Introduction¶

DeepSeek Harness (DSH) uses a plugin architecture and extends its Web UI functionality through plugins. When debugging or using DeepSeek series models, models often declare their supported “thinking levels” or reasoning effort levels. Manually filtering these levels in the UI can sometimes be inefficient. The dsh-thinking-level plugin addresses this scenario by providing filtering and default-value configuration.

Plugin Purpose¶

nyzeep/dsh-thinking-level is a thinking-level selector that runs on the DeepSeek Harness Web UI side. It is maintained by developer Nyzeep. It is primarily used to resolve the issue of cluttered level lists in the UI, ensuring users only see the levels declared and preferred by the current model.

Core Features¶

• Level Filtering: Based on the current model’s state, only display the thinking levels declared and preferred by that model.
• Default Setting: Selects the highest level by default.

Installation and Enablement¶

No official installation command is currently provided. Please obtain the source code through the plugin directory or GitHub repository, and enable it using Harness’s generic plugin mechanism.

Use Cases and Notes¶

• Use Cases: Developers who need to precisely control reasoning effort levels and want to keep the UI clean.
• Notes:
  • The plugin runs with the permissions of the current DSH process.
  • Since the license and specific functional details cannot be confirmed, be sure to review the source code before installation.

Conclusion¶

This plugin provides DSH users with more fine-grained level control. For more details, refer to the plugin directory or source repository.