helibeiqi/dsh-compaction-pro ↗★ 0
dsh-compaction-pro
High-fidelity, faithful, bilingual, recursive compaction backend for DeepSeek Harness — a drop-in upgrade over dsh-compaction-basic.
AI Analysis
核心用途是作为内置压缩插件的升级版，通过递归分块和强制保留精确数值/路径来压缩上下文。适合需要处理长对话、防止关键信息在上下文窗口超出时被截断丢弃的用户。
Install
This plugin has no verified bundle, or compatibility checks failed. Read the repository notes first. Read the full README ↗
README
Read the full README ↗配置项
| 配置 | 类型 | 默认 | 含义 |
|---|---|---|---|
faithful | boolean | true | 强制保留精确数值/路径/命令/标识符，绝不改写 |
summaryLanguage | 'en' | 'zh' | 'auto' | 'auto' | 摘要语言；auto 跟随会话语言 |
recursive | boolean | true | 大区段先分块压、再合并，避免截断丢信息 |
chunkMessages | number | 40 | recursive 开启时每个分块的消息数 |
customInstruction | string | — | 整段替换内置摘要提示词 |
thresholdRatio | number | 0.8 | 上下文窗口的压缩触发比例（继承自官方） |
retainRatio | number | 0.16 | 保留近期尾部的比例（继承自官方） |
maxTokens | number | 8192 | 摘要生成上限（继承自官方） |
summarizationProvider / summarizationModel | string | '' | 留空则复用会话路由模型，否则钉死摘要模型 |