---
phase: 13
plan: 07
type: execute
wave: 2
depends_on:
- 13-01
- 13-02
- 13-03
- 13-04
- 13-05
- 13-06
files_modified:
- kb_workflow/pipeline/processors/srt.py
- tests/test_srt_processor_integration.py
autonomous: true
requirements:
- CONFIG-03
must_haves:
truths:
- "SRTProcessor 支持启用/禁用 LLM 语义分段"
- "启用时使用 SemanticSegmentationEngine，禁用时使用规则分段"
- "知乎/文章平台继续使用 AnalyzeProcessor，不受影响"
artifacts:
- path: "kb_workflow/pipeline/processors/srt.py"
provides: "SRTProcessor 集成 LLM 分段"
min_lines: 350
exports: ["SRTProcessor.__init__", "SRTProcessor.process"]
- path: "tests/test_srt_processor_integration.py"
provides: "SRTProcessor 集成测试"
min_lines: 100
key_links:
- from: "SRTProcessor.__init__"
to: "SemanticSegmentationEngine"
via: "enable_semantic_segmentation 参数"
pattern: "enable_semantic.*segmentation"
- from: "SRTProcessor.process"
to: "SemanticSegmentationEngine.segment"
via: "条件分支选择分段方式"
pattern: "if.*semantic.*else.*rule"
- from: "Pipeline.run"
to: "AnalyzeProcessor/SRTProcessor"
via: "平台路由逻辑（保持不变）"
pattern: "platform.*bilibili.*srt"
---
集成 SemanticSegmentationEngine 到 SRTProcessor
目的：让 SRTProcessor 支持 LLM 语义分段（可选），保持向后兼容
输出：更新的 SRTProcessor，集成测试
@/home/starmen1265/.claude/get-shit-done/workflows/execute-plan.md
@/home/starmen1265/.claude/get-shit-done/templates/summary.md
@.planning/phases/13-llm/13-CONTEXT.md
@.planning/phases/13-llm/13-RESEARCH.md
# User Decision (Locked)
- **SRTProcessor 集成**: 配置开关控制 — 新增 `enable_semantic_segmentation` 参数，True 时用 LLM 分段，False 时保留规则分段作为后备选项
- **现有 SemanticSegmenter**: 完全替代 — 新代码完全替代旧代码，SemanticSegmenter 被 SemanticSegmentationEngine 取代
# Existing Code (from srt.py)
```python
class SRTProcessor(BaseProcessor):
def __init__(
self,
state_manager: Optional[WorkflowStateManager] = None,
enable_correction: bool = True
):
...
self.segmenter = SemanticSegmenter() # 规则分段
```
Task 1: 集成 SemanticSegmentationEngine 到 SRTProcessor
kb_workflow/pipeline/processors/srt.py, tests/test_srt_processor_integration.py
更新 SRTProcessor，集成 LLM 语义分段：
1. 在 __init__ 中添加 `enable_semantic_segmentation` 参数：
```python
def __init__(
self,
state_manager: Optional[WorkflowStateManager] = None,
enable_correction: bool = True,
enable_semantic_segmentation: bool = False # 新增参数
):
super().__init__(state_manager)
self.enable_correction = enable_correction
self.enable_semantic_segmentation = enable_semantic_segmentation
# 初始化分段器
if enable_semantic_segmentation:
try:
from kb_workflow.srt.semantic_engine import SemanticSegmentationEngine
from kb_workflow.deep_analyzer import SimpleAIClient
ai_client = SimpleAIClient()
if ai_client.api_key:
self.segmenter = SemanticSegmentationEngine(ai_client=ai_client)
logger.info("Semantic segmentation enabled")
else:
logger.info("AI API key not configured, using rule-based segmenter")
self.segmenter = SemanticSegmenter()
except Exception as e:
logger.warning(f"Failed to initialize semantic engine: {e}, using rule-based segmenter")
self.segmenter = SemanticSegmenter()
else:
self.segmenter = SemanticSegmenter()
# 初始化纠错器（保持不变）
...
```
2. 更新 process() 方法中的分段调用：
```python
# Step 2: 语义分段
logger.info("Segmenting subtitles into paragraphs")
if self.enable_semantic_segmentation and isinstance(self.segmenter, SemanticSegmentationEngine):
# LLM 语义分段
paragraphs = self.segmenter.segment(segments, metadata=metadata)
else:
# 规则分段
paragraphs = self.segmenter.segment(segments)
logger.info(f"Generated {len(paragraphs)} paragraphs")
```
3. 更新 _generate_markdown() 以支持带标题的分段：
```python
def _generate_markdown(
self,
output_file: Path,
metadata: Dict[str, Any],
paragraphs: List[str], # 或 List[VideoSegment]
tags: List[str]
) -> None:
...
# 如果 paragraphs 包含 VideoSegment（LLM 分段结果）
if paragraphs and hasattr(paragraphs[0], 'title'):
for seg in paragraphs:
if seg.title:
content_parts.append(f"## {seg.title}")
content_parts.append(seg.content)
content_parts.append("")
else:
# 规则分段（纯文本）
content_parts.append("## 内容")
content_parts.append("")
for para in paragraphs:
content_parts.append(para)
content_parts.append("")
...
```
4. 更新 SRTInput/Output Schema（如果需要）：
- 考虑是否需要添加 semantic_segmentation_enabled 字段
5. 创建集成测试：
```python
# tests/test_srt_processor_integration.py
def test_srt_processor_semantic_mode():
"""测试 SRTProcessor 语义分段模式"""
# mock AI client
# enable_semantic_segmentation=True
# 验证调用 SemanticSegmentationEngine
def test_srt_processor_rule_based_mode():
"""测试 SRTProcessor 规则分段模式"""
# enable_semantic_segmentation=False
# 验证使用 SemanticSegmenter
def test_srt_processor_semantic_fallback():
"""测试 AI 不可用时降级到规则分段"""
# AI client 无效
# 验证降级到 SemanticSegmenter
```
6. 确保 Pipeline 平台路由逻辑不受影响（Phase 12 已实现）：
- B站视频使用 SRTProcessor
- 知乎/文章继续使用 AnalyzeProcessor
- 不修改 Pipeline.run() 中的路由逻辑
pytest tests/test_srt_processor_integration.py::test_srt_processor_semantic_mode -x
- SRTProcessor 支持 enable_semantic_segmentation 参数
- 启用时使用 SemanticSegmentationEngine
- 禁用时使用 SemanticSegmenter（向后兼容）
- AI 不可用时自动降级到规则分段
- _generate_markdown() 支持带标题的分段
- 集成测试验证所有场景
- 平台路由逻辑不受影响（知乎/文章继续使用 AnalyzeProcessor）
- pytest tests/test_srt_processor_integration.py -v 全部通过
- pytest tests/test_srt_processor.py -v 现有测试不回归
- 平台路由逻辑不受影响
- 向后兼容（enable_semantic_segmentation=False 时行为不变）
1. SRTProcessor 支持 enable_semantic_segmentation 参数
2. 启用时使用 LLM 分段，禁用时使用规则分段
3. AI 不可用时自动降级
4. 集成测试验证所有场景
5. 平台路由逻辑不受影响