AI Agent Hub
Back to plugins
🖥️

dsh-model-router

Client Updated 2026.08.26

Run the following command in DeepSeek Harness:

dsh plugin install fonlan/dsh-model-router

Paste the following prompt into your AI chat to install this plugin:

Run dsh plugin install fonlan/dsh-model-router in your DeepSeek Harness terminal to install the plugin; source code is available at https://github.com/fonlan/dsh-model-router

About this plugin

When you run several LLM providers in parallel, every model is locked to a single vendor and switching means editing config and restarting the profile. dsh-model-router registers a virtual provider called model-router inside DSH that merges all your configured providers and models by model id into one unified list in the model picker, with each entry showing which provider is currently active, giving you a single entry point to manage every model.

Routing resolves at request time: the session always carries provider=model-router, the process picks the active provider in-memory and delegates to the real adapter so credential lookup, retries, streaming, and metering all flow through the original pipeline. Main agent, sub-agents, and tool LLM calls are treated equally. You can switch providers from the settings card or directly from the model picker dropdown, and the change takes effect globally on the very next request across all sessions.

The plugin also handles model-ID prefix merging, automatic cleanup and active-provider fallback when providers are added or removed, configurable display order (custom, alphabetical, or most-recently-used), and keeps core routing logic decoupled from the UI so web, CLI, and headless profiles all route identically out of the box. It is built for developers who maintain multiple LLM services and want to swap vendors on the fly without repeatedly hand-editing YAML.

Use Cases

  • Juggling multiple LLM vendors and switching the active provider per model without editing YAML
  • Toggling providers directly from the model picker so the very next request across all sessions goes to the new vendor
  • Collapsing identical models from different providers into one entry with the active source labelled

Best For

  • Developers maintaining multiple LLM services who need on-the-fly vendor switching
  • Teams running DSH across web, CLI, and headless profiles with consistent routing behaviour
  • Ops engineers who prefer instant provider switching over hand-editing config files