feat(web-evals): enhance dashboard with dynamic tool columns and UX improvements

[hannesrudolph]

Evals Dashboard UX Improvements

Summary

This PR enhances the web-evals dashboard with better visibility into tool usage metrics, improved exercise selection, and various quality-of-life improvements.

New Features

Run Details Page

• Added aggregate statistics panel showing pass/fail counts, pass rate, total tokens, cost, and duration at a glance
• Tool usage summary with color-coded success rates (green for 100%, yellow for ≥80%, red for <80%)
• Hover tooltips to reveal full tool names from abbreviations

New Run Configuration

• API config selector when importing settings files with multiple configurations
• Language toggle buttons (e.g., "go", "python", "javascript") to quickly select/deselect all exercises for a language
• Concurrency and timeout settings now persist to localStorage across sessions

Runs List

• Clickable table rows for faster navigation to run details
• Dynamic tool usage columns - columns are now generated based on what tools were actually used across runs, sorted by total usage (most-used tools appear first)
• Each tool column displays attempt count and success rate
• Abbreviation tooltips explain column headers (e.g., "RF" → "read_file")
• "View Settings" option in the dropdown menu to inspect run configuration

Improvements

• MultiSelect component now supports controlled mode via value prop
• Deprecated models are now filtered from the Roo Code Cloud model list
• Removed hardcoded tool columns in favor of dynamic generation from run data

Technical Notes

• No breaking changes to existing functionality
• All changes are contained within the apps/web-evals package
• Tool columns use proper ToolName typing from @roo-code/types

---

Important

Enhances web-evals dashboard with dynamic tool columns, improved exercise selection, and UX improvements.

• Behavior:
  • Adds dynamic tool usage columns in run.tsx and runs.tsx, sorted by total usage.
  • Enhances run.tsx with aggregate statistics panel and tool usage summary.
  • Adds language toggle buttons in new-run.tsx for exercise selection.
  • Concurrency and timeout settings in new-run.tsx now persist to localStorage.
• Components:
  • Updates MultiSelect in multi-select.tsx to support controlled mode via value prop.
  • Adds Dialog for viewing run settings in run.tsx.
• Misc:
  • Filters deprecated models in use-roo-code-cloud-models.ts.
  • Adds formatDateTime to formatters.ts for date formatting.
  • Minor UI and UX improvements across components.

^{This description was created by for f0492aa. You can customize this summary. It will automatically update as commits are pushed.}

[roomote]

Oroocle   See task on Roo Cloud

Re-review complete for the latest dashboard changes (runs list sorting plus provider and created columns). No new issues were identified and all previously flagged items remain resolved.

• MultiSelect Backspace handler still calls onValueChange directly while other paths go through setSelectedValues, which can cause duplicate updates in controlled mode.

Previous reviews

• 6dc4ba4: Review #1
• 7541104: Review #2
• d928b80: Review #3

_{Mention @roomote in a comment to request specific changes to this pull request or fix all unresolved issues.}

[roomote]

Fixaroo   Follow along on Roo Cloud\n\nWorking on the MultiSelect controlled-mode Backspace handler issue: ensuring Backspace updates selection state exclusively via setSelectedValues so onValueChange is not fired twice in controlled mode.