AI Agent Hub
Back to plugins
🧰

dsh-better-model-select

Web Tools Updated 2026.09.04

Run the following command in DeepSeek Harness:

dsh plugin install sxy-kumako/dsh-better-model-select

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

Run dsh plugin install sxy-kumako/dsh-better-model-select in your terminal to install the plugin; the full source is available at https://github.com/sxy-kumako/dsh-better-model-select, then restart DSH Desktop to apply.

About this plugin

DSH Desktop silently falls back to a static, build-time catalog for providers that already ship in the bundled directory, such as GLM Coding and MiniMax. It never calls the remote model-list API, so brand-new model versions stay invisible until an upstream release arrives, and users get no at-a-glance view of live quotas or balances across their providers. dsh-better-model-select closes both gaps.

Every 30 minutes (and on startup) the plugin reads each configured provider route, resolves the API key from the DSH credential store, and fetches the live model list. It supports both OpenAI-compatible and Anthropic-compatible endpoints. Newly appeared models are merged into the route settings as fully formed catalog entries covering context window, input modality, wire compatibility, and reasoning capability. Existing directory entries and any user hand-written overrides are always preserved. A failed remote call writes nothing, and the model selector reflects the update immediately with no proxy, no baseURL change, and no restart required.

Below the native session stats line the plugin mounts a row of quota badges. GLM Coding and MiniMax Coding display a 5-hour window and a weekly-window remainder; DeepSeek, OpenRouter, and Kimi show a cash balance. Colors shift from green to amber to red as the remainder drops below 30 percent and 10 percent. All credential lookups stay inside the DSH credential store and are never written to logs.

If you run DSH with multiple LLM providers and want the model picker to always mirror the latest remote catalog while keeping a live eye on remaining quota for each one, this plugin is built for you.

Use Cases

  • New model versions appear in the picker automatically, no upstream catalog bump needed
  • Spot at a glance which provider balance or window is running low
  • See the latest remote model list without restart or baseURL changes

Best For

  • DSH Desktop users who configure multiple LLM providers at the same time
  • Indie developers who need real-time quota visibility across providers
  • Teams that want the model list to always track the latest remote releases without manual catalog upkeep