dsh-advanced-model-editor

When using DeepSeek Harness for model inference, complex configuration needs often exceed the capacity of the default settings page. Many users require fine-grained management of LLM provider configurations, including reasoning/thinking options, custom request headers, and retry policies, which the official interface often struggles to handle simultaneously across different namespaces (such as llm-deepseek and llm-pi-ai).

dsh-advanced-model-editor plugin is designed specifically to address this pain point. By adding a dedicated "Advanced Model Settings" section to the settings page, it offers a more flexible configuration entry point. The plugin not only supports managing both built-in and custom providers but also allows users to adjust reasoning parameters, set request headers, and define retry policies for specific models. Crucially, it fully inherits the validation and conflict rules of the official settings UI, ensuring configuration safety and consistency.

This tool is ideal for developers and power users who require deep customization of large language model inference behavior. Whether you wish to enable a model's "thinking" mode or need to inject custom headers for specific requests, this plugin significantly enhances configuration convenience and functionality without breaking system rules, allowing DeepSeek Harness to better align with your specific business scenarios.