dsh-think-in-chinese
Run the following command in DeepSeek Harness:
dsh plugin install Mr-cjf/dsh-think-in-chinese
Paste the following prompt into your AI chat to install this plugin:
Run dsh plugin install Mr-cjf/dsh-think-in-chinese in the terminal, or clone from https://github.com/Mr-cjf/dsh-think-in-chinese to place files in the global plugin directory, append the patch line to cordis.patch.yml, and restart DSH.
About this plugin
You ask in Chinese, the agent answers in Chinese, yet peek behind the curtain and the internal reasoning trace is still all in English. That mismatch between output language and thinking language is the exact gap dsh-think-in-chinese closes for DeepSeek Harness users.
The plugin stays lean: it registers a single prompt line via systemPrompt.section() at order -50, ahead of the persona block, instructing the model to reason in Simplified Chinese. Because the injection is global, it survives every preset switch and requires no translation API or external service—just one prompt directive that pulls the model's internal monologue into your language.
It is built for any Chinese-speaking user who wants the full chain—reasoning, drafting, and final reply—rendered in Simplified Chinese. Whether you are chatting casually, annotating code, or asking the model to unpack a long article, the thinking text and the output finally read as one continuous, unbroken voice.
Use Cases
- Asking in Chinese while wanting the internal reasoning trace rendered in Simplified Chinese
- Switching presets without losing the Chinese thinking language every time
- Reading the model's thinking output without the jarring mix of Chinese and English
- Annotating code or unpacking long articles when reasoning and reply should stay in sync
Best For
- DSH users whose primary working language is Simplified Chinese
- Developers and researchers who audit the model's reasoning step by step
- Chinese users who expect the thinking and reply lanes to speak the same language
Related Plugins
A unified suite combining hot runtime injection, task-aware thinking-mode routing, and a graded session protocol with red-team gates to sustain model diligence across long-horizon inference.
ModLens is a vision plugin for DeepSeek Harness that gives text-only models sight by reading images pasted directly into chat, with zero-config setup and multiple vision engines.
On-demand vision for text-only DeepSeek Harness agents: built-in free keyless vision chain and 14 vision tools, routing image turns as tool calls to vision models with pixel fidelity, no Python needed, one-command install.
Give text-only models in DeepSeek Harness eyes, enabling image Q&A, long-screenshot OCR, UI restoration, and GUI visual tasks.