dsh-compaction
Compaction backend replacing LLM summarization with a deterministic semantic extractor (keeps code/paths/commands, drops chatter) plus 28.4x KV-compression accounting.
压缩后端:用确定性语义提取器替换 LLM 摘要(保留代码/路径/命令,剔除闲聊),附带 28.4x KV 压缩记账。
How to install
dsh plugin add github:ljsysfurryACE/dsh-compaction About
dsh-compaction-agentframe AgentFrame 压缩后端插件 —— 把 DeepSeek Harness 的默认 LLM 摘要压缩替换为 **语义 + 物理双轨压缩**(28.4x KV 压缩思路)。 为什么替换默认压缩 默认 compaction-basic 用 LLM 摘要(有损总结旧对话),每次压缩都要一次 LLM 调用,且摘要可能丢细节。 AgentFrame 的思路: **语义轨**:MemoryDirector 判断哪些 token 值得保留(去闲聊、留关键) **物理轨**:吸收式 MLA + INT4 量化(270KB→7.6KB/token,28.4x) **效果**:保留关键信息 + 大幅省 token + 不额外调用 LLM 使用 在 profile 的 cordis.patch.yml 中把 compaction-basic 替换为: yaml id: compaction-agentframe name: '@deepseek-ai/dsh-compaction-agentframe' config: semantic: true retainRatio: 0.2 physical: true 配置 | 字段 | 默认 | 说明 | |------|------|------| | semantic | true | 语义压缩(保…
Recommendation signals
Meta
- License
- GPL-3.0
- Language
- JavaScript
- GitHub stars
- 0
- mo. downloads
- –
- Last push
- 2026-09-24
- Created
- 2026-08-14
Links
Basic safety check
- Findings
- curated 收录但无 npm 包/安装命令
- Sources
- curated:awesome-dsh-plugin.com, curated:awesome-dsh-plugin/awesome-dsh-plugin
- Topics
- context-compression, deepseek-harness, dsh, llm, plugin
Related plugins
hindsight
vectorize-io/hindsight/tree/main/hindsight-integrations/coding-agents
Hindsight, agent memory that learns: long-term project memory with auto recall and retain, knowledge pages, deep reflection, and per-repo memory banks.
firecrawl-mcp-server
firecrawl/firecrawl-mcp-server
🔥 Official Firecrawl MCP Server - Adds powerful web scraping and search to Cursor, Claude and any other LLM clients.
dsh-mnemon
omdsh-dev/dsh-mnemon
Cross-agent, local-first persistent memory plugin for DeepSeek Harness (DSH), powered by Mnemon. It shares long-term memory across Mnemon-enabled agents and adds runtime memory, searchable project documents, semantic recall, knowledge graph, and a Sidebar UI.
get-shit-done
gsd-build/get-shit-done
A light-weight and powerful meta-prompting, context engineering and spec-driven development system for Claude Code by TÂCHES.