AI Agent Hub
Back to plugins
🖥️

llmtrim-stats-plugin

Client Updated 2026.08.26

Run the following command in DeepSeek Harness:

dsh plugin install Zenjibad/llmtrim-stats-plugin

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

Run dsh plugin install Zenjibad/llmtrim-stats-plugin in your terminal to install; source code is at https://github.com/Zenjibad/llmtrim-stats-plugin

About this plugin

You compress your DeepSeek requests with llmtrim every day, but checking how much you have actually saved still means hopping to a terminal and rerunning status. llmtrim-stats-plugin brings that savings data straight into the DSH web UI so the numbers are one glance away in your browser, with no more window-switching.

The plugin offers two layers of visibility. On the Settings page you get a full KPI dashboard: four money cards (You paid, Would have cost, Saved today, Saved this week), total tokens trimmed, request counts, a per-model breakdown of saved percentage and USD, plus a daemon health badge and binary version. Below the chat composer sits a configurable stats strip that can cycle through your chosen metrics in rotating mode (one stat every four seconds) or display all selected stats at once in static mode. Nine individual metrics are freely toggleable, and the strip refreshes on a five-second poll.

Under the hood the plugin is strictly read-only. It shells out to llmtrim status --json, parses the JSON payload, and never writes to llmtrim files, ledger, or daemon state, so every figure matches what the CLI reports. All colors are driven by DSH design tokens for automatic light-dark theme support, and the plugin installs once as a profile bundle that auto-loads on every boot. It is built for developers and small teams who already wire llmtrim into DSH and want their daily compression savings visible at a glance instead of buried in terminal output.

Use Cases

  • Monitor daily llmtrim compression savings in tokens and USD at a glance
  • Rotate key savings metrics in a strip below the chat composer
  • Compare per-model compression savings percentage and cost

Best For

  • Developers already using llmtrim compression in their DeepSeek workflows
  • Small teams needing real-time visibility into API cost savings
  • Users who prefer a web dashboard over frequent terminal checks