Introduction¶

In DSH, switching models and reasoning effort usually requires two steps: first changing the model, then adjusting the reasoning effort. With many models and many levels, toggling between sets is annoying. What dsh-2d-model-picker does is combine these two actions into one drag: in a 2D grid, rows are models, columns are reasoning effort levels, and releasing on a specific combination applies the settings. Below is an introduction to the plugin’s positioning, installation, and usage.

The plugin comes from Fyerl/dsh-2d-model-picker on GitHub, by Fyerl, under the MIT license. It replaces DSH’s native model selector, belongs to the “Model Reasoning” category, and is positioned as a purely auxiliary tool at the interaction layer that does not alter the behavior of the model itself.

What is this Plugin¶

One sentence: Replace DSH’s native model selector with a 2D grid, dragging once sets both the model and reasoning effort.

• Maintainer: Fyerl
• License: MIT
• Version: 1.0.1 as stated in package.json
• Code Organization: index.js is the entry point, client.js is the Web client code. There is also cordis.patch.yml for bundle patching and locale/*.json for multilingual text.
• Client Configuration: Points to the Web platform, injecting @deepseek-ai/dsh-client-ui-conversation.

Core Features¶

1. Grid Selection: Rows are models, columns are reasoning effort levels.
2. One-Drag Setup: Set model and reasoning effort in one drag; release on a supported combination to apply.
3. Replacement: Replaces the original model selector without creating a parallel entry.

The demo provided in the README is a GitHub video showing the drag-and-select process on the grid.

Installation and Enablement¶

In DSH, open Plugins → Add plugin and enter:

dsh-2d-model-picker

Or use the terminal, targeting the Web profile:

dsh plugin --profile web add dsh-2d-model-picker

If you want to use the Discover method, you need to first enable Custom plugin indexes in Experimental plugins, then add this index in Plugins → Discover:

https://raw.githubusercontent.com/Fyerl/dsh-2d-model-picker/main/docs/index.json

After installing via any of the above methods, restart or refresh the DSH client to see the grid-style selector.

Typical Usage¶

After installation, the daily process of switching models and reasoning effort is:

1. Open the model selector in the conversation interface; a 2D grid appears.
2. Drag horizontally to change the model, vertically to change the reasoning effort.
3. Stop on a supported cell and release; both settings take effect at once.

There are no additional configuration files to write, and no environment variables to configure. If you find that a certain model or intensity level lacks a corresponding cell, it means that combination is not within the plugin’s supported range.

Suitable Scenarios and Notes¶

Suitable Scenarios:
* Users who frequently switch between multiple models and have clear preferences for reasoning effort.
* People who need a fixed “model + intensity” combination in a workflow and do not want to click twice every time.

Points to Note:

1. The plugin runs with the permissions of the current DSH process. The models and configurations it can access depend on what the process has already connected to; the plugin does not expand the model list on its own.
2. Before installing, it is recommended to review the source code and the MIT license. Confirm that the loaded index.js, client.js, and cordis.patch.yml match your expectations before deciding whether to enable it.
3. It injects UI into the Web client, so the functionality is available under the Web profile; the behavior under other profiles is not explained in the README.
4. The current star count on the repository is low (2 stars displayed on the directory page), so there isn’t much ecosystem accumulation yet. Evaluate whether it is worth introducing before using it.

Conclusion¶

The value of this plugin is focused: turning the two-step selection of “model + reasoning effort” into one drag, saving repetitive operations during each switch. If you have no issues with the native selector, or if there are few model levels to begin with, the benefit is limited; if you frequently adjust between the two, it is worth trying.

Directory Page: https://www.skillhub.cn/plugins/Fyerl/dsh-2d-model-picker
GitHub: https://github.com/Fyerl/dsh-2d-model-picker