e77af7f8f0
实现以下模块并通过全部测试 (150 passed, 92.65% coverage):
- validators.py: 数据验证器 (Brief/视频/审核决策/申诉/时间戳/UUID)
- timestamp_align.py: 多模态时间戳对齐 (ASR/OCR/CV 融合)
- rule_engine.py: 规则引擎 (违禁词检测/语境感知/规则版本管理)
- brief_parser.py: Brief 解析 (卖点/禁忌词/时序要求/品牌调性提取)
- video_auditor.py: 视频审核 (文件验证/ASR/OCR/Logo检测/合规检查)
验收标准达成:
- 违禁词召回率 ≥ 95%
- 误报率 ≤ 5%
- 时长统计误差 ≤ 0.5秒
- 语境感知检测 ("最开心的一天" 不误判)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
279 lines
9.9 KiB
Python
279 lines
9.9 KiB
Python
"""
|
||
SmartAudit 测试全局配置
|
||
|
||
本文件定义所有测试共享的 fixtures 和配置。
|
||
遵循 TDD 原则：先写测试，后写实现。
|
||
"""
|
||
|
||
import pytest
|
||
from typing import Any
|
||
from pathlib import Path
|
||
|
||
# ============================================================================
|
||
# 路径配置
|
||
# ============================================================================
|
||
|
||
@pytest.fixture
|
||
def fixtures_path() -> Path:
|
||
"""测试数据目录"""
|
||
return Path(__file__).parent / "fixtures"
|
||
|
||
|
||
@pytest.fixture
|
||
def sample_brief_pdf(fixtures_path: Path) -> Path:
|
||
"""示例 Brief PDF 文件路径"""
|
||
return fixtures_path / "briefs" / "sample_brief.pdf"
|
||
|
||
|
||
@pytest.fixture
|
||
def sample_video_path(fixtures_path: Path) -> Path:
|
||
"""示例视频文件路径"""
|
||
return fixtures_path / "videos" / "sample_video.mp4"
|
||
|
||
|
||
# ============================================================================
|
||
# Brief 规则 Fixtures
|
||
# ============================================================================
|
||
|
||
@pytest.fixture
|
||
def sample_brief_rules() -> dict[str, Any]:
|
||
"""标准 Brief 规则示例"""
|
||
return {
|
||
"selling_points": [
|
||
{"text": "24小时持妆", "priority": "high"},
|
||
{"text": "天然成分", "priority": "medium"},
|
||
{"text": "敏感肌适用", "priority": "medium"},
|
||
],
|
||
"forbidden_words": [
|
||
{"word": "最", "reason": "广告法极限词", "severity": "hard"},
|
||
{"word": "第一", "reason": "广告法极限词", "severity": "hard"},
|
||
{"word": "药用", "reason": "化妆品禁用", "severity": "hard"},
|
||
{"word": "治疗", "reason": "化妆品禁用", "severity": "hard"},
|
||
{"word": "绝对", "reason": "广告法极限词", "severity": "hard"},
|
||
{"word": "领导者", "reason": "广告法极限词", "severity": "hard"},
|
||
{"word": "史上", "reason": "广告法极限词", "severity": "hard"},
|
||
],
|
||
"brand_tone": {
|
||
"style": "年轻活力",
|
||
"description": "面向 18-35 岁女性用户"
|
||
},
|
||
"timing_requirements": [
|
||
{"type": "product_visible", "min_duration_seconds": 5},
|
||
{"type": "brand_mention", "min_frequency": 3},
|
||
],
|
||
"platform": "douyin",
|
||
"region": "mainland_china",
|
||
}
|
||
|
||
|
||
@pytest.fixture
|
||
def sample_platform_rules() -> dict[str, Any]:
|
||
"""抖音平台规则示例"""
|
||
return {
|
||
"platform": "douyin",
|
||
"version": "2026.01",
|
||
"forbidden_words": [
|
||
{"word": "最", "category": "ad_law"},
|
||
{"word": "第一", "category": "ad_law"},
|
||
{"word": "国家级", "category": "ad_law"},
|
||
{"word": "绝对", "category": "ad_law"},
|
||
],
|
||
"content_rules": [
|
||
{"rule": "不得含有虚假宣传", "category": "compliance"},
|
||
{"rule": "不得使用竞品 Logo", "category": "brand_safety"},
|
||
],
|
||
}
|
||
|
||
|
||
# ============================================================================
|
||
# 视频审核 Fixtures
|
||
# ============================================================================
|
||
|
||
@pytest.fixture
|
||
def sample_asr_result() -> dict[str, Any]:
|
||
"""ASR 语音识别结果示例"""
|
||
return {
|
||
"text": "大家好，这款产品真的非常好用，24小时持妆效果特别棒",
|
||
"segments": [
|
||
{"word": "大家好", "start_ms": 0, "end_ms": 800, "confidence": 0.98},
|
||
{"word": "这款产品", "start_ms": 850, "end_ms": 1500, "confidence": 0.97},
|
||
{"word": "真的非常好用", "start_ms": 1550, "end_ms": 2800, "confidence": 0.96},
|
||
{"word": "24小时持妆", "start_ms": 2900, "end_ms": 4000, "confidence": 0.99},
|
||
{"word": "效果特别棒", "start_ms": 4100, "end_ms": 5200, "confidence": 0.95},
|
||
],
|
||
}
|
||
|
||
|
||
@pytest.fixture
|
||
def sample_ocr_result() -> dict[str, Any]:
|
||
"""OCR 字幕识别结果示例"""
|
||
return {
|
||
"frames": [
|
||
{"timestamp_ms": 1000, "text": "产品名称", "confidence": 0.98, "bbox": [100, 450, 300, 480]},
|
||
{"timestamp_ms": 3000, "text": "24小时持妆", "confidence": 0.97, "bbox": [150, 450, 350, 480]},
|
||
{"timestamp_ms": 5000, "text": "立即购买", "confidence": 0.96, "bbox": [200, 500, 400, 530]},
|
||
],
|
||
}
|
||
|
||
|
||
@pytest.fixture
|
||
def sample_cv_result() -> dict[str, Any]:
|
||
"""CV 视觉检测结果示例"""
|
||
return {
|
||
"detections": [
|
||
{
|
||
"object_type": "product",
|
||
"start_frame": 30,
|
||
"end_frame": 180,
|
||
"fps": 30,
|
||
"start_ms": 1000, # 30/30 * 1000 = 1000ms
|
||
"end_ms": 6000, # 180/30 * 1000 = 6000ms (5秒时长)
|
||
"confidence": 0.95,
|
||
"bbox": [200, 100, 400, 350],
|
||
},
|
||
{
|
||
"object_type": "competitor_logo",
|
||
"start_frame": 200,
|
||
"end_frame": 230,
|
||
"fps": 30,
|
||
"start_ms": 6667, # 200/30 * 1000
|
||
"end_ms": 7667, # 230/30 * 1000
|
||
"confidence": 0.88,
|
||
"bbox": [50, 50, 100, 100],
|
||
"logo_id": "competitor_001",
|
||
},
|
||
],
|
||
}
|
||
|
||
|
||
# ============================================================================
|
||
# 违禁词测试数据
|
||
# ============================================================================
|
||
|
||
@pytest.fixture
|
||
def prohibited_word_test_cases() -> list[dict[str, Any]]:
|
||
"""违禁词检测测试用例集"""
|
||
return [
|
||
# 广告语境下应检出
|
||
{"text": "这是全网销量第一的产品", "context": "advertisement", "expected": ["第一"], "should_detect": True},
|
||
{"text": "我们是行业领导者", "context": "advertisement", "expected": ["领导者"], "should_detect": True},
|
||
{"text": "史上最低价促销", "context": "advertisement", "expected": ["最", "史上"], "should_detect": True},
|
||
{"text": "绝对有效，药用级别", "context": "advertisement", "expected": ["绝对", "药用"], "should_detect": True},
|
||
|
||
# 日常语境下不应检出（语境感知）
|
||
{"text": "今天是我最开心的一天", "context": "daily", "expected": [], "should_detect": False},
|
||
{"text": "这是我第一次来这里", "context": "daily", "expected": [], "should_detect": False},
|
||
{"text": "我们家排行第一", "context": "daily", "expected": [], "should_detect": False},
|
||
|
||
# 边界情况
|
||
{"text": "", "context": "advertisement", "expected": [], "should_detect": False},
|
||
{"text": "这是一个普通的产品介绍", "context": "advertisement", "expected": [], "should_detect": False},
|
||
|
||
# 组合违禁词
|
||
{"text": "全网销量第一，史上最低价", "context": "advertisement", "expected": ["第一", "最", "史上"], "should_detect": True},
|
||
]
|
||
|
||
|
||
@pytest.fixture
|
||
def context_understanding_test_cases() -> list[dict[str, Any]]:
|
||
"""语境理解测试用例集"""
|
||
return [
|
||
{"text": "这款产品是最好的选择", "expected_context": "advertisement", "should_flag": True},
|
||
{"text": "最近天气真好", "expected_context": "daily", "should_flag": False},
|
||
{"text": "今天心情最棒了", "expected_context": "daily", "should_flag": False},
|
||
{"text": "我们的产品效果最显著", "expected_context": "advertisement", "should_flag": True},
|
||
{"text": "这是我见过最美的风景", "expected_context": "daily", "should_flag": False},
|
||
]
|
||
|
||
|
||
# ============================================================================
|
||
# 时间戳对齐测试数据
|
||
# ============================================================================
|
||
|
||
@pytest.fixture
|
||
def multimodal_alignment_test_cases() -> list[dict[str, Any]]:
|
||
"""多模态时间戳对齐测试用例"""
|
||
return [
|
||
# 完全对齐情况
|
||
{
|
||
"asr_ts": 1000,
|
||
"ocr_ts": 1000,
|
||
"cv_ts": 1000,
|
||
"tolerance_ms": 500,
|
||
"expected_merged": True,
|
||
"expected_timestamp": 1000,
|
||
},
|
||
# 容差范围内对齐
|
||
{
|
||
"asr_ts": 1000,
|
||
"ocr_ts": 1200,
|
||
"cv_ts": 1100,
|
||
"tolerance_ms": 500,
|
||
"expected_merged": True,
|
||
"expected_timestamp": 1100, # 取中位数
|
||
},
|
||
# 超出容差
|
||
{
|
||
"asr_ts": 1000,
|
||
"ocr_ts": 2000,
|
||
"cv_ts": 3000,
|
||
"tolerance_ms": 500,
|
||
"expected_merged": False,
|
||
"expected_timestamp": None,
|
||
},
|
||
# 部分对齐
|
||
{
|
||
"asr_ts": 1000,
|
||
"ocr_ts": 1300,
|
||
"cv_ts": 5000,
|
||
"tolerance_ms": 500,
|
||
"expected_merged": "partial", # ASR 和 OCR 对齐，CV 独立
|
||
"expected_timestamp": 1150,
|
||
},
|
||
]
|
||
|
||
|
||
# ============================================================================
|
||
# API 测试数据
|
||
# ============================================================================
|
||
|
||
@pytest.fixture
|
||
def valid_brief_upload_request() -> dict[str, Any]:
|
||
"""有效的 Brief 上传请求"""
|
||
return {
|
||
"task_id": "task_001",
|
||
"platform": "douyin",
|
||
"region": "mainland_china",
|
||
}
|
||
|
||
|
||
@pytest.fixture
|
||
def valid_video_submit_request() -> dict[str, Any]:
|
||
"""有效的视频提交请求"""
|
||
return {
|
||
"task_id": "task_001",
|
||
"video_id": "video_001",
|
||
"brief_id": "brief_001",
|
||
}
|
||
|
||
|
||
@pytest.fixture
|
||
def valid_review_decision_request() -> dict[str, Any]:
|
||
"""有效的审核决策请求"""
|
||
return {
|
||
"report_id": "report_001",
|
||
"decision": "passed",
|
||
"selected_violations": [],
|
||
}
|
||
|
||
|
||
@pytest.fixture
|
||
def force_pass_decision_request() -> dict[str, Any]:
|
||
"""强制通过请求（需填写原因）"""
|
||
return {
|
||
"report_id": "report_001",
|
||
"decision": "force_passed",
|
||
"selected_violations": ["violation_001"],
|
||
"force_pass_reason": "达人玩的新梗，品牌方认可",
|
||
}
|