Fix/consistent generic type handling

[HezziCode]

Fix: Preserve Generic Type Information in ToolContext

Problem

There was an inconsistency in how RunContextWrapper[T] generics were handled between dynamic instructions and tool validation functions:

• Dynamic Instructions: ctx.context.age worked correctly ✅
• Tool Validation: ctx.context.age threw AttributeError: 'dict' object has no attribute 'age' ❌

This forced developers to manually convert context using UserDataType.model_validate(ctx.context) in tool validation functions.

Root Cause

The issue was in ToolContext.from_agent_context() method which used field-by-field copying:

# Old problematic code
for f in fields(RunContextWrapper):
    setattr(tool_context, f.name, getattr(context, f.name))



# New fixed code
return cls(
    context=context.context,  # Direct assignment preserves type
    usage=context.usage,
    tool_name=tool_name,
    tool_call_id=tool_call_id,
)



# Before (tool validation)
async def tool_validation(ctx: RunContextWrapper[UserDataType], agent):
    user_data = UserDataType.model_validate(ctx.context)  # Manual conversion needed
    if user_data.age >= 18:
        return True

# After (tool validation) 
async def tool_validation(ctx: RunContextWrapper[UserDataType], agent):
    if ctx.context.age >= 18:  # Direct access works now!
        return True

[seratch]

I don't think this change is necessary. The following code works without any issues:

import asyncio
from pydantic import BaseModel
from agents import Agent, RunContextWrapper, Runner, function_tool

class MyContext(BaseModel):
    foo: str

@function_tool
def get_weather(cxt: RunContextWrapper[MyContext], city: str) -> str:
    print(cxt.context.foo)
    return "Sunny with wind."

agent = Agent(
    name="Hello world",
    instructions="You are a helpful agent. You must call the tools.",
    tools=[get_weather],
)

async def main():
    result = await Runner.run(
        agent, input="What's the weather in Tokyo?", context=MyContext(foo="FOO")
    )
    print(result.final_output)

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