跳过主要内容

插件

浏览、筛选并安装 DeepSeek-Harness 插件。

共 579 个插件

I

dsh-llm-provider

imchangchang/dsh-llm-provider

A dsh plugin that replaces the built-in llm-pi-ai and llm-deepseek adapters, the model selector and the Models settings page with self-maintained versions, tracks upstream @earendil-works/pi-ai, and adds quota/balance lookups plus a provider management UI

015天前用量与计费MIT
Z

dsh-tier-router

zhangzhangco/dsh-tier-router

Automatic tier-based model routing for DeepSeek Harness: a virtual `smart` model classifies each request by difficulty (hard / normal / easy) and by vision need, then delegates it to the models you already configured. 三级难度 + 视觉自动路由,虚拟 smart 模型零配置接入。

014天前视觉、语音与多模态
X

dsh-xingtu-skills

xingtu1996/dsh-xingtu-skills

XingTu 26 production-grade AI agent skills, packaged as a DeepSeek Harness (DSH) skills plugin — token compression, code review, safe refactor, migration, evidence review & more. 26 个生产级 SKILL.md 技能,DSH 一键安装即用。

018天前Git 与代码评审MIT
N

dsh-memory

navid-kianfar/dsh-memory

Persistent, searchable, per-project memory for the DeepSeek Harness: decisions, rules, and sprint context in a queryable DuckDB file, rules enforced on every model request, with a full management UI in the Web Client

015天前UI 增强MIT
N

dsh-usage-info

navid-kianfar/dsh-usage-info

Context occupancy and account balance for the DeepSeek Harness Web Client: a session-header readout, a balance capability seam, and a DeepSeek provider

015天前用量与计费MIT
A

dsh-reasoning-only-guard

apex-mochen/dsh-reasoning-only-guard

防止「只有推理、没有正文」的一轮把整个会话变成不可用。某一轮没有可见文本也没有 tool-call 时,assistant 消息会以空内容落盘,此后该会话的每一次请求都会被网关以 "content or tool_calls must be set" 拒绝,会话彻底死亡且无法再从对话里找回。本插件只注册一个 llm/stream waterfall 监听器,且仅当这一轮没有任何可见产出时,在终止 finish 之前注入一小段文本,使落盘消息永远不为空。附带 DSH 自带 mock 造不出的夹具(严格只有推理的 SSE 服务,因为 llm-mock-server 不允许空 successText,且 reasoning_success 总会再补一段正文),以及把真实落盘会话经 DSH 自己的 serializeMessages 重放的端到端复现。属预防而非修复。零依赖、单文件、不访问进程与文件系统。

0会话与消息可能已停更
W

dsh-llm-hub

webkubor/dsh-llm-hub

补上官方 pi-ai 适配器在模型页缺的那半:网关可达性与延迟探测、一键拉取网关模型目录并勾选写回配置、DeepSeek 余额与各家配额就地常驻、每个 provider 实际接入的协议与地址;模型下拉只列当前真能用的分组 —— key 失效、余额耗尽、配额用尽都会被摘掉,修好后自动恢复。

0模型与账号接入可能已停更
M

dsh-image-guard

mafeis/dsh-image-guard

发送前把请求中的历史图片裁剪至保留张数,并从上游 400 错误解析单次图片上限、按更少张数降级重试,保证含图会话不被图片数量限制打断。

017天前视觉、语音与多模态MIT
1

dsh-reasoning-tiers

1069137617/dsh-reasoning-tiers

在 llm-pi-ai 设置节为第三方模型提供商补齐推理档位(reasoning effort)声明,让 DSH 自带的思考强度选择器对 pi-ai 目录之外的模型真正可用。

018天前模型与账号接入MIT
A

dsh-llm-balance-siderbar

aliengene/dsh-llm-balance-siderbar

DeepSeek Harness 服务提供商用量查询插件:在宿主侧读取 DeepSeek、Moonshot 余额与 Kimi For Coding、OpenCode Go 的配额窗口,在常驻卡片中展示,支持三种视图、边缘磁吸与收起为字母气泡。

019天前用量与计费MIT
Z

dsh-context-budget

zpis666/dsh-context-budget

在模型设置页为每个 pi-ai 路由新增上下文窗口与压缩阈值字段,并在输入框右下角按厂商提供思考强度选择器,写入 llm-pi-ai 容量、reasoningEfforts 与 compaction-basic 补丁行。

0模型与账号接入可能已停更
P

dsh-project-context

p02-1010751281/dsh-project-context

dsh (DeepSeek Harness) plugins for project-level persistent context: session archive with a mechanical index, memory consolidation (CONTEXT.md + MEMORY.md), low-frequency skill autolearn, and automatic handoff.

018天前记忆MIT
D

dsh-web-search-local

ddowbnac/dsh-web-search-local

A web-search provider for the DeepSeek Harness, fully local no API key needed.

018天前模型与账号接入
D

dsh-claude-auth-proxy

ddowbnac/dsh-claude-auth-proxy

LLM provider plugin for the DeepSeek Harness (dsh): streams Claude through your Claude Code subscription (claude.ai OAuth credentials), no Anthropic API key required.

018天前模型与账号接入
Z

dsh-permission-matrix

zhang8019/dsh-permission-matrix

把 DSH 权限拆成 9 个可选预设(3 种沙箱 × 4 种审批),规则 + LLM 双通道四档风险分级,高风险操作需输入批准密码才放行,附 JSONL 审计日志。

020天前安全与权限MIT
W

dsh-resume-screening

wangzhanchao883/dsh-resume-screening

面向 HR 的简历筛选插件:把 Word/PDF/Excel 简历统一收进简历库建档(技能/粗分类/入库时间),新简历自动入库、重复自动去重;用大白话下指令即可筛选,硬性条件靠库内规则精准过滤,入围的少数人再由 LLM 精判适配度,结果可导出、可追溯。

0前天工具与能力MIT
J

dsh-markitdown

jiekesu967/dsh-markitdown

提供 markitdown 工具,把 PDF、Word、Excel、PowerPoint、HTML、CSV、EPUB 与网址转成 Markdown:能找到真实的 Microsoft MarkItDown 时驱动它,否则使用内置的零依赖转换器。

017天前工具与能力MIT
X

dsh-model-probe

xiaomao49/dsh-model-probe

直接实测 llm-pi-ai 端点真正接受的能力——从它自己的模型列表取上下文窗口,从它自己的报错反推输出上限与推理档位,用生成的测试图验证图像支持——再报告与证据不符的声明值,确认后才改写。

0模型与账号接入可能已停更
B

dsh-answer-reviewer

bycall/dsh-answer-reviewer

智能体回答审查器:每一轮 agent 的最终输出由独立 LLM 重新打分(1-100),低于阈值时携带具体反馈引导 agent 自我修正;评分以角标形式展示在该回答的操作行上(通过=绿、未达标=红,悬停可见重试次数与原因)。闸值配置经 127.0.0.1 本地 HTTP 服务实时修改,并支持输入框上方的折叠面板与 dsh-better-sidebar 侧栏配置页。

0记忆可能已停更
S

dsh-livedocs

sogoodayo/dsh-livedocs

DSH 实时库文档:写代码前自动拉取项目安装版本的官方文档,消灭幻觉 API。零配置零 Key。Version-pinned live library docs for DeepSeek Harness — kill hallucinated APIs

019天前工具与能力MIT
F

dsh-cron

fan56/dsh-cron

dsh plugin: cron scheduling — bounded tasks with calendar & interval rules, delivered to live agents

019天前工具与能力MIT
K

dsh-cwl

kalifun/dsh-cwl

CWL — Context Window Lifecycle for DeepSeek Harness: structured context eviction (arXiv:2606.11213). Graduated, deterministic, zero-LLM eviction of exploration/action episodes when context pressure exceeds budget — no summarization lossiness, no hallucina

026天前工具与能力MIT
F

dsh-llm-stats

fan56/dsh-llm-stats

dsh plugin: persistent whole-machine LLM usage ledger with a markdown /llm-stats report (GFM tables + emoji: tokens, cache hit, model/tool time, per-day bars)

020天前用量与计费MIT
V

dsh-livebench-panel

vithrive/dsh-livebench-panel

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

019天前UI 增强MIT