from fam_edge.model_adapters.ollama_adapter import OllamaAdapter


def _cfg(**overrides):
    base = {
        "provider": "ollama",
        "role": "text",
        "model_name": "qwen2.5:7b",
        "base_url": "http://localhost:11434",
        "circuit_breaker": {"enabled": False},
    }
    base.update(overrides)
    return base


class _FakeResp:
    def __init__(self, status_code=200, text=""):
        self.status_code = status_code
        self.text = text

    def json(self):
        return {"response": "ok"}


def test_warm_up_success(monkeypatch):
    calls = {}

    def fake_post(url, json=None, timeout=None):
        calls["url"] = url
        calls["json"] = json
        calls["timeout"] = timeout
        return _FakeResp(200)
    monkeypatch.setattr(
        "fam_edge.model_adapters.ollama_adapter.requests.post", fake_post)
    a = OllamaAdapter(_cfg())
    assert a.warm_up() is True
    assert calls["url"] == "http://localhost:11434/api/generate"
    assert calls["json"]["model"] == "qwen2.5:7b"
    # 只为触发加载，不需要真的生成长文本
    assert calls["json"]["options"]["num_predict"] == 1


def test_warm_up_http_error_returns_false(monkeypatch):
    monkeypatch.setattr(
        "fam_edge.model_adapters.ollama_adapter.requests.post",
        lambda url, json=None, timeout=None: _FakeResp(500, "boom"))
    a = OllamaAdapter(_cfg())
    assert a.warm_up() is False


def test_warm_up_exception_does_not_raise(monkeypatch):
    """核心诉求: 预热失败(比如 Ollama 服务当时没起来)不能抛异常影响主服务
    启动，只应该记警告日志、返回 False。"""
    def raise_err(url, json=None, timeout=None):
        raise ConnectionError("refused")
    monkeypatch.setattr(
        "fam_edge.model_adapters.ollama_adapter.requests.post", raise_err)
    a = OllamaAdapter(_cfg())
    assert a.warm_up() is False


def test_warm_up_uses_generous_timeout_for_cold_start(monkeypatch):
    """核心诉求: 冷启动实测能到 1-2 分钟，预热请求的超时不能沿用问答的短超时。"""
    captured = {}

    def fake_post(url, json=None, timeout=None):
        captured["timeout"] = timeout
        return _FakeResp(200)
    monkeypatch.setattr(
        "fam_edge.model_adapters.ollama_adapter.requests.post", fake_post)
    a = OllamaAdapter(_cfg(timeout=20))  # chat() 用的短超时
    a.warm_up()
    assert captured["timeout"] >= 120