初始化提交

yuanqi/tests/test_asr.py

@@ -0,0 +1,184 @@
+"""
+ASR 引擎和端点测试。
+"""
+
+from __future__ import annotations
+
+import io
+import wave
+from unittest.mock import patch
+
+import pytest
+
+from src.endpoints.asr import _ensure_loaded, _validate_format
+from src.engine.asr import _SUPPORTED_FORMATS, SenseVoiceASREngine
+
+
+def _make_test_wav(duration_sec: float = 0.5, sample_rate: int = 16000) -> bytes:
+    """生成测试用的 WAV 文件字节（静音音频）。"""
+    buf = io.BytesIO()
+    num_samples = int(duration_sec * sample_rate)
+    with wave.open(buf, "wb") as wf:
+        wf.setnchannels(1)
+        wf.setsampwidth(2)
+        wf.setframerate(sample_rate)
+        wf.writeframes(b"\x00\x00" * num_samples)
+    return buf.getvalue()
+
+
+# ── 测试 _SUPPORTED_FORMATS ────────────────────────────────────────────────
+
+
+class TestSupportedFormats:
+    def test_common_formats_present(self) -> None:
+        """应包含常见音频格式。"""
+        for fmt in ("wav", "mp3", "m4a", "flac", "ogg", "opus", "aac"):
+            assert fmt in _SUPPORTED_FORMATS, f"缺少常用格式: {fmt}"
+
+
+# ── 测试 _validate_format ──────────────────────────────────────────────────
+
+
+class TestValidateFormat:
+    def test_supported_format(self) -> None:
+        """支持的格式返回扩展名。"""
+        assert _validate_format("audio.wav") == "wav"
+        assert _validate_format("speech.mp3") == "mp3"
+
+    @pytest.mark.asyncio
+    async def test_no_extension_raises(self) -> None:
+        """无扩展名应抛出异常。"""
+        from fastapi import HTTPException
+
+        with pytest.raises(HTTPException) as exc:
+            _validate_format("no_ext")
+        assert exc.value.status_code == 400
+        assert str(exc.value.detail["code"]) == "unsupported_format"
+
+    @pytest.mark.asyncio
+    async def test_unsupported_format(self) -> None:
+        """不支持的格式应抛出异常。"""
+        from fastapi import HTTPException
+
+        with pytest.raises(HTTPException) as exc:
+            _validate_format("video.avi")
+        assert exc.value.status_code == 400
+        assert str(exc.value.detail["code"]) == "unsupported_format"
+
+
+# ── 测试 SenseVoiceASREngine ──────────────────────────────────────────────
+
+
+class TestSenseVoiceASREngine:
+    def test_init(self) -> None:
+        """引擎初始化应使用默认配置。"""
+        engine = SenseVoiceASREngine()
+        assert engine.name == "asr"
+        assert engine.model_loaded is False
+        assert engine._model is None
+
+    @pytest.mark.asyncio
+    async def test_health_unloaded(self) -> None:
+        """模型未加载时 health 应返回 unavailable。"""
+        engine = SenseVoiceASREngine()
+        h = await engine.health()
+        assert h["status"] == "unavailable"
+
+    @pytest.mark.asyncio
+    async def test_load_without_funasr(self) -> None:
+        """未安装 funasr 时加载应优雅失败。"""
+        engine = SenseVoiceASREngine()
+        # 模拟 funasr 不可用
+        with patch.dict("sys.modules", {"funasr": None}):
+            await engine.load()
+        assert engine.model_loaded is False
+
+    @pytest.mark.asyncio
+    async def test_transcribe_without_model(self) -> None:
+        """模型未加载时 transcribe 返回空结果。"""
+        engine = SenseVoiceASREngine()
+        result = await engine.transcribe(b"fake-audio", format="wav")
+        assert result["text"] == ""
+        assert result["model"] == engine._model_name
+
+    @pytest.mark.asyncio
+    async def test_normalize_language(self) -> None:
+        """语言归一化应正确映射。"""
+        engine = SenseVoiceASREngine()
+        assert engine._normalize_language(None) == "auto"
+        assert engine._normalize_language("") == "auto"
+        assert engine._normalize_language("auto") == "auto"
+        assert engine._normalize_language("zh") == "zh"
+        assert engine._normalize_language("en") == "en"
+        assert engine._normalize_language("ja") == "ja"
+        assert engine._normalize_language("ko") == "ko"
+        assert engine._normalize_language("yue") == "yue"
+        # 不支持的应回退 auto
+        assert engine._normalize_language("zz") == "auto"
+
+    @pytest.mark.asyncio
+    async def test_unload(self) -> None:
+        """卸载应清理模型引用。"""
+        engine = SenseVoiceASREngine()
+        await engine.unload()
+        assert engine.model_loaded is False
+        assert engine._model is None
+
+    @pytest.mark.asyncio
+    async def test_close(self) -> None:
+        """close 应调用 unload。"""
+        engine = SenseVoiceASREngine()
+        await engine.close()
+        assert engine.model_loaded is False
+
+
+# ── 测试端点逻辑 ──────────────────────────────────────────────────────────
+
+
+class TestEndpoints:
+    def test_ensure_loaded_returns_503_when_engine_none(self) -> None:
+        """引擎未初始化时返回 503。"""
+        with pytest.raises(Exception):
+            # 模拟 asr_engine 为 None 的场景
+            with patch("src.endpoints.asr.asr_engine", None):
+                _ensure_loaded()
+
+    @pytest.mark.asyncio
+    async def test_asr_endpoint_no_engine(self) -> None:
+        """ASR 引擎未就绪时应有适当的错误响应状态码。"""
+        from fastapi import HTTPException
+
+        with patch("src.endpoints.asr.asr_engine", None):
+            with pytest.raises(HTTPException) as exc:
+                _ensure_loaded()
+            assert exc.value.status_code == 503
+
+
+# ── 集成测试（需要实际模型，标记为跳过） ────────────────────────
+
+
+@pytest.mark.skip(reason="需要真实 SenseVoiceSmall 模型和 FunASR 环境")
+class TestIntegrationASR:
+    """集成测试 — 需要实际模型文件。"""
+
+    @pytest.mark.asyncio
+    async def test_transcribe_wav(self) -> None:
+        """转写 WAV 音频。"""
+
+        engine = SenseVoiceASREngine()
+        await engine.load()
+        assert engine.model_loaded
+
+        audio = _make_test_wav()
+        result = await engine.transcribe(audio, format="wav")
+        assert isinstance(result["text"], str)
+
+    @pytest.mark.asyncio
+    async def test_transcribe_with_language_hint(self) -> None:
+        """指定语言提示的转写。"""
+        engine = SenseVoiceASREngine()
+        await engine.load()
+
+        audio = _make_test_wav()
+        result = await engine.transcribe(audio, format="wav", language="zh")
+        assert isinstance(result["text"], str)

yuanqi/tests/test_health.py

@@ -0,0 +1,25 @@
+"""元炁健康检查端点测试。"""
+
+from __future__ import annotations
+
+from fastapi import status
+from fastapi.testclient import TestClient
+
+from src.main import app
+
+client = TestClient(app)
+
+
+def test_healthz_returns_200() -> None:
+    """健康检查端点应始终返回 200（即使在未加载模型时也返回可用状态信息）。"""
+    resp = client.get("/healthz")
+    assert resp.status_code == status.HTTP_200_OK
+
+    data = resp.json()
+    assert data["status"] in ("ok", "degraded", "error")
+    assert "engines" in data
+    assert "llm" in data["engines"]
+    assert "embedding" in data["engines"]
+    assert "reranker" in data["engines"]
+    assert "version" in data
+    assert isinstance(data["loaded_models"], list)

yuanqi/tests/test_rerank.py

@@ -0,0 +1,37 @@
+"""元炁 Reranker 端点测试。"""
+
+from __future__ import annotations
+
+from fastapi import status
+from fastapi.testclient import TestClient
+
+from src.main import app
+
+client = TestClient(app)
+
+
+def test_rerank_without_model_returns_503() -> None:
+    """Reranker 端点未加载模型时应返回 503。"""
+    resp = client.post(
+        "/v1/rerank",
+        json={
+            "model": "test-model",
+            "query": "测试查询",
+            "documents": ["文档A", "文档B", "文档C"],
+        },
+    )
+    # 未加载模型时返回 503；安装后若模型就绪则返回 200
+    assert resp.status_code in (
+        status.HTTP_200_OK,
+        status.HTTP_503_SERVICE_UNAVAILABLE,
+    )
+
+
+def test_rerank_invalid_body() -> None:
+    """Reranker 端点应对无效请求体返回 422。"""
+    resp = client.post(
+        "/v1/rerank",
+        json={"model": "test", "query": "test"},
+        # 缺少 documents 字段
+    )
+    assert resp.status_code == status.HTTP_422_UNPROCESSABLE_ENTITY