AI Agent Hub
Back to plugins
🖥️

dsh-session-stats-panel

Client Updated 2026.08.25

Run the following command in DeepSeek Harness:

dsh plugin install a1113622001/dsh-session-stats-panel

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

Run the following command in DeepSeek Harness to install this plugin; the source repository is https://github.com/a1113622001/dsh-session-stats-panel .

About this plugin

Running an Agent task and wondering how many tokens were actually consumed, how much of that was cache hits, what it cost, and whether the account balance can still sustain a few more runs—these questions usually only surface when the monthly bill arrives. dsh-session-stats-panel mounts a non-invasive panel on the right side of the Harness UI, surfacing token cache hit rate, real-time cost estimates based on DeepSeek official peak/off-peak pricing, remaining account balance, cumulative tokens, and model request step count so that cost and efficiency for every turn are visible at a glance.

The panel ships with the 2026 peak/off-peak pricing rules built in: standard pricing during weekday peak hours, across-the-board 50 percent discount during off-peak windows, and weekend-long off-peak pricing covering models such as deepseek-v4-flash and deepseek-v4-pro. Balance lookups go through a server-side credential-isolation route and refresh every two minutes, keeping the API key entirely out of the browser while staying current.

If you are debugging Agent workflows, tuning Prompt Caching, or need to track inference spend for a team, this dashboard is the small tool you glance at to confirm you are still on track.

Use Cases

  • Monitoring token consumption and cache hit rate while debugging an Agent workflow
  • Comparing peak vs off-peak inference costs for team-level budget tracking
  • Checking account balance before launching a long-running Agent task

Best For

  • Engineers building Agent applications on DeepSeek Harness
  • Developers tuning Prompt Caching and controlling inference spend
  • Team leads tracking multi-model inference costs