AI Agent Hub
Back to plugins
🖥️

dsh-plugin-cost

Client Updated 2026.09.11

Run the following command in DeepSeek Harness:

dsh plugin install xiaml666/dsh-plugin-cost

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

Run dsh plugin install xiaml666/dsh-plugin-cost in your terminal (source: https://github.com/xiaml666/dsh-plugin-cost), then restart the dsh web process to see per-turn cost, session total, and account balance in the chat UI.

About this plugin

Running multi-turn conversations on the DeepSeek API means invoices arrive weekly, and the DSH chat UI shows no per-turn cost, no running total, and no balance. dsh-plugin-cost embeds spend directly into the conversation: each assistant reply carries its per-turn amount with an expandable breakdown of model routing, call count, and the four non-overlapping token buckets; a session total refreshes below the input box every three seconds; and account balance stays currency-synced with the cumulative total, so switching the primary currency converts the balance with it and you never see a dollar total next to a yuan balance.

Costs are computed from the provider-reported token buckets (cache-miss input, cache-hit, cache-write, output), each multiplied by its own rate, with peak and off-peak windows applied automatically. The plugin parses the official pricing page for model lists, per-bucket rates, and peak windows, then renders both peak and off-peak intervals in your local timezone. When the official site changes its schedule, the next refresh picks it up with no plugin update. Multi-currency display shows a primary and optional secondary code side by side, with FX rates auto-fetched every six hours. Third-party multiplier plans (e.g. a relay service listing a model at 1.20x) map to a known route via a single config entry, and the rate table supports three-level wildcards; unconfigured prices are clearly labelled rather than shown as zero.

Built for solo developers and small teams who run DeepSeek for production work or agent pipelines and need to watch API spend in real time, and for users on third-party relay channels where rates are opaque. Zero runtime dependencies; install, restart the web process once, and costs start showing up immediately.

Use Cases

  • Monitor per-turn and cumulative spend live during multi-turn conversations
  • Verify multiplier rates and token-bucket billing when calling APIs via relay services
  • Track account balance and burn rate to avoid monthly overage

Best For

  • Solo developers running DeepSeek API for production workloads
  • Small teams calling models via relay services or multiplier channels
  • Ops personnel of agent pipelines who need real-time API spend visibility