When developing or deploying models locally, distinguishing between Docker Desktop and the WSL engine, and confirming whether a vLLM container is running and accessible, usually requires manual checks. The dsh-wsl-docker plugin for DeepSeek Harness (DSH) provides automated diagnostic capabilities.

Plugin Introduction

This is a DeepSeek Harness plugin for diagnosing Docker context in a WSL environment and providing status hints for vLLM or OpenAI :8000 containers. The plugin is maintained by user 173787247 and is provided under the MIT License.

Core Features

The plugin primarily provides the docker_doctor tool, supporting the following three inspection focuses:

  1. focus=all: Checks the Docker CLI path, current context, server version, and NVIDIA runtime hints.
  2. focus=daemon: Focuses on the configuration and environment information of the Docker daemon itself.
  3. focus=vllm: Lists running containers and filters out containers named similar to OpenAI/:8000. This helps confirm whether the local inference port is ready.

Additionally, the plugin recommends using this plugin together with the host_reach service when configuring the baseURL of DeepSeek Harness.

Installation and Enabling

The plugin is included in the KIT_SET=llm and full configuration sets and can also be installed individually.

  1. Run the installation command:
    dsh plugin --profile web add github:173787247/dsh-wsl-docker
  1. After installation is complete, restart the dsh web service and open a new session.

Typical Usage

After successful installation, you can use docker_doctor in the command line or in tool calls.

  • Check the overall environment:
    docker_doctor focus=all
  • Focus on checking the daemon:
    docker_doctor focus=daemon
  • Focus on checking vLLM-related containers:
    docker_doctor focus=vllm

Applicable Scenarios and Notes

This plugin is suitable for developers debugging local model deployments in a WSL environment.

  • Dependency Requirements: Requires DeepSeek Harness version ≥ 0.1.2 and Node.js ≥ 18.
  • Model Configuration: Cloud Flash models (such as deepseek-flash) should be configured in ~/.dsh/settings.yaml; this plugin does not handle that configuration.
  • Upstream Dependencies: Agent Teams is an upstream experimental feature, and this plugin does not depend on it.

Conclusion

With dsh-wsl-docker, developers can quickly obtain the runtime status of Docker in a WSL environment and reduce troubleshooting time. The complete source code and troubleshooting documentation can be referenced in the project repository.