← DeepSeek Harness Plugins
DeepSeek Harness PluginManifest Valid

dsh-argp

Guarded dual-engine context compaction for DeepSeek Harness with per-atom shrinking, graph eviction, and exact recall.

Developer ToolsSkills & Workflows
11GitHub Stars1ForksUpdated2026-10-17

Install

$ dsh plugin --profile <name> add dsh-argp # 安装即挂;装好后在会话里输入 /argp 查看实时状态

Plugin Overview & Capabilities

AI-assisted organization based on the public repository snapshot. The content must be grounded in source evidence and does not replace compatibility or security verification.

source-grounded
dsh-argp is a guarded context compaction plugin for DeepSeek Harness (dsh). It replaces free-form summary rewrites with a dual-engine architecture: Stage-1 eager per-atom shrinking governed by deterministic verbatim guards, and Stage-2 lazy 0-LLM reference graph eviction with exact budget convergence. Preserving original tokens in an append-only log, it provides verifiable byte-exact recall while maintaining prompt prefix cache stability across multi-turn sessions.

Key Capabilities

  • Stage-1 eager per-atom compression with deterministic guard enforcement (extract/summary/keep decisions)
  • Stage-2 lazy 0-LLM reference graph pruning across three-tiered triggers (turn start, mid-turn pressure, truncation continuation)
  • Append-only factual history log with two-tier detail/summary zoom recall
  • Preservation of system and request prefix cache fingerprints
  • Real-time session status monitoring via the /argp command
  • Seamless drop-in installation with automatic disabling of stock basic compaction

Useful For

  • Long-running autonomous agent sessions requiring extended multi-turn context without prompt bloat
  • High-precision coding tasks where verbatim paths, error stacks, and configs must not be destroyed by LLM summaries
  • Cost and token optimization while preserving prompt cache hit rates across turns

Who It Fits

  • DeepSeek Harness developers and power users running long-horizon multi-turn agent workflows
  • Engineers requiring strictly budgeted context consumption with zero-loss exact recall

Documented Limitations

  • Per-atom shrink proposals depend on model instruction-following quality (though guards prevent corrupted data from passing)
  • Stage-1 eager compaction incurs side-channel output token costs during per-turn LLM passes
  • Susceptible to host-level context window truncation (B-6) where non-replaced live nodes sliced off by request assembly cannot be recalled
  • Disabled by default under the 'minimal' preset unless explicitly overridden in skipPresets configuration

DSH Compatibility

Version-specific runtime evidence collected by DSH Plugin. A missing result means we have not tested that combination yet.

Not tested yetNo runtime compatibility tests have been published yet.

Security Signals

Objective signals discovered from package metadata and source inspection. These are not a guarantee that a plugin is safe.

Dependency Counts

package.json declares 0 runtime, 25 development, 8 peer, and 0 optional dependencies.

info
Dsh Bundle Declared

package.json declares DSH bundle metadata.

info
License Declared

GitHub reports the repository license as MIT.

info
Package Manifest Available

A root package.json was captured and can be inspected by the registry.

info
Source Available

Public GitHub source metadata is available for this registry snapshot.

info

Source & Registry Notes

Public provenance, Registry classification, and the latest source check for this entry, kept separate from runtime verification.

Source repository
yoza10635/dsh-argp
Registry source
GitHub · dsh-plugin topic
Registry classification
Plugin
Source checked
825760b · 2026-10-18

This project is independently indexed from public source information. DSH Plugin is not affiliated with DeepSeek or the plugin author. Always check the author repository before installation.

Repository Activity

GitHub stars
11GitHub stars
Forks
1
Open issues
0
Last commit
2026-10-17
Last release
2026-10-17
For maintainers

Maintaining this plugin?

This listing is generated from public repository data. If you maintain this project, you can review the information and share this listing with your users if you find it useful.

Add to README
Listed on DSHPlugin.app