Problem
LLMeter benchmarks produce several metrics that are fractions of a total — error rate, cache hit rate, throttle rate, SLA compliance, etc. Currently there's no standard chart type for these. Users end up building ad-hoc bar charts or just printing numbers, losing the spatial context of where a value sits in the 0–100% range.
Additionally, the plotting module hardcodes "plotly_white" as the template string in every function independently. There's no centralized way to change the theme or to get a consistent color cycle across custom charts built alongside LLMeter's built-in visualizations.
What would be useful
- A dedicated visualization primitive for percentage metrics that immediately conveys how a rate sits relative to the full 0–100% range.
- A centralized place for theme and color configuration so all charts are visually consistent and users can switch themes without patching every function.
Problem
LLMeter benchmarks produce several metrics that are fractions of a total — error rate, cache hit rate, throttle rate, SLA compliance, etc. Currently there's no standard chart type for these. Users end up building ad-hoc bar charts or just printing numbers, losing the spatial context of where a value sits in the 0–100% range.
Additionally, the plotting module hardcodes
"plotly_white"as the template string in every function independently. There's no centralized way to change the theme or to get a consistent color cycle across custom charts built alongside LLMeter's built-in visualizations.What would be useful