
MEMZ-JZY/DSH-Claude-Style-Reasoning-Slider
dsh-client-ui-effort-slider
dsh-client-ui-effort-slider is a community DeepSeek Harness plugin. Read the repository README before installing.
Install
npx @deepseek-ai/dsh plugin --profile web add github:MEMZ-JZY/DSH-Claude-Style-Reasoning-SliderRestart `dsh web` after install. Bundle APIs can change during the developer preview.
README badge
[](https://dshhub.dev/plugins/dsh-claude-style-reasoning-slider)Paste this into your README. The star count updates with every catalog sync.
From the README
Excerpt from MEMZ-JZY/DSH-Claude-Style-Reasoning-Slider, cleaned of badges and images.
DSH Claude Style Reasoning Slider
An animated reasoning-effort slider and model picker for DeepSeek Harness (DSH).
It replaces the native model selector's conversation.input.model slot with a
compatible selector whose reasoning pane is a Claude-style animated slider.
π¨π³ δΈζηθ―΄ζθ―·θ§ README.zh.md
Preview
Features
- Drop-in model selector β replaces only the
conversation.input.modelslot; the rest of the chat input stays untouched. - Canonical effort slider β always shows
Off | Low | Medium | High | Extra | Maxpositions, regardless of how the provider names its levels. - Tolerant name matching β aliases such as
off,none,disabled,ultracodeβmax,medβmedium,extremeβextra, and similar display names are mapped to canonical slider positions. - Unsupported levels stay visible β a level the current model does not support is dimmed but remains clickable. Clicking it keeps the thumb on the chosen slot and applies the nearest supported level below it, with a toast explaining the downgrade.
- Model switching preserves effort β when you switch models, the current effort is kept when possible, or automatically downgraded to the nearest supported level below it (with a toast).
- Default and adapter-specific strengths β
Defaultis offered as a pill below the slider and submits withoutreasoningEffort(the provider's own default applies). Other provider strengths that do not map to a canonical slider position are also shown as selectable pills. - Models without reasoning metadata stay manageable β when DSH does not
expose
reasoning/effortsfor the selected model (for example some OpenRouter or hand-declared custom entries), the Effort menu still appears: it shows a short βno effort levelsβ notice plus the safeDefaultpill, instead of hiding every thinking-strength control. - Liang Calibrator (ζ»ε¨εη₯ε¨) β optional, off by default. Binds the six
canonical levels to stage names:
Off β ε°ιΎζ’,Low β η’ζ’,Medium β ζ’ε,High β ζ’ε£,Extra β ζ’η₯,Max β ζ’η₯. When enabled, a 31-frame portrait (frame-00β¦frame-30) appears above the slider and follows the thumb continuously while dragging, then snaps to the segment's first frame on release. Labels and the trigger showMax ζ’η₯-style suffixes. The setting persists inlocalStorage. - Big Fat Fish thumb β optional, off by default. Replaces the slider thumb
with an 8-frame running chibi fish sprite. The idle animation loops at
720ms, speeds up to420mswhile dragging, and freezes underprefers-reduced-motion. Toggle it in DSH Settings β General β Big Fat Fish slider; changes apply immediately and persist inlocalStorage. - High / Extra ripple field β
HighandExtranow show a sparse blue / purple dot-matrix field with random flicker and radial water ripples around the thumb, as a lighter prelude to theMaxpixel field. - Max / Ultracode treatment β the model's
max/ultracodelevel turns the slider into an animated purple pixel field, and theMaxlabel becomes a flowing multi-color gradient text. - Localized and theme-aware β ships with English and Simplified Chinese dictionaries and follows DSH light/dark design tokens.
- Accessible β keyboard-operable slider (arrows, Home/End, PageUp/PageDown),
ARIA labels, focus management, and
prefers-reduced-motionsupport.
Install
β¦

