dsh-language-control
Run the following command in DeepSeek Harness:
dsh plugin install wodongx123/dsh-language-control
Paste the following prompt into your AI chat to install this plugin:
Run dsh plugin install wodongx123/dsh-language-control in the DeepSeek Harness terminal, or visit https://github.com/wodongx123/dsh-language-control for the full source code; the plugin activates automatically after installation.
About this plugin
When an agent calls a tool that returns English documentation or code, its chain-of-thought quietly drifts into English—planning, reasoning, and even the final answer to the user all follow, silently breaking the reading experience of Chinese-speaking users. dsh-language-control is a lightweight plugin built to eliminate this language drift.
The approach is simple yet precise: before every reasoning step, the plugin injects a language constraint at the very top of the system prompt with the highest priority (order: -200), mandating that chain-of-thought reasoning, planning, tool-call deliberation, and internal thinking all stay in Chinese. Crucially, this is not a one-time injection at session start—the constraint is re-evaluated on every assemble() call, so even after a tool returns a wall of English text, the directive remains undiluted.
A visual settings panel lets you edit custom constraint text, toggle between custom and default modes, or restore defaults with one click. Zero dependencies, zero configuration, plug and play—ideal for anyone who wants their agent to think and respond in Chinese throughout the entire workflow.
Use Cases
- Agent drifts into English after calling a tool that returns foreign content; forces Chinese reasoning
- Multi-step planning workflow where every internal reasoning step must stay in Chinese
- Team works primarily in Chinese and needs consistent language in both thinking and output
Best For
- Chinese-first agent users and content creators
- Developers building multi-step reasoning workflows
- Team leads who need uniform language across agent output
Related Plugins
A method pack that makes AI coding agents plan against your real baseline, prove completion with fresh evidence, and reduce reworks and unsafe changes.
Turns the DeepSeek Harness session into a captain that builds a durable sub-agent team, splits goals into dependency-aware tasks, and coordinates work via direct messages and a live Web UI.
Gives coding agents design judgment, letting Claude Code, Cursor, and 70+ agents generate and iterate high-quality UI, presentations, and graphics right from the terminal.
Run the Pi ecosystem's plugins on DeepSeek Harness, unmodified, via a compatibility layer that implements Pi's public extension ABI on DSH's native services.