Timeout

Timeout limits how long a wrapped call may run. If the deadline expires, it cancels the call and raises TimeoutError instead of allowing a slow node or model request to halt the workflow indefinitely.

import railtracks as rt
from railtracks.prebuilt.middleware import Timeout

TimedAgent = rt.agent_node(
    name="timeout-demo",
    llm=rt.llm.OpenAILLM("gpt-4o"),
    middleware=[Timeout(seconds=30)],
)

Because it only controls the wrapped call, Timeout works in both middleware= and model_middleware=. The deadline covers the complete call in the selected slot.