bojansandhaus/jev-lcm-dsh-compaction ↗★ 0

@bojansandhaus/jev-lcm-dsh-compaction

Jev ranking and SQLite lossless context management for DeepSeek Harness 适合需要精细化控制长对话上下文、防止超出模型Token限制的任务。

パッケージ
@bojansandhaus/jev-lcm-dsh-compaction
互換性
未検証
Harness ピア範囲
0.1.6-alpha.2
Cordis ピア範囲
4.0.2
バージョン
1.0.0-rc.1
ライセンス
MIT
最終更新
2026/09/22

インストール

$npx -p @deepseek-ai/dsh dsh plugin --profile web add github:bojansandhaus/jev-lcm-dsh-compaction

ドキュメント

README 全文を読む ↗

What configuration does the DSH bundle expose?

The bundle schema in src/index.ts currently exposes these DSH-level values:

SettingDefaultMeaning
databasePathjev-lcm.sqliteLocal archive path.
thresholdRatio0.8Host compaction trigger ratio.
retainRatio0.16Host retention ratio.
maxTokens8192Engine context budget.
autotrueEnable automatic engine behavior.
jevhost-supplied valueJev engine configuration passed to the implementation.

The engine defaults in src/settings.ts include jev_provider: auto, TypeSafe base https://api.typesafe.ai/v1, OpenRouter base https://openrouter.ai/api, TypeSafe path /systemone, OpenRouter model ~typesafe/jev-latest, a 0.15 fallback threshold, a 0.40 threshold cap, a 0.10 minimum keep rate, a 500-sample calibration window, a 50-sample calibration minimum, a three-turn batch window, a 300-candidate batch cap, a 25,000-token state cap, and a 30,000-token request cap. See docs/reference.md for the complete table and validation rules.

Provider environment variables are:

export TYPESAFE_API_KEY='set-through-your-secret-manager'