AI Eval Score Chart
Prompt evaluation panel comparing candidate and production versions with a score distribution.
About this block
What it's for
- Prompt evaluation panel comparing candidate and production versions with a score distribution.
- Prompt engineering console watching improvements or regressions in factuality and safety tests.
- CI/CD dashboard for language models that validates acceptance thresholds before release.
Includes
- Score distribution
- Prompt version comparison
- Decile histogram
- Pass threshold control
- Interactive rubric filter
- Drilldown case inspection
- Accessible data table
- Keyboard navigation
Works with
- Radix UI primitives (the registry variant it ships in)
- React 19, Next.js App Router and Tailwind CSS 4, installed with the shadcn CLI
- Your theme tokens, in light and dark mode
Install
Included with Agentic UI Kit (one-time license). Install needs your license key: see pricing · connect your project.
Agentic UI CLI
agentic-ui add ai-eval-score-chartInstalls through shadcn and adapts the component to your design tokens.
shadcn CLI
npx shadcn@4.17.0 add @agentic-ui/ai-eval-score-chartUsed in
A standalone component: drop it into any screen.
- Type
- Block
- Variants
- Radix UI
- Themes
- Light and dark
- Widths
- Preview it at desktop, tablet and mobile widths above.
- Motion
- Animated, with a reduced-motion fallback
- npm packages
lucide-react- Registry dependencies
lib-primitives- Files
1 file
- components/agentic-ui/ai-eval-score-chart/index.tsx
Related components
AI Action Confirmation
Operations agent that asks to drop production tables and shows the rows, size and last use of each one.
Details and live previewAI Agent Memory Viewer
Assistant settings where the user reviews and corrects what the agent knows about them (time zone, style preferences).
Details and live previewAI Agent Status Badge
Header of an agent run that moves from working to waiting-approval and opens the approval on click.
Details and live previewAI Chat Empty State
Assistant chat with no messages that offers four starting points configured by the workspace.
Details and live preview