AI Agent Hub
Back to plugins
🤖

dsh-peak-pricing

Model Inference Updated 2026.08.27

Run the following command in DeepSeek Harness:

dsh plugin install Oliver0804/dsh-peak-pricing

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

Run dsh plugin install Oliver0804/dsh-peak-pricing (source: https://github.com/Oliver0804/dsh-peak-pricing), then restart dsh web to see the real-time pricing row under the composer.

About this plugin

DeepSeek prices by time of day in Beijing: weekdays 09:00-12:00 and 14:00-18:00 are peak, everything else is off-peak at exactly half the price. Running a long session that crosses 18:00 means you cannot tell which tokens were billed at which rate until the invoice arrives. dsh-peak-pricing puts that answer right under the composer: a real-time indicator of the current period, a countdown to the second, and the running session cost, with a full breakdown on hover.

The core mechanic is incremental pricing. Rather than repricing the cumulative total at whatever rate is in force now, the plugin diffs each new token-usage reading against the previous one and prices only the delta at the rate that was active when those tokens arrived. A session spanning 17:55 to 18:05 therefore splits cleanly into ten peak minutes and five off-peak minutes. Accumulated totals persist in localStorage keyed by session id, so a page reload continues the tally instead of resetting. The hover panel shows per-model input (cache hit and miss) and output rates, four cumulative token buckets, the peak windows for the week, and a sparkline of the cache-hit rate over the most recent 120 input-bearing increments.

Built for developers who regularly run long generation sessions that drift across the peak boundary, teams that want to pace batch jobs around the cheaper off-peak window, and anyone mixing DeepSeek with other providers who needs a single glance at per-currency subtotals without switching dashboards.

Use Cases

  • Check whether you are in a peak or off-peak window before kicking off a long generation run
  • Auto-split billing when a session crosses the 18:00 boundary into the opposite rate tier
  • See per-currency subtotals when a single session touches both DeepSeek and another provider

Best For

  • Developers who routinely run long sessions that drift across the peak-off-peak boundary
  • Teams that schedule batch jobs to land inside the cheaper off-peak window
  • Engineers mixing multiple model providers who want a single consolidated cost view