DeepSeek Harness Configurator

Generate a DeepSeek Harness launch command, provider YAML and optional plugin commands. Preview the local setup without entering API secrets or running a server here.

Launch a local Web profile

Choose your operating system and terminal, port and automatic-browser preference. Install Node.js, then run the generated command in your project directory. Open the printed local URL and choose a workspace in the Web UI.

'npx' '-y' '@deepseek-ai/[email protected]' 'web' '--port' '3080' '--no-open'

The initial example uses the shipped DeepSeek provider and sets its default reasoning effort to high. It does not start a server in this page or require credentials here.

Provider configuration and plugins

- id: llm-deepseek
  config:
    reasoningEffort: high

After first launch, merge the YAML into $DSH_HOME/profiles/web/cordis.patch.yml. The default home is ~/.dsh. Preserve other overrides: a Cordis entry config override replaces its complete config. Configure authentication in your own Harness Models settings or the environment described by your installed adapter.

Custom routes let you choose provider ID, base URL, one of three supported protocols, provider-supplied model ID and key environment variable name. The example endpoint and model are placeholders, not verified service presets. Optional plugin package commands require an exact npm version; REA integration uses its official setup preview.

  • @deepseek-ai/dsh-base (shipped)
  • @deepseek-ai/dsh-web-app (shipped)
  • llm-deepseek provider (shipped)

Developer preview and sources

npm 0.2.0-rc.2, source revision 687f282e8625c6b8d11ff58211a90524be2e4317, reviewed 2026-10-18. Preview releases can make breaking changes. Generated YAML is a configuration document, not proof of a successful model connection or plugin compatibility.

Launch documentation · provider configuration · profile/plugin commands