How can I ensure that the context passed to a large language model using Tavily results does not exceed token limits?

How can I ensure that the context passed to a large language model using Tavily results does not exceed token limits during create langgraphic?