134 results
DSH web plugin: a LiveBench tab in the Trajectory view (right of 对话/轨迹). Run LiveBench evaluations against every model configured in the DeepSeek Harness — pick provider/model, category, task, release and question range from dropdowns, watch progress, and read scores in place. Ships a Baseline probe set built from the questions a reference model cannot solve.
DSH local folder knowledge base: manual-sync indexing, Chinese hybrid retrieval, explicit-invocation gate, in-panel Office preview, and a Web panel.
Independent community DeepSeek Harness plugin for secure multi-file upload, native images, local document retrieval, and coordinate-aware text/PDF/DOCX/XLSX/PPTX reading.
DSH 试金石：给「自改造」补上评测这一环——用一套金标准用例，把改动前 vs 改动后跑一遍、打分、对比，好就留、不好就撤。
Community plugin for event-grounded, privacy-first agent reliability analysis in DeepSeek Harness.
Measure whether a dsh setup change actually helped: register repeatable cases, run them, and diff the results before and after you change rules, skills, or models.
EVA-style companion HUD for DeepSeek Harness: remaining balance as active time, session cache-hit share as sync ratio.
DSH 超长记忆包：跨会话长期记忆、长文分块检索、聊天记录提取与精炼、上下文总结记录与翻阅。纯 JS 混合检索（BM25 + CJK 二元分词），零第三方依赖。
EVA / MAGI 主题皮肤：把 NERV 中央教条区的界面语言装进 DSH Web GUI（--dsw-* 令牌层覆盖 + CRT 扫描线 / 硬边 / 等宽数字，明朝体汉字标题）
注意力调度与记忆检索规则 —— DeepSeek Harness 技能（内嵌插件分发，可 dsh plugin add 一键安装）
A DeepSeek Harness plugin that adds a "Check updates" button to the Plugins settings section: it scans every installed plugin bundle for published updates and evaluates its peer-dependency compatibility against the running DSH runtime.
DSH Web UI plugin: colors sidebar Session rows by their official status — blue rail = pending interaction, red rail = the selected session, purple rail = finished but unviewed, static rainbow title = pinned. Reads official Client services only; no polling, no host half.
Market Intelligence Analysis Expert for DeepSeek Harness — CaSee (凯见) market/competition/customer intelligence retrieval with two-stage trusted sourcing and T-Score credibility grading.
Search a local Paperless-ngx document library with hybrid semantic + keyword retrieval, returning titles and key snippets as agent context.