AI Agent Hub
Back to plugins
dsh-model-palette preview

dsh-model-palette

Model Inference Updated 2026.08.27

Run the following command in DeepSeek Harness:

dsh plugin install Jensen-Yao/dsh-model-palette

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

Browse the plugin source at https://github.com/Jensen-Yao/dsh-model-palette, run dsh plugin install Jensen-Yao/dsh-model-palette in your terminal, restart dsh web, and press Alt+M to open the palette.

About this plugin

Switching between providers and tuning model parameters in DeepSeek Harness Web used to mean scrolling through a limited native selector, hunting down protocol details, and editing config files by hand. dsh-model-palette replaces that friction with a single global command palette: press Alt+M from anywhere, type a few characters, and fuzzy search resolves model names, model IDs, provider names, and provider IDs at once. A provider filter rail on the left shows live model counts, while starred favorites and recently used models stay pinned to the top so your go-to model is always one Enter press away.

Beyond quick switching, the plugin turns the Web UI into a full configuration workspace. Add, edit, or duplicate provider profiles with base URL, protocol, and credential reference in a few clicks; test connectivity and pull live model metadata in one step; auto-fill context windows, max outputs, and input modalities from 44 bundled presets that refresh online; and set reasoning effort from off to max per model or across an entire route. Optional OpenRouter media tools let you generate images, submit async video jobs, and download results without leaking prompts into the conversation.

This is the plugin for DeepSeek Harness Web users juggling OpenAI, DeepSeek, Qwen, GLM, Together, OpenRouter, or any custom OpenAI-compatible gateway. If you regularly adjust reasoning levels, repair DeepSeek-dialect replay fields, validate protocol compatibility, or need a loopback-only B.AI relay to work around network blocks, dsh-model-palette has you covered. Security is built in: API keys are masked by default, credential reveal is restricted to 127.0.0.1 connections, and every plugin API endpoint ships with CSRF protection, so your multi-provider workflow stays fast without sacrificing safety.

Screenshots

Use Cases

  • Fuzzy-search and switch models across OpenAI, DeepSeek, Qwen, and other providers in one keystroke
  • Bulk-configure reasoning effort, context windows, and input modalities instead of editing each model manually
  • Generate images or submit async video jobs via OpenRouter without leaving the Web UI

Best For

  • DeepSeek Harness Web users juggling multiple LLM providers simultaneously
  • Developers who regularly tune reasoning levels, protocol compatibility, and dialect replay fields
  • Content creators who prefer in-browser image and video generation over standalone scripts