Preface

DeepSeek Harness (DSH) adopts an “everything is a plugin” architectural philosophy. In local deployment or private environments, developers may sometimes need to completely disconnect the hardware from the network, or apply strict restrictions to plugin network behavior. In the existing plugin ecosystem, some inference-related plugins may rely on external network resources. The following section introduces dsh-argon, which addresses the need for “model inference in offline environments.”

What It Is

dsh-argon (the element Argon) is a DSH plugin focused on model inference. It is maintained by the developer uckkk. Its core positioning is to provide lightweight inference capabilities that do not depend on a network environment.

Core Features

  • Pure Node implementation: Built on the Node.js ecosystem, with no dependency on other runtime environments.
  • No network dependency: The plugin does not establish any external network connections during execution.

Installation and Activation

In a DSH environment, install the plugin using the following command:

dsh plugin add github:uckkk/dsh-argon

After installation, DSH automatically handles subsequent dependency loading.

Dependencies and Configuration

The plugin has specific runtime dependency requirements. Ensure that the following versions are met:

  • @deepseek-ai/cordis: requires version ^4.0.1 or higher.
  • @deepseek-ai/dsh-tools: requires version >=0.1.0-rc.6 or higher.

Before installation, verify that the current DSH environment already includes the above dependencies.

Use Cases and Notes

  • Use cases: fully offline local deployment environments, intranet environments with strict network policy restrictions, and scenarios requiring self-auditing of plugin source code.
  • Permissions and security: The plugin runs with the permissions of the current DSH process. Before installation, it is recommended to review the code logic via the GitHub repository to ensure compliance with security audit requirements.
  • License: The plugin is open-sourced under the MIT License.

Conclusion

dsh-argon provides a concise, pure Node.js inference solution, suitable for developers who追求 an ultimate offline experience. For more details and source code, refer to its directory page and GitHub repository.