deepseek-harness-model-config: A Model Configuration DSH plugin for DeepSeek Harness
marvekg/deepseek-harness-model-config
Adds a model advanced configuration page to DeepSeek Harness Web UI for custom endpoints and capability settings.
Adds a model advanced configuration page to the DeepSeek Harness Web UI, supporting custom endpoints, protocols, and model capability parameters.
dsh plugin --profile web add github:marvekg/deepseek-harness-model-config- Category
- Models & Reasoning
- Platform
- DSH-Plugin
- Author
- marvekg
- Distribution
- Plugin
deepseek-harness-model-config Key Features
deepseek-harness-model-config Repository
deepseek-harness-model-config Repo Summary
deepseek-harness-model-config is a model configuration plugin (DSH plugin) for DeepSeek Harness, adding a "Model Advanced Configuration" page to the Web UI to solve the pain point of custom model endpoints and capability parameter configuration. Maintained by MarvekG under the MIT license, last updated in 2026-08, it has 23 stars. The plugin supports creating custom endpoints (name, URL, API Key, protocol), configuring custom request headers at the provider level, and is compatible with openai-completions, openai-responses, and anthropic-messages protocols. It fetches candidate models via a unified GET /models flow, supports select all, invert selection, and select none, and can complete missing context window, max output, input modalities, and reasoning capabilities from models.dev. The UI text is bilingual (Chinese and English), following the Harness language setting, and API Keys are only written through the credential store, never exposed in configuration previews.
What core features does the deepseek-harness-model-config DSH plugin support?
This DSH plugin provides a complete model management experience: when adding an endpoint, you can set name, URL, API Key, and protocol; custom headers can be configured at the provider level and applied to model requests, but "Fetch Available Models" still uses the API Key. Before saving, you can edit each model's capacity, text/image input, and reasoningEfforts, and preview the configuration without keys. Additionally, the plugin automatically completes missing context window, max output, and other parameters from models.dev, reducing manual input.
How to install deepseek-harness-model-config?
Before installation, ensure dsh is installed and runnable. Install from GitHub by default:
dsh plugin --profile web add github:MarvekG/deepseek-harness-model-config
dsh web
After installation, open "Settings → Model Advanced Configuration" in the Web UI. To pin a version, append a commit SHA to the repository address, e.g., github:MarvekG/deepseek-harness-model-config#<sha>.
How to debug locally and uninstall deepseek-harness-model-config?
For local debugging, clone the repository and install with a local link: dependency in the root:
dsh plugin --profile web add .
dsh web
To uninstall, remove the plugin from the Web profile:
dsh plugin --profile web remove dsh-models-config-plugin
Updates use the "uninstall old version, then install new version" approach; for local debugging, replace the second command with dsh plugin --profile web add ..
Which DeepSeek Harness versions and platforms are compatible with deepseek-harness-model-config?
This plugin targets the DeepSeek Harness Web UI, compatible with openai-completions, openai-responses, and anthropic-messages protocols, suitable for scenarios requiring custom model endpoints. Maintained by MarvekG under the MIT license, last updated in 2026-08, it has 23 stars and is part of the dsh-plugin ecosystem as a verified plugin.