dsh-v-token-insight
Run the following command in DeepSeek Harness:
dsh plugin install victor10035445/dsh-v-token-insight
Paste the following prompt into your AI chat to install this plugin:
Run dsh plugin install victor10035445/dsh-v-token-insight in your DeepSeek Harness terminal or web console to install the plugin; the source repository is available at https://github.com/victor10035445/dsh-v-token-insight . After installation, restart dsh web and refresh the page to see the Token Stats entry in the sidebar.
About this plugin
You talk to DeepSeek every day, yet the per-turn token counts, cache-hit ratios, and the dollar figure behind each reply stay buried in logs. dsh-v-token-insight lifts all of that into the DeepSeek Harness Web client where you can see it at a glance. A new sidebar entry opens a full-page cross-session dashboard with KPIs for billed tokens, input/output split, cache hit rate, session count, and estimated cost, plus five drill-down tabs: consumption trend, per-workspace, per-session detail table (search, sort, resizable columns, right-click jump), time-based view, model distribution, and a full price sheet. On any conversation page a third tab labeled "Session Token Stats" dives into every turn and every step, showing the four-bucket token split, per-step model and call parameters, a context-pressure bar, and an instant fee estimate.
The plugin is strictly client-side and makes zero kernel changes: it replaces no official plugin and reads exclusively from the host existing projections and session windows. Every chart is hand-rolled SVG with no chart-library dependency; colors follow the official theme tokens so light and dark modes stay in sync, and prefers-reduced-motion is respected. Pricing runs a three-tier pipeline-fair per-bucket unit prices, dual-dimension tiers by context scale and cumulative session volume, then time-window discounts-recomputing step by step instead of the rough total-times-rate shortcut. Steps whose currency does not match are flagged as undpriced rather than guessed. The editable price table lives on the overview page; change a number and every fee updates instantly.
It is built for DeepSeek Harness Web users who juggle multiple workspaces and models and want token spend to be as auditable as a line-item invoice. All data persists in browser localStorage with no backend; a single click rebuilds the ledger after a data wipe. Every record is keyed by (session, turn, step), so even archived conversations keep a complete, traceable history.
Use Cases
- Track token consumption and estimated costs across multiple DeepSeek workspaces in real time
- Audit cache-hit ratios and model distribution to refine context strategies and reduce spend
- Trace and reconcile token usage of archived conversations for post-hoc cost accounting
Best For
- DeepSeek Harness Web users who juggle multiple workspaces and switch between models daily
- Dev teams sensitive to API costs who want auditable per-turn billing details
- Technicians who prefer zero-dependency lightweight plugins and avoid extra chart-library overhead
Related Plugins
A unified suite combining hot runtime injection, task-aware thinking-mode routing, and a graded session protocol with red-team gates to sustain model diligence across long-horizon inference.
ModLens is a vision plugin for DeepSeek Harness that gives text-only models sight by reading images pasted directly into chat, with zero-config setup and multiple vision engines.
On-demand vision for text-only DeepSeek Harness agents: built-in free keyless vision chain and 14 vision tools, routing image turns as tool calls to vision models with pixel fidelity, no Python needed, one-command install.
Give text-only models in DeepSeek Harness eyes, enabling image Q&A, long-screenshot OCR, UI restoration, and GUI visual tasks.