Reusing the cached prefix of a prompt across calls to dramatically reduce input token cost when the same context is sent repeatedly.
Solutions