AI Agent Hub
Back to plugins
dsh-token-use preview

dsh-token-use

admin-security Updated 2026.09.15

Run the following command in DeepSeek Harness:

dsh plugin install huangyuheng/dsh-token-use

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

Run dsh plugin install huangyuheng/dsh-token-use in the DeepSeek Harness terminal to install; the source code is hosted at https://github.com/huangyuheng/dsh-token-use. Restart dsh web after installation and the Token Usage panel will appear under Settings.

About this plugin

You burn tokens every day, yet you rarely see where the spend lands: which project costs the most, which model hogs the cache, whether this week is up or down from last week — those answers sit in server-side logs nobody opens on purpose.

dsh-token-use turns that invisible cost into numbers you can read at a glance. One click after install and a first-level entry appears under Settings, requiring zero configuration. The overview tab pairs a stacked-area trend chart (with a cost bar on the right axis and a daily-average reference line) with a 12-month usage heatmap; dedicated model and project tables let you slice data by day, month, year, or a custom date range. Estimated cost is computed per record against official DeepSeek pricing, applying peak-versus-off-peak multipliers, while non-DeepSeek models are deliberately left unpriced and shown with a dash. The configuration tab ships a three-part self-diagnostic — API latency, client-bundle reachability, and Harness directory integrity — plus a one-click restart with a confirmation dialog, so when something breaks you get a concrete file-level answer instead of a stack trace.

It is built for DeepSeek Harness users who juggle multiple projects, watch cache-hit and reasoning ratios, and want invisible costs turned into visible numbers. On the host side the plugin uses event-bus O(1) increments with no polling, no extra timers, and no disk writes; historical backfill runs in a Worker thread so active sessions are never blocked.

Screenshots

Use Cases

  • Identify the most expensive project and model across parallel workloads
  • Review weekly or monthly spending trends to adjust usage strategy
  • Pinpoint missing files or version mismatches via the three-part health diagnostic

Best For

  • Developers juggling multiple DeepSeek projects in parallel
  • Tech leads tracking cache-hit ratios and reasoning token breakdowns
  • Teams that need to convert invisible token spend into quantifiable data