prompt is too long

[atopheim]

What is the best way to handle this when summarizing very long inputs?

Is there a way to run llama.cpp just to get the token count?
If so I can chunk into batches, but pointless if the tooling does this already and there's just a flag I haven't found yet.
Command in python

with subprocess.Popen(
            [
                LLAMA_CPP_EXECUTABLE_PATH,
                "-m",
                LLAMA_CPP_MODEL_PATH,
                "-t",
                "8",
                "--ctx-size",
                "4096",
                "-p",
                prompt,
            ],
            stdout=subprocess.PIPE,
            stderr=subprocess.PIPE,
            text=True,
            bufsize=1, 
        )

Feel like pushing --ctx-size over 4096 will take too long and be not beneficial anyways.