Compaction backend replacing LLM summarization with a deterministic semantic extractor (keeps code/paths/commands, drops chatter) plus 28.4x KV-compression accounting.
Install
# from GitHub (first run asks for allowBuilds approval — follow the hint, retry)
dsh plugin --profile web add github:ljsysfurryACE/dsh-compaction
Any plugin you install runs third-party code with your own permissions — it can read your files, use your credentials, and reach the network, and tool approvals don’t sandbox it. GitHub-sourced plugins also run build scripts at install time — pnpm blocks those until you allow them, so an install can stop with ERR_PNPM_GIT_DEP_PREPARE_NOT_ALLOWED or ERR_PNPM_IGNORED_BUILDS; dsh prints the exact key to add under allowBuilds in your profile’s pnpm-workspace.yaml, and the install works on the next run. Allowing a build is a trust decision: only install sources you trust, and pin a commit (github:owner/repo#sha).
README
This plugin publishes its README in Chinese only.
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：
- 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（保留高信息行） |
retainRatio |
0.2 | 压缩后保留比例 |
physical |
true | 物理压缩记账（bytes/token） |
bytesPerToken |
7776 | INT4 压缩后每 token 字节 |
auto |
true | 自动压缩 |
实现
继承 CompactionEngine（compaction seam），实现三个抽象方法：
compactIfNeeded— 自动压力触发compactNow— 手动 /compact 命令compactRegion— 指定范围压缩
保持 dsh 的 surface/事件/持久化契约（compaction/start → replace → compaction/end）， 只替换"如何压缩"。
License
GPL-3.0 © Cloud LTE Studio / AgentFrame
Links
More in this category
vectorize-io/hindsight#coding-agents★ 60728
Hindsight, agent memory that learns: long-term project memory with auto recall and retain, knowledge pages, deep reflection, and per-repo memory banks.
volcengine/OpenViking#examples/dsh-memory-plugin★ 42125
OpenViking memory and context bundle for DeepSeek Harness: pre-step auto-recall and profile injection, session capture, `viking://` URI guarding, and recall/write memory tools backed by an OpenViking server.
MemTensor/MemOS#memos-local-plugin★ 12376
Local-first memory plugin for DeepSeek Harness: layered L1/L2/L3 memory with automatic capture, manual memos_* retrieval tools, and a built-in memory viewer.
agentscope-ai/ReMe#dsh★ 3796
Connects DeepSeek Harness to ReMe's local-first, self-evolving personal knowledge base: automatically captures completed main-agent conversations as user-owned Markdown memory, searches conversations and source material through reme_search with BM25, optional embeddings, and wikilink expansion, and schedules daily memory consolidation.
zilliztech/memsearch#MemSearch★ 2866
Shared Markdown memory for DSH and other coding agents, with automatic capture, pre-step context injection, searchable recall, and memory-to-skill self-evolution through a review panel.
vshulcz/deja-vu#extensions/dsh★ 1529
Reads the session files forty-two other coding agents on this machine already wrote (Claude Code, Codex, Cursor, VS Code Copilot Chat, opencode, OpenClaw, Hermes, Kimi, Cline, Zed and more), including sessions from before it was installed: six tools (deja_recall, deja_session, deja_blame, deja_fix, deja_how, deja_remember), a /deja command, and optional automatic recall added to the runtime context. Local BM25 index, no LLM, no embeddings, no network (dsh plugin --profile web add dsh-deja).
Community comments
Comments are public GitHub Discussions. Loading them connects to GitHub and Giscus; a GitHub account is required to post.