from adam.ollama import OllamaClient


def test_qwen3_chat_has_a_larger_reply_budget_for_reasoning() -> None:
    client = OllamaClient("http://localhost:11434", "qwen3:4b")

    assert client._chat_system("You are ADAM.") == "You are ADAM."
    assert client._num_predict(180) == 1024


def test_other_models_keep_their_original_system_prompt() -> None:
    client = OllamaClient("http://localhost:11434", "qwen2.5:1.5b")

    assert client._chat_system("You are ADAM.") == "You are ADAM."
    assert client._num_predict(180) == 180


def test_configured_chat_limit_overrides_the_model_default() -> None:
    client = OllamaClient("http://localhost:11434", "qwen3:4b", chat_max_tokens=1536)

    assert client._num_predict(180) == 1536
    assert client._num_predict(300, chat=False) == 1024