MarvekG/deepseek-harness-model-config

This is a native plugin built exclusively for DeepSeek Harness (DSH), that adds an advanced model configuration page to the DSH web user interface. It lets DSH users add custom model endpoints, fill in basic connection details like endpoint name, base URL, and API key, and configure per-model capability parameters. It supports multiple common API protocols including openai-completions, openai-responses, and anthropic-messages, and allows custom request headers to be added for all model requests. To install the plugin, you first need a working DSH installation, and you can install it directly via the DSH CLI command.

After adding an endpoint, users can fetch the list of available models through a unified GET /models endpoint flow. They can search for models by ID or display name, and use bulk selection tools like select all, invert selection, and clear all to quickly pick the models they want. If the model list cannot be fetched, users can manually enter model IDs to add models directly, and duplicate IDs will automatically match existing entries. All endpoint edits use a shared editor, and existing checked models retain their selection status after you refresh the model list.

To install this plugin, you run a DSH CLI command pointing to the GitHub repository, and you can pin a specific version by appending a commit SHA to the repository address. For local development, you can clone the repository and install it via a local path link with the DSH CLI. Automatic parameter completion for models pulls data from models.dev to fill in context window size, input modality, and other common parameters. The project is released under the open source MIT license, and API keys are only stored via DSH's built-in credential system. Uninstalling or updating the plugin is also done via the DSH CLI remove and add commands.