starmen1265
5983e95352
feat(15-02): integrate KBSettings config for semantic segmentation
...
- Modified SRTProcessor to read enable_semantic_segmentation from KBSettings
- Changed constructor parameter default from False to None (None = read from config)
- Added config priority: constructor parameter > KBSettings > default (True)
- Added logging for config status (INFO level)
- Added comprehensive tests for config integration (TestConfigIntegration)
- Added tests for output format compatibility (TestOutputFormatBenefits)
- Fixed MockAIClient to accept disable_reasoning parameter
- All 34 tests passing
Requirements satisfied:
- CONFIG-01: SRTProcessor reads from KBSettings
- CONFIG-02: Disabled config uses rule-based segmentation
- OUTPUT-01: LLM segmentation outputs H2 titles
- OUTPUT-02: title=null skips title output
- OUTPUT-03: Rule-based segmentation outputs "## 内容"
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com >
2026-03-23 23:03:54 +08:00
starmen1265
bb321d9cd4
feat(14-05): integrate cache into SRTProcessor
...
- Add enable_cache parameter to SRTProcessor.__init__()
- Create SemanticCache instance when semantic segmentation is enabled
- Pass cache to UnifiedSegmentationStrategy through SemanticSegmentationEngine
- Add logging for cache status (enabled/disabled)
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com >
2026-03-23 22:40:47 +08:00
starmen1265
ef4a660faa
feat(13-07): integrate SemanticSegmentationEngine into SRTProcessor
...
- 添加 enable_semantic_segmentation 参数到 SRTProcessor.__init__
- 启用时使用 SemanticSegmentationEngine，禁用时使用 SemanticSegmenter
- AI 不可用时自动降级到规则分段
- 更新 _generate_markdown() 支持带标题的 VideoSegment 输出
- 调整步骤顺序：先加载元数据，再进行语义分段
- 更新 UnifiedSegmentationStrategy.segment() 返回 List[VideoSegment]
- 添加代理方法到 SemanticSegmentationEngine 以保持向后兼容
- 创建全面的集成测试（9 个测试用例）
- 更新现有语义引擎测试以适配新 API
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com >
2026-03-22 00:32:14 +08:00
starmen1265
7ac4937c48
fix(deep-analyzer, srt): correct Zhipu API thinking parameter and fix metadata handling
...
Root cause:
1. thinking.type="disabled" caused ASR correction accuracy degradation
2. SRTProcessor didn't flatten nested metadata structure
3. Low-quality auto-generated summary section
4. duration field type handling issue
Changes:
- DeepAnalyzer: Remove thinking.type="disabled" to enable default thinking mode
- SRTProcessor: Add metadata flattening for nested structures
- SRTProcessor: Remove low-quality summary section generation
- SRTProcessor: Add duration type checking (string or int)
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com >
2026-03-21 21:41:29 +08:00
starmen1265
87b1919483
feat(12-01): implement SRTProcessor and Markdown generation
...
- Add SRTInput/SRTOutput schemas to pipeline
- Create SRTProcessor following BaseProcessor pattern
- Generate Markdown with YAML frontmatter
- Support metadata extraction and formatting
- Handle missing metadata gracefully
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com >
2026-03-19 23:38:56 +08:00
starmen1265