dsh-token-usage

In AI model inference, token consumption is a critical metric for assessing usage costs and efficiency, yet native environments often lack intuitive monitoring tools, leaving users unable to track expenses and optimize resources in real time. The dsh-token-usage plugin addresses this pain point by seamlessly integrating a token usage statistics panel into the DeepSeek Harness Web UI, making data readily accessible and actionable.

The plugin’s core strength lies in its multi-dimensional visual analysis: it automatically categorizes token consumption by API provider, model, and date, enhanced with smooth animations (e.g., number scrolling, chart growth) for an engaging experience. The panel features overview statistics cards, trend charts, daily heatmaps, model summary tables, and data export functions, alongside automatic price calculation and manual overrides to ensure accurate cost estimation. All data is persisted, and the interface automatically adapts to dark/light themes, blending perfectly with the DSH ecosystem.

This tool is especially valuable for AI developers, researchers, or team managers. It enables users to monitor usage trends, identify high-consumption models, and export reports for analysis, thereby effectively controlling costs and optimizing inference workflows. Whether for individual experiments or team collaborations, it delivers clear insights that transform token management from a challenge into a straightforward task.