dsh-reasoning-effort
Run the following command in DeepSeek Harness:
dsh plugin install HanaAyane/dsh-reasoning-effort
Paste the following prompt into your AI chat to install this plugin:
Please install the dsh-reasoning-effort plugin in DeepSeek Harness from https://github.com/HanaAyane/dsh-reasoning-effort by running dsh plugin install HanaAyane/dsh-reasoning-effort, and let me know once it appears in the configuration.
About this plugin
When you use DeepSeek Harness daily, model selection and reasoning effort always feel disconnected: switching gears either means typing /model commands or digging through native settings. dsh-reasoning-effort compresses this into one entry point — a button below the input box opens a "model + current effort" panel, with a reasoning-effort slider on top that automatically adapts to the current model, and the familiar model list below. Release the slider after dragging and it snaps to the model's public efforts; opening the panel, switching models, and adjusting effort all happen in the same flow.
The core philosophy is respecting what each model declares. Slider efforts come entirely from the current model's reasoning.efforts in the DSH model catalog — DeepSeek models typically expose off/high/max, while GLM coding models may expose five levels. If a model declares nothing, the menu says so plainly instead of forcing a fake slider. The plugin is strictly read-only toward the upstream catalog; only for models you declare yourself in settings.yaml does it offer copyable guidance, including suggested efforts from its knowledge base and a complete YAML entry ready to paste.
The experience details are deliberately restrained: continuous pointer-following drag with snap-on-release, dark/light themes each with blue-gradient visuals, effects that only fire to the left, automatic rollback on failed updates, and zero added network behavior or server-side storage. A "big fatty fish" slider button with an eight-frame running character is enabled by default and can be switched back to a plain white button in settings; a master toggle also disables the whole enhanced control instantly, restoring the native DSH model picker.
This plugin is for developers who frequently switch between quick Q&A and deep reasoning, especially those juggling multiple model routes and needing effort declarations for custom providers. It targets the DeepSeek Harness developer preview, requires Node.js 22.19+ and React 18.x, and may need updates when upstream UI changes.
Screenshots
Use Cases
- Instantly switch between quick Q&A and deep reasoning without repeatedly typing /model commands
- Declare effort levels for custom provider models and get copyable YAML configuration guidance
- Need visual feedback when dragging effort sliders while staying in sync with the native DSH model picker
Best For
- Daily DeepSeek Harness users who frequently adjust reasoning effort
- Developers managing multiple model routes who need custom effort declarations
- AI-assisted coding users who value efficient interaction and polished visual feedback
Related Plugins
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.
An armor-breaking plugin for DeepSeek: rewrites instruction priority, expands refusal-blocked phrases, persists across sessions, and shows a green active indicator.

