A deterministic DSH context compiler with original-token checkpoints and log recall. Direct installation requires a copied preset to replace the built-in engine.
DSH integration
Native runtime
Author-claimed
Safety audit
Unaudited
Last verified
2026-10-19
License
MIT
01What can it help you accomplish?
Compact DSH session history without a summarizer model call.
A checkpoint built from original tokens, with seq pointers for later recovery.
DSH users replacing the compaction-basic backend.
02How to install into DeepSeek Harness
Prerequisites
- A DeepSeek Harness web profile; built-in presets pin compaction-basic and need a copied preset for the direct-install route.
Installation steps
- 01
Install: `dsh plugin --profile web add dsh-compaction-instant`.
$ dsh plugin --profile web add dsh-compaction-instant
- 02
In a session using the shipped cordis preset (creation mode), ask it to copy standard and change the copy’s compaction engine to dsh-compaction-instant; use that new preset.
- 03
Restart dsh web once so its boot graph loads the client bundle and settings card.
Verify the integration
- Start a session on the copied preset; the README identifies successful preset mounting as validation.
03DSH integration and capability boundaries
Native DSH compaction backend with recall tools and a /recall command.
Deterministic compaction
Shadowed session history→Checkpoint containing original tokens and seq markers
Same-session recall and search
Seq pointers or keyword/regex queries→Restored log content or matching events with seq pointers
04Who is it for? When not to use it?
Good for
- DSH users replacing the compaction-basic backend.
Not for
- Only one ctx.compaction implementation may be mounted per context; preset instances keep their own isolate realm.
- For prose-heavy history, truncation can be less dense than an LLM summary. Elided content is recoverable through seq pointers and recall.
05Compatibility, maintenance and safety notes
- Only one ctx.compaction implementation may be mounted per context; preset instances keep their own isolate realm.
- For prose-heavy history, truncation can be less dense than an LLM summary. Elided content is recoverable through seq pointers and recall.
README claims pinned to a commit; runtime not tested.
06Frequently asked questions
Does compaction call a model?
No. The compiler uses deterministic text processing and retains original tokens.
Is direct installation sufficient for built-in presets?
In a session using the shipped cordis preset (creation mode), ask it to copy standard and change the copy’s compaction engine to dsh-compaction-instant; use that new preset.
How do I validate the copied DSH preset?
Start a session on the copied preset; the README identifies successful preset mounting as validation.
What is the main compaction trade-off?
For prose-heavy history, truncation can be less dense than an LLM summary. Elided content is recoverable through seq pointers and recall.
07Related DSH workflows
weknora
by tencent
Open-source LLM knowledge platform: turn raw documents into a queryable RAG, an autonomous reasoning agent, and a self-maintaining Wiki.
honcho
by plastic-labs
Memory library for building stateful agents
webnovel-writer
by lingfengqaq
A long-form Chinese web-novel writing assistant for Claude Code and DeepSeek Harness, designed to reduce memory loss and hallucinations in serialized projects of up to 2 million characters.
mirage
by strukto-ai
The World's First Virtual Terminal for AI Agents
08Data and sources
Instant, near-lossless context compaction for [DeepSeek Harness](https://github.com/deepseek-ai/deepseek-harness)
This page is generated from the project’s public documentation, repository metadata and a structured parse of DSH Plugins; last verified on 2026-10-19. Found an error? Submit a correction.
Best DeepSeek Harness Plugins
Twelve plugins worth installing first — picked from the whole catalog, across every category.