Skip to content

armgit / ODS

Public Imported from GitHub
0

ODS V3 Pre-Release: Public testing and refinement ahead of the official V3 launch. Turn your PC, Mac, or Linux box into a private AI server.

9.1 KB
Raw Blame
"""Bark TTS API Server Tests"""

import pytest
import base64
import threading
import numpy as np
from unittest.mock import patch
from fastapi.testclient import TestClient

# Import the module under test
import server

client = TestClient(server.app)


# Fixtures
@pytest.fixture(autouse=True)
def reset_globals():
    """Reset global state before each test."""
    # Use setattr to modify the module's attribute, not a local variable
    original = server._models_loaded
    server._models_loaded = False
    yield
    server._models_loaded = original


@pytest.fixture
def mock_bark_generate_audio():
    """Mock bark.generate_audio to return a simple audio array."""
    with patch("bark.generate_audio") as mock:
        # Generate a simple 1-second audio array at 24kHz (numpy array, not list)
        mock.return_value = np.array([0.1] * 24000, dtype=np.float32)
        yield mock


@pytest.fixture
def mock_bark_preload_models():
    """Mock bark.preload_models."""
    with patch("bark.preload_models") as mock:
        yield mock


@pytest.fixture
def mock_soundfile_write():
    """Mock soundfile.write."""
    with patch("server.sf.write") as mock:
        # Make it write to the provided buffer
        def side_effect(buf, audio_array, sample_rate, format=None):
            # Simulate writing by just seeking to end
            buf.seek(0)
            buf.write(b'\x00' * 100)  # fake WAV data
            buf.seek(0)
        mock.side_effect = side_effect
        yield mock


@pytest.fixture
def mock_soundfile_read():
    """Mock soundfile.read for stream endpoint."""
    with patch("server.sf.read") as mock:
        mock.return_value = (np.array([0.1] * 24000, dtype=np.float32), 24000)
        yield mock


# Tests for /health endpoint
def test_health_initial():
    """Test health endpoint before models are loaded."""
    response = client.get("/health")
    assert response.status_code == 200
    data = response.json()
    assert data["status"] == "ok"
    assert data["models_loaded"] is False


def test_health_after_load(mock_bark_preload_models):
    """Test health endpoint after models are loaded."""
    # Trigger model loading
    server._load_models()
    response = client.get("/health")
    assert response.status_code == 200
    data = response.json()
    assert data["status"] == "ok"
    assert data["models_loaded"] is True


# Tests for /tts endpoint
def test_tts_success(mock_bark_generate_audio, mock_soundfile_write):
    """Test successful TTS request."""
    with patch("server._models_loaded", True):
        response = client.post("/tts", json={
            "text": "Hello, world!",
            "voice_preset": "v2/en_speaker_6",
            "output_format": "wav"
        })
        assert response.status_code == 200
        data = response.json()
        assert "audio_base64" in data
        assert data["sample_rate"] == 24000
        assert data["format"] == "wav"
        assert base64.b64decode(data["audio_base64"])


def test_tts_default_format(mock_bark_generate_audio, mock_soundfile_write):
    """Test TTS with default format (wav)."""
    with patch("server._models_loaded", True):
        response = client.post("/tts", json={
            "text": "Hello, world!",
        })
        assert response.status_code == 200
        data = response.json()
        assert data["format"] == "wav"


def test_tts_case_insensitive_format(mock_bark_generate_audio, mock_soundfile_write):
    """Test TTS with lowercase format."""
    with patch("server._models_loaded", True):
        response = client.post("/tts", json={
            "text": "Hello, world!",
            "output_format": "mp3"
        })
        assert response.status_code == 200
        data = response.json()
        assert data["format"] == "mp3"


def test_tts_invalid_format():
    """Test TTS with invalid format."""
    response = client.post("/tts", json={
        "text": "Hello, world!",
        "output_format": "avi"
    })
    assert response.status_code == 422
    assert "output_format" in response.json()["detail"].lower()


def test_tts_text_too_long():
    """Test TTS with text exceeding MAX_TEXT_LENGTH."""
    long_text = "a" * (server.MAX_TEXT_LENGTH + 1)
    response = client.post("/tts", json={
        "text": long_text
    })
    assert response.status_code == 422
    assert "text" in response.json()["detail"].lower()


def test_tts_text_empty():
    """Test TTS with empty text."""
    response = client.post("/tts", json={
        "text": ""
    })
    assert response.status_code == 200  # Empty text is allowed by Pydantic


def test_tts_model_loading_on_first_request(mock_bark_preload_models, mock_bark_generate_audio, mock_soundfile_write):
    """Test that models are loaded on first request."""
    # Ensure models are not loaded
    with patch("server._models_loaded", False):
        response = client.post("/tts", json={
            "text": "Hello, world!",
        })
        assert response.status_code == 200
        # Verify preload_models was called
        mock_bark_preload_models.assert_called_once()


def test_tts_concurrent_requests(mock_bark_preload_models, mock_bark_generate_audio, mock_soundfile_write):
    """Test concurrent TTS requests."""
    with patch("server._models_loaded", False):
        # Make multiple concurrent requests
        threads = []
        results = []

        def make_request():
            try:
                resp = client.post("/tts", json={"text": "Hello!"})
                results.append(resp.status_code)
            except Exception as e:
                results.append(str(e))

        for _ in range(3):
            t = threading.Thread(target=make_request)
            threads.append(t)
            t.start()

        for t in threads:
            t.join()

        # All requests should succeed
        assert all(code == 200 for code in results)
        # preload_models should be called exactly once
        assert mock_bark_preload_models.call_count == 1


# Tests for /tts/stream endpoint
def test_tts_stream_success(mock_bark_generate_audio, mock_soundfile_write):
    """Test successful TTS stream request."""
    with patch("server._models_loaded", True):
        response = client.post("/tts/stream", json={
            "text": "Hello, world!",
            "voice_preset": "v2/en_speaker_6"
        })
        assert response.status_code == 200
        assert response.headers["content-type"] == "audio/wav"
        assert "bark_output.wav" in response.headers["content-disposition"]
        assert len(response.content) > 0


def test_tts_stream_default_format(mock_bark_generate_audio, mock_soundfile_write):
    """Test TTS stream always returns WAV regardless of format."""
    with patch("server._models_loaded", True):
        response = client.post("/tts/stream", json={
            "text": "Hello, world!",
            "output_format": "mp3"  # This should be ignored for stream endpoint
        })
        assert response.status_code == 200
        assert response.headers["content-type"] == "audio/wav"


# Tests for validation
def test_tts_invalid_voice_preset():
    """Test TTS with invalid voice preset (now rejected by server validator)."""
    response = client.post("/tts", json={
        "text": "Hello, world!",
        "voice_preset": "invalid_preset"
    })
    assert response.status_code == 422
    assert "voice_preset" in response.json()["detail"].lower()


def test_tts_text_max_length_boundary():
    """Test TTS with text at MAX_TEXT_LENGTH boundary."""
    text = "a" * server.MAX_TEXT_LENGTH
    response = client.post("/tts", json={
        "text": text
    })
    assert response.status_code == 200


# Tests for error handling
def test_tts_generation_error(mock_bark_generate_audio):
    """Test TTS when bark.generate_audio raises an exception."""
    mock_bark_generate_audio.side_effect = Exception("Bark error")
    with patch("server._models_loaded", True):
        response = client.post("/tts", json={
            "text": "Hello, world!",
        })
        assert response.status_code == 500
        assert "TTS generation failed" in response.json()["detail"]


def test_tts_stream_generation_error(mock_bark_generate_audio):
    """Test TTS stream when bark.generate_audio raises an exception."""
    mock_bark_generate_audio.side_effect = Exception("Bark error")
    with patch("server._models_loaded", True):
        response = client.post("/tts/stream", json={
            "text": "Hello, world!",
        })
        assert response.status_code == 500
        assert "TTS generation failed" in response.json()["detail"]


def test_load_models_thread_safety(mock_bark_preload_models):
    """Test that model loading is thread-safe."""
    # Reset global state
    server._models_loaded = False

    results = []
    errors = []

    def load_and_check():
        try:
            server._load_models()
            results.append(server._models_loaded)
        except Exception as e:
            errors.append(e)

    threads = [threading.Thread(target=load_and_check) for _ in range(10)]
    for t in threads:
        t.start()
    for t in threads:
        t.join()

    assert len(errors) == 0
    assert len(results) == 10
    assert all(results)
    # preload_models should only be called once due to lock
    assert mock_bark_preload_models.call_count == 1