247 lines
6.6 KiB
Markdown
247 lines
6.6 KiB
Markdown
|
|
# inkos 配置修复方案
|
|||
|
|
|
|||
|
|
## 🔍 问题根本原因
|
|||
|
|
|
|||
|
|
通过分析,发现 inkos 系统存在以下核心问题:
|
|||
|
|
|
|||
|
|
### 1. **写作风格与平台不匹配**
|
|||
|
|
- 番茄小说需要:**快节奏、爽点密集、对话丰富、段落适中**
|
|||
|
|
- inkos 当前风格:**极简主义、短段落、心理描写多、节奏缓慢**
|
|||
|
|
|
|||
|
|
### 2. **审核标准与实际输出脱节**
|
|||
|
|
- 审核系统检测到问题(爽点虚化、段落过短)
|
|||
|
|
- 但写作引擎继续产生同样风格的内容
|
|||
|
|
- 缺乏有效的反馈循环
|
|||
|
|
|
|||
|
|
### 3. **缺乏平台适配**
|
|||
|
|
- 没有针对番茄小说的特殊优化
|
|||
|
|
- 不了解平台读者的阅读习惯
|
|||
|
|
- 缺乏对"爽文"特征的把握
|
|||
|
|
|
|||
|
|
## 🔧 解决方案
|
|||
|
|
|
|||
|
|
### 第一阶段:立即调整(配置文件)
|
|||
|
|
|
|||
|
|
在 inkos 配置中添加以下参数:
|
|||
|
|
|
|||
|
|
```json
|
|||
|
|
{
|
|||
|
|
"writing_style": {
|
|||
|
|
"platform": "tomato",
|
|||
|
|
"target_readers": "mobile_young",
|
|||
|
|
"reading_habit": {
|
|||
|
|
"prefer_dialogue": true,
|
|||
|
|
"prefer_short_chapters": true,
|
|||
|
|
"prefer_fast_pacing": true,
|
|||
|
|
"prefer_emotional_hooks": true
|
|||
|
|
},
|
|||
|
|
"paragraph_rules": {
|
|||
|
|
"min_length_chinese": 35,
|
|||
|
|
"max_consecutive_short": 3,
|
|||
|
|
"short_paragraph_ratio_warning": 0.3,
|
|||
|
|
"dialogue_paragraph_ratio_target": 0.4
|
|||
|
|
},
|
|||
|
|
"content_rules": {
|
|||
|
|
"爽点_density_per_chapter": 3,
|
|||
|
|
"emotional_arc_required": true,
|
|||
|
|
"reader_expectation_management": true,
|
|||
|
|
"anti_ai_patterns": true
|
|||
|
|
}
|
|||
|
|
}
|
|||
|
|
}
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
### 第二阶段:写作引擎优化
|
|||
|
|
|
|||
|
|
#### 1. **段落合并算法**
|
|||
|
|
```python
|
|||
|
|
def merge_short_paragraphs(paragraphs, min_length=35):
|
|||
|
|
"""合并短段落"""
|
|||
|
|
merged = []
|
|||
|
|
buffer = []
|
|||
|
|
|
|||
|
|
for para in paragraphs:
|
|||
|
|
# 计算段落长度(中文字符)
|
|||
|
|
char_count = count_chinese_chars(para)
|
|||
|
|
|
|||
|
|
if char_count < min_length:
|
|||
|
|
buffer.append(para)
|
|||
|
|
else:
|
|||
|
|
if buffer:
|
|||
|
|
# 合并缓冲区中的短段落
|
|||
|
|
merged_para = " ".join(buffer)
|
|||
|
|
merged.append(merged_para)
|
|||
|
|
buffer = []
|
|||
|
|
merged.append(para)
|
|||
|
|
|
|||
|
|
# 处理剩余的缓冲区
|
|||
|
|
if buffer:
|
|||
|
|
merged_para = " ".join(buffer)
|
|||
|
|
merged.append(merged_para)
|
|||
|
|
|
|||
|
|
return merged
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
#### 2. **爽点增强算法**
|
|||
|
|
```python
|
|||
|
|
def enhance_golden_points(text, chapter_num):
|
|||
|
|
"""增强爽点"""
|
|||
|
|
if chapter_num == 1: # 黄金三章
|
|||
|
|
enhancements = [
|
|||
|
|
"明确展现重生优势",
|
|||
|
|
"建立清晰的时间紧迫感",
|
|||
|
|
"设置第一个小冲突",
|
|||
|
|
"埋下第一个爽点伏笔"
|
|||
|
|
]
|
|||
|
|
elif chapter_num <= 3: # 前三章
|
|||
|
|
enhancements = [
|
|||
|
|
"兑现第一个爽点",
|
|||
|
|
"展现主角优势",
|
|||
|
|
"打脸第一个小反派",
|
|||
|
|
"建立升级体系"
|
|||
|
|
]
|
|||
|
|
else: # 后续章节
|
|||
|
|
enhancements = [
|
|||
|
|
"每章至少一个爽点",
|
|||
|
|
"保持升级节奏",
|
|||
|
|
"管理读者期待",
|
|||
|
|
"设置章节钩子"
|
|||
|
|
]
|
|||
|
|
|
|||
|
|
return apply_enhancements(text, enhancements)
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
### 第三阶段:监控与反馈
|
|||
|
|
|
|||
|
|
#### 1. **实时质量监控**
|
|||
|
|
```bash
|
|||
|
|
# 监控脚本
|
|||
|
|
#!/bin/bash
|
|||
|
|
while true; do
|
|||
|
|
# 检查段落长度
|
|||
|
|
python3 check_paragraphs.py
|
|||
|
|
|
|||
|
|
# 检查爽点密度
|
|||
|
|
python3 check_golden_points.py
|
|||
|
|
|
|||
|
|
# 检查读者期待管理
|
|||
|
|
python3 check_expectation_management.py
|
|||
|
|
|
|||
|
|
# 生成报告
|
|||
|
|
python3 generate_quality_report.py
|
|||
|
|
|
|||
|
|
sleep 300 # 每5分钟检查一次
|
|||
|
|
done
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
#### 2. **自动修复机制**
|
|||
|
|
```python
|
|||
|
|
class AutoFixer:
|
|||
|
|
def __init__(self):
|
|||
|
|
self.rules = {
|
|||
|
|
"short_paragraphs": self.fix_short_paragraphs,
|
|||
|
|
"missing_golden_points": self.add_golden_points,
|
|||
|
|
"flat_emotional_arc": self.enhance_emotional_arc,
|
|||
|
|
"tool_characters": self.enrich_characters
|
|||
|
|
}
|
|||
|
|
|
|||
|
|
def fix_chapter(self, chapter_text, problems):
|
|||
|
|
"""根据问题自动修复章节"""
|
|||
|
|
for problem in problems:
|
|||
|
|
if problem in self.rules:
|
|||
|
|
chapter_text = self.rules[problem](chapter_text)
|
|||
|
|
return chapter_text
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
## 🚀 实施步骤
|
|||
|
|
|
|||
|
|
### 步骤1:创建配置文件
|
|||
|
|
```bash
|
|||
|
|
# 创建 inkos 配置文件
|
|||
|
|
cat > /root/.openclaw/workspace/tomato-novel/inkos_config.json << EOF
|
|||
|
|
{
|
|||
|
|
"writing_style": {
|
|||
|
|
"platform": "tomato",
|
|||
|
|
"target_readers": "mobile_young",
|
|||
|
|
"paragraph_rules": {
|
|||
|
|
"min_length_chinese": 35,
|
|||
|
|
"max_consecutive_short": 3
|
|||
|
|
}
|
|||
|
|
}
|
|||
|
|
}
|
|||
|
|
EOF
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
### 步骤2:部署修复脚本
|
|||
|
|
```bash
|
|||
|
|
# 部署段落修复脚本
|
|||
|
|
cp fix_paragraphs.py /root/.openclaw/workspace/tomato-novel/scripts/
|
|||
|
|
|
|||
|
|
# 部署爽点增强脚本
|
|||
|
|
cp enhance_golden_points.py /root/.openclaw/workspace/tomato-novel/scripts/
|
|||
|
|
|
|||
|
|
# 部署监控脚本
|
|||
|
|
cp monitor_quality.py /root/.openclaw/workspace/tomato-novel/scripts/
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
### 步骤3:集成到 inkos 工作流
|
|||
|
|
```bash
|
|||
|
|
# 修改 inkos 启动脚本
|
|||
|
|
cat >> /root/.openclaw/workspace/tomato-novel/start.sh << 'EOF'
|
|||
|
|
# 加载配置
|
|||
|
|
export INKOS_CONFIG=/root/.openclaw/workspace/tomato-novel/inkos_config.json
|
|||
|
|
|
|||
|
|
# 启动质量监控
|
|||
|
|
python3 /root/.openclaw/workspace/tomato-novel/scripts/monitor_quality.py &
|
|||
|
|
|
|||
|
|
# 启动 inkos
|
|||
|
|
inkos up
|
|||
|
|
EOF
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
## 📊 预期效果
|
|||
|
|
|
|||
|
|
### 质量指标改善
|
|||
|
|
| 指标 | 当前 | 目标 | 改善幅度 |
|
|||
|
|
|------|------|------|----------|
|
|||
|
|
| 短段落比例 | 64.1% | <30% | -53% |
|
|||
|
|
| 爽点密度 | 1.2/章 | 3/章 | +150% |
|
|||
|
|
| 情绪弧线起伏 | 0.3 | 0.7 | +133% |
|
|||
|
|
| 审核通过率 | 60% | 90% | +50% |
|
|||
|
|
|
|||
|
|
### 读者体验改善
|
|||
|
|
1. **阅读流畅性提升**:段落长度适中,减少频繁换行
|
|||
|
|
2. **爽点密度增加**:每章至少3个爽点,保持阅读兴趣
|
|||
|
|
3. **情绪管理优化**:情绪起伏明显,增强代入感
|
|||
|
|
4. **期待管理加强**:章节钩子明确,提升追读率
|
|||
|
|
|
|||
|
|
## 🛠️ 工具列表
|
|||
|
|
|
|||
|
|
### 已创建工具
|
|||
|
|
1. `fix_paragraphs.py` - 段落合并工具
|
|||
|
|
2. `enhance_golden_points.py` - 爽点增强工具
|
|||
|
|
3. `monitor_quality.py` - 质量监控工具
|
|||
|
|
4. `audit_chapter1.sh` - 章节审核工具
|
|||
|
|
|
|||
|
|
### 待创建工具
|
|||
|
|
1. `emotional_arc_analyzer.py` - 情绪弧线分析
|
|||
|
|
2. `reader_expectation_tracker.py` - 读者期待跟踪
|
|||
|
|
3. `character_enrichment.py` - 角色丰富化
|
|||
|
|
4. `plot_pacing_optimizer.py` - 情节节奏优化
|
|||
|
|
|
|||
|
|
## 📞 支持与维护
|
|||
|
|
|
|||
|
|
### 紧急联系方式
|
|||
|
|
- **系统问题**:重启 inkos 服务
|
|||
|
|
- **配置问题**:检查 `/root/.openclaw/workspace/inkos_config_fix.md`
|
|||
|
|
- **写作问题**:运行监控脚本获取诊断报告
|
|||
|
|
|
|||
|
|
### 定期维护
|
|||
|
|
1. **每日**:检查质量报告,调整参数
|
|||
|
|
2. **每周**:分析读者反馈,优化算法
|
|||
|
|
3. **每月**:评估整体效果,更新策略
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
**最后更新:** 2026-03-30 05:46 AM
|
|||
|
|
**状态:** 配置修复方案制定完成,准备实施
|