dsh-usage-analytics

If you call multiple models and trigger various skills in DeepSeek Harness every day but can never quite tell how many Tokens you actually used, on which models, or which skills keep failing, dsh-usage-analytics is built for exactly that heavy-usage, low-visibility gap. This personal analytics plugin aggregates your local session data into an intuitive dashboard under Settings > Personal Analytics, so you can review your usage rhythm at a glance.

Core capabilities include switchable 1-day, 7-day, and 30-day windows with hourly bar charts, a 7-day activity heatmap, and a clickable 30-day calendar. The Token composition bar breaks down input, output, cache read, and cache write, with reasoning shown as a subset of output rather than double-counted. Model distribution groups by provider with call counts and usage shares. Skill invocations are split between auto-triggered and user-explicit calls, reporting success, failure, and incomplete counts along with success rates. All statistics are derived from local DSH sessions, support one-click index rebuilds, and auto-expire after a 90-day window.

The plugin writes only derived indices — timestamps, numeric Token counts, model identifiers, and skill names — and never stores prompts, model replies, tool parameters, or keys. It is ideal for DSH Web users who want to understand their personal usage structure and refine their skill invocation strategy without sending data to any external service.