Process History
ProcessHistory is a capability that wraps a history processor: a function that receives the message history before each model request and returns the (possibly modified) list of messages to send. Use it to trim old turns, redact sensitive content, or summarize long conversations:
from pydantic_ai import Agent
from pydantic_ai.capabilities import ProcessHistory
from pydantic_ai.messages import ModelMessage
def keep_recent(messages: list[ModelMessage]) -> list[ModelMessage]:
return messages[-5:] # (1)
agent = Agent('openai:gpt-5.2', capabilities=[ProcessHistory(keep_recent)])Keep only the five most recent messages. In practice you'll want to keep the first request too, so the system prompt survives — see Processing Message History for complete patterns.
The processor may be sync or async, and may optionally take a RunContext as its first argument to access dependencies and run state. Multiple ProcessHistory capabilities apply in registration order. Note that the processed messages replace the run’s message history, so make a copy first if you need to keep the original.
ProcessHistory is a thin wrapper around the before_model_request lifecycle hook — hook that event directly for richer control, like short-circuiting the model call. See Processing Message History for the full guide, including summarization examples and interactions with new_messages().