Introduction

DeepSeek Harness (DSH) extends client capabilities as a plugin. For agent developers, a common requirement is to connect an existing OpenAI-compatible provider to DSH, or to configure the provider, model, and API key on a custom endpoint.

loongport is a DSH Cordis bundle and setup CLI maintained by SailingLoong. It targets verified LoongPort OpenAI-compatible providers but also provides an advanced CLI for custom OpenAI-compatible endpoints.

What is this

The package name is loongport, version 0.2.0, licensed under MIT, and requires Node.js >=20.

It primarily addresses two things:

  1. Provides a configuration entry in DSH via Settings → LoongPort.
  2. Provides a CLI to write custom OpenAI-compatible endpoints into the DSH configuration.

Core Features

Below are the main capabilities currently provided by the plugin.

  • Provides the DeepSeek Harness (DSH) Cordis bundle.
  • Targets verified LoongPort OpenAI-compatible providers.
  • Provides an advanced CLI for custom OpenAI-compatible endpoints.
  • Supports selecting a provider, pasting an API key, and choosing a model in DSH’s Settings → LoongPort.
  • CLI supports --base-url, --model, --route, --credential-name, and --write.
  • --model can be reused to register multiple models.
  • The CLI performs a dry run by default; adding --write updates the DSH configuration.
  • Uses DSH’s built-in llm-pi-ai transport, without implementing streaming, retries, or tool calls itself.

Installation and Usage

First, install it to the currently active DSH profile:

dsh plugin --profile <profile> add loongport

After installation, open Settings → LoongPort in DSH and configure as follows:

  1. Select a provider
  2. Paste the API key
  3. Select deepseek-v4-flash or deepseek-v4-pro
  4. Save

Custom Endpoint Usage

If your endpoint is a manually provided OpenAI-compatible endpoint, you can run the CLI directly:

LOONGPORT_API_KEY='your-api-key' npx loongport dsh setup \
  --base-url https://relay.example.com/v1 \
  --model model-id \
  --write

This example writes a provider route and the corresponding credential to DSH. The API key is read only from LOONGPORT_API_KEY and cannot be passed as a command-line argument.

By default, the command performs a dry run. It only updates the DSH configuration after adding --write. If you need to register multiple models, you can reuse the --model flag.

The relevant CLI options include:

--base-url
--model
--route
--credential-name
--write

Configuration Writing

The CLI prioritizes $DSH_HOME; if not set, it uses ~/.dsh.

When using --write, it writes:

  • A selected llm-pi-ai.providers.<route> entry in settings.yaml
  • A selected credential name and API key in .credentials.yaml (owner-only)

Other DSH settings, provider routes, and credentials are preserved.

Applicable Scenarios and Notes

Suitable for the following scenarios:

  • Need to complete provider selection, API key entry, and model selection within DSH’s Settings → LoongPort.
  • Already have a manually provided OpenAI-compatible endpoint and want to explicitly write it into the DSH configuration.
  • Want to reuse DSH’s built-in llm-pi-ai transport instead of re-implementing streaming, retries, or tool calls within the plugin.

Please note:

  • Version 0.2.0 does not perform automatic browser authorization, does not collect browser sessions, and does not access cookies or local storage.
  • Registration/Sign-in and API key creation are still completed on the provider’s website.
  • VeriDrop data is for display-only health observation and cannot alter the provider configuration.
  • The plugin runs with the permissions of the current dsh process; please check the source code and license before installing.
  • The community directory is an independent site with no official affiliation with DeepSeek or SailingLoong.

Links

  • Directory Page: https://www.skillhub.cn/plugins/SailingLoong/loongport-dsh
  • GitHub: https://github.com/SailingLoong/loongport-dsh