1.50K
total weekly downloads
7 packages
dsh-agent-bench
Regression benchmark for DSH agents: replay fixed cases in isolated sandboxes and diff scorecards across model, prompt and DSH versions. DSH Agent 回归基准。
dsh-fine-tune
Local LoRA / QLoRA fine-tuning for DeepSeek Harness: import JSONL, train on your own GPU (or a remote SSH host), evaluate, and export an adapter. One-click environment install; Windows + NVIDIA verified. 本地微调插件:数据集→训练→评测,支持本机与远程 GPU。
dsh-paper-review-revision
DeepSeek Harness skill bundle for four-round academic review and revision of engineering and applied-mathematics papers.
dsh-publisher-radar
Market-gap radar for DSH plugin developers: splits 6,000+ plugins by category and topic, tags each topic empty / thin / contested / crowded, and ranks the emptiest directions first — so you know what to build before writing it. Also browses the full marke
dsh-readprune
Saves tokens in DeepSeek Harness sessions: read any file by line range or symbol mention instead of whole, and bound oversized tool results while keeping a recovery path. Pruning happens at the source, not after content reaches the context.
dsh-token-frugal
Frugal tool layer for DeepSeek Harness: precise file reading, a self-serve tool catalog, and slimmed oversized tool results, so long sessions stop re-billing whole files and whole pages.
dsh-token-prune
Prunes a session's input tokens at the source: read any file by line range or symbol mention instead of whole, and bound oversized tool results at the boundary while keeping a recovery path. Unlike compression, it stops unneeded content from entering the