"""Tests for local token/cache mode benchmark harness.""" from benchmarks.proxy_mode_benchmark import run_local_benchmark def test_local_mode_benchmark_shows_compression_and_cache_tradeoff() -> None: results = run_local_benchmark(turns=6) baseline = results["baseline"] token = results["token"] cache = results["cache"] assert token.total_tokens_saved > 0 assert cache.total_tokens_saved > 0 assert token.total_sent_tokens < baseline.total_sent_tokens assert cache.total_sent_tokens < baseline.total_sent_tokens # Cache mode should preserve prefix better than token mode. assert cache.total_cache_read_tokens >= token.total_cache_read_tokens