cool, trying to understand how context comes into play, i noticed pydantic-ai doesnt have the task abstraction like marvin/cf have, so im trying to understand how the context comes into play and gets passed around and subsequently passed to the LLM