ARGUS
AI benchmark comparison with ECharts GL plots and agent execution timelines
You can select a benchmark run in the charts to open its model calls, token usage, and execution timeline. Built the frontend in React and TypeScript with Vite to display imported benchmark run JSON.
Implemented 3D plots of score, token count, and token efficiency with ECharts GL, line and bar charts with Unovis, and selection and filtering controls with Base UI.
Built an execution timeline from imported run data with React, CSS, and SVG. Positioned calls by start time and duration, assigned colors by model, and linked related tasks. Added replay for viewing the sequence and overlap of calls.
ARGUS won the Final Winner award at JunctionX Korea 2026.