AI Agent Hub
Back to plugins
🖥️

dsh-effort-bars

Client Updated 2026.08.19

Run the following command in DeepSeek Harness:

dsh plugin install imdeniil/dsh-effort-bars

Paste the following prompt into your AI chat to install this plugin:

Run dsh plugin install imdeniil/dsh-effort-bars in DeepSeek Harness to install this plugin; the source repository is https://github.com/imdeniil/dsh-effort-bars.

About this plugin

Switching a reasoning effort level in DeepSeek Harness or DSH Desktop used to mean: open the model menu, drill into the Effort submenu, pick a value. dsh-effort-bars collapses that into a single click. A row of equal-height bars appears beside the model button—think of it as a volume fader for depth of thought—each bar representing one level, gradient-filled up to the active setting. Click a bar and the level changes instantly, no menu in sight.

The model menu itself is streamlined too. It opens straight into a searchable list with provider filter chips, a result counter, and adaptive height, reusing the full retrieval capabilities of dsh-model-picker-search. Models without reasoning levels simply get no bars. Full keyboard navigation, radiogroup semantics, and hover tooltips are included. The plugin registers in the conversation.input.model slot at priority −2, so it layers cleanly over the stock picker and dsh-model-picker-search, and gracefully falls back if it is removed or fails to render.

Built for DeepSeek Harness users who toggle between Off, Medium, High, and Max often—especially those juggling multiple models and providers who want to control thinking depth with the fewest possible interactions. Localised in Russian, English, and Chinese.

Use Cases

  • Cut down clicks when toggling reasoning depth to a single bar tap
  • Repeatedly adjusting thinking depth between Off, Medium, High, and Max
  • Quickly locating and switching effort levels across multiple models and providers

Best For

  • Daily DeepSeek Harness users who toggle reasoning depth often
  • Heavy desktop users managing multiple models and providers
  • Developers who prefer minimal-step interactions for maximum efficiency