dsh-token-heatmap

At the end of each billing cycle, do you ever wonder which days or models actually consumed the most tokens? dsh-token-heatmap renders your local session logs into a GitHub-style usage heatmap embedded directly in the DSH Web settings page—entirely offline, with zero calls to any billing or balance API.

The plugin offers four switchable views: weekly (hourly bar columns), monthly (calendar grid), quarterly (90-day lattice), and yearly (full-year two-row layout). Colors scale across four intensity tiers within each window, while a summary bar tracks today, this week, this month, and cumulative totals. Top 5 rankings switch dynamically between model and vendor, auto-refresh can be set anywhere from 5 to 60 minutes or disabled entirely, and the interface toggles between Chinese and English. All history is persisted to disk, so deleting a session never erases your statistics.

Built for developers and AI engineers who run multi-model inference through DSH Web and want a private, local-first, zero-network way to inspect their token consumption patterns, peak hours, and model-level spending.