_history'