Clarification on the need for RunContextWrapper[UserContext] and Agent[UserContext]

[qudratullah0708]

I noticed that Agent[UserContext] uses generics, but in the current implementation, it always works with UserContext.

Questions for clarification:

Are there existing or planned use cases where the agent works with other types of contexts (e.g., AdminContext, CustomerContext)?
If UserContext is the only expected type, would it make sense to simplify this by removing generics?
If generics are needed for extensibility, could we add documentation/comments to clarify this?

https://github.com/openai/openai-agents-python/blob/main/docs/agents.md

Understanding the reasoning behind this design could help contributors navigate and extend the codebase more easily. Looking forward to your thoughts!

class UserContext:
def init(self, name, role):
self.name = name
self.role = role # Example: "admin" or "guest"

Now, instead of RunContextWrapper, we can just pass the user directly:

class Agent:
def init(self, name, instructions):
self.name = name
self.instructions = instructions # This will be a function

Dynamic instructions function

def dynamic_instructions(user: UserContext, agent: Agent):
return f"The user's name is {user.name}. They are a {user.role}. Assist them accordingly."

Creating a user

user = UserContext(name="Alice", role="admin")

Creating an agent and assigning the dynamic function

agent = Agent(name="Support Agent", instructions=dynamic_instructions)

Getting dynamic instructions

instructions_text = dynamic_instructions(user, agent)
print(instructions_text)

[rm-openai]

The rationale is that every tool, callback, hook etc in a Runner.run() call should use the same context type. Generics are a way to enforce that during type checking. I think we didn't do a great job enforcing that, but let me know if the idea makes sense?