dsh-price-less
Run the following command in DeepSeek Harness:
dsh plugin install 54shitaimzf/dsh-price-less
Paste the following prompt into your AI chat to install this plugin:
Clone the source from https://github.com/54shitaimzf/dsh-price-less into a local DeepSeek Harness tree (Node 20+), build it with DSH_CHECKOUT=… npm run build, then load it via dev_inject_plugin and restart DSH to activate the plugin.
About this plugin
Long AI-coding sessions hit an uncomfortable paradox: the context window keeps growing, every request pays for early turns that are now irrelevant, and the information that actually matters gets buried under hundreds of rounds of tool output and verbose logs. dsh-price-less is an unofficial context steward for DeepSeek Harness that uses explicit task boundaries, a persistent project brief, tool-output shear, and boundary-aware compaction to fold stale content into a versioned archive, so only the lean context the current task needs reaches the model.
Four defence layers fire in sequence: write-time shaping trims oversized process logs before they are booked, exchange-pair shear drops reads that have been superseded by a later write, compaction folds closed tasks into an archive with a hot tail when the window hits the 35 percent pressure line, and restore replays state with pure functions on session resume. Every path follows a fail-lazy rule - on any doubt nothing changes - and every action leaves a replayable fact log, so your content is never silently lost.
This plugin is aimed at developers who run long, multi-task coding sessions inside DeepSeek Harness and want every token sent to the model to earn its place. If you are comfortable working in a local harness source tree (Node 20+, ESM) and do not want to hand-roll truncation rules or risk losing critical context, this is worth exploring. It is still experimental and tightly coupled to a specific host build, so back up your sessions directory before first use.
Use Cases
- Long coding sessions where the context window bloats and token costs climb steadily
- Multi-task workflows that need clear boundaries and automatic archiving of stale content
- Iterative development where tool output can be auto-trimmed to keep only task-relevant context
Best For
- Developers running long, multi-task coding sessions inside DeepSeek Harness
- Users who want to reduce per-request token cost without losing critical context
- Technical explorers comfortable building experimental plugins in a local source tree
Related Plugins
Memory layer for coding agents that indexes local session history and auto-recalls relevant context before edits, commands, and failures, with no manual search needed.
Traceable, searchable cross-session memory for AI agents that turns conversation knowledge into a typed knowledge graph and recalls relevant subgraphs instead of replaying full history, natively integrated with DeepSeek Harness.
Gives DSH AI cross-session long-term memory, to-do and skill management, plus multi-session orchestration, external AI delegation, and an infinite canvas that grows with you.
dsh-mnemon is a three-tier, pluggable, Agent-driven memory system for DeepSeek Harness, combining Runtime memory, Project Documents, and replaceable Memory Spaces with nine long-term providers.