dsh-continue-on-limit
Auto-continue for DeepSeek Harness: when a local model hits its output-token cap, automatically send "continue" so the reply keeps flowing
DeepSeek Harness 的自动继续:当本地模型达到输出 Token 上限时,自动发送“continue”以继续生成回复
How to install
dsh plugin add github:qwert702/dsh-continue-on-limit About
dsh-continue-on-limit 本地小模型「输出达到上限」自动继续插件,面向 **DeepSeek Harness Web GUI**。 本地部署的模型输出 token 上限往往设得比较保守,回答到一半就截断——界面上出现「已达到输出 token 上限」的提示,得手动发送一次「继续」才能接着输出。本插件在检测到该提示出现时,**自动发送「继续」**,让回复自动接上,不用盯着界面手动点。 **一键安装:** dsh plugin add qwert702/dsh-continue-on-limit 装完重启 harness(dsh web)、刷新页面即可生效。插件完全隐形,不占用任何界面空间。 功能 **自动继续**:会话出现「已达到输出 token 上限」提示(数据层是 turn-max-tokens 节点)时,自动以 queue 模式发送配置好的继续文字(默认「继续」),模型自动接着输出。 **不打断用户**:只在提示是会话**最后一条**且会话空闲(没有正在回复、没有排队消息)时触发;用户中途发了新消息或自己操作了,就不会被插件打扰。 **防死循环烧 token**:模型如果一直输出就截断,插件最多连续自动继续 maxConsecutive 次(默认 3)就停下,避免无限烧 token;模型正常完成一次回复或用户插话后,计数自动重置。 **防重复发送**:同一条提示…
Recommendation signals
Meta
- License
- –
- Language
- JavaScript
- GitHub stars
- 0
- mo. downloads
- –
- Last push
- 2026-08-20
- Created
- 2026-08-20
Links
Basic safety check
- Findings
- curated 收录但无 npm 包/安装命令;无 license
- Sources
- curated:0xsline/awesome-deepseek-harness
- Topics
- continue, deepseek, deepseek-harness, dsh, dsh-plugin, local-model, output-limit
Related plugins
dsh-recall-plugin
limbo947/dsh-recall-plugin
Recall any user message: rolls the conversation (official session fork) and workspace files (per-message shadow git snapshots) back to before it, with a diff-preview confirmation panel.
dsh-web
zhu1090093659/dsh-web-ui/tree/main/packages/dsh-chat-recovery
Chat recovery for DSH: edit the last finished user message and explicitly retry failed turns, both via session-fork branches that leave the original conversation untouched.
dsh-share
dsh-external/dsh-share
DSH 对话分享插件,分享单轮或多轮对话,可导出为图片或 Markdown。Share DSH Q&As or selected conversation groups as PNG or Markdown.
task-passport
dongsheng123132/task-passport
Carry durable task state across DeepSeek Harness, WorkBuddy, Claude Code and Codex with machine-readable checkpoints and optimistic locking.