jiay98528-dev/dsh-model-sync

dsh-model-sync is a DSH-native plugin that adds two custom settings pages to DeepSeek Harness, for syncing model lists and tracking API usage quota. One page handles syncing the latest available model IDs from each LLM provider directly into your DSH provider settings, so you don’t have to manually add new model IDs when providers update their catalogs. You can install it via the DSH CLI using either the public npm package or directly from the GitHub source, and it activates automatically after installation without extra configuration. Its UI matches DSH’s active language setting, supporting both English and Chinese out of the box.

To sync new models, navigate to Settings → Model Sync, then click Refresh to pull the latest list from your configured providers. New model IDs not already in your DSH settings are marked with a blue outline, and you can apply all new models in one click to add them to your provider configuration. The plugin also adds a colored progress ring next to the active model name in the DSH composer, which updates automatically to match the current session’s model. The usage tracking page shows 5-hour and 7-day quota windows for most providers, with colored progress bars, reset countdown timers, and current balance for metered accounts.

This plugin is released under the permissive MIT license, and it uses the API credentials you already have stored in DSH, so it never stores your API keys separately. It currently supports quota tracking for 7 major LLM providers including OpenAI Codex, Kimi Coding, DeepSeek, MiniMax, Z.ai, XAI, and Xiaomi, with custom endpoints for each provider’s quota API. It polls for updated quota data every 60 seconds by default, though you can adjust the polling interval in the plugin configuration. It only works with the DSH web profile, and it is bound to DSH 0.1.0-rc.6 composer class names, so changes to the plugin require a dsh web restart to take effect.