dsh-usage-insights

Every DSH user knows tokens cost money, but few can say exactly where they went. dsh-usage-insights opens that black box in Settings to Personal Analytics: switch between 1-day, 7-day and 30-day windows to see hourly bar charts, heat maps, a clickable monthly calendar, and a stacked token composition bar that breaks down input, output, cache read, cache write, and reasoning usage.

Beyond tokens, the plugin tracks model distribution and skill invocations. A model card lists each provider with call count, token volume, and share, expandable into a full list. Skill calls are split into auto-triggered and user-initiated, each reporting success, failure, and pending counts with an overall success rate. All date and hour details render in your local time zone, and reasoning tokens are displayed as a subset of output without double counting.

On the data side, the plugin writes only its own derived index (timestamps, numeric token counts, model ids, skill names) and never stores prompts, model replies, tool parameters, or secrets. Details are pruned on a 90-day schedule. It is designed for Web-platform users who want to quantify personal DSH usage, identify which model or skill consumes the most, and maintain a clear privacy boundary.