🪢 Open source agent evals & observability: Trace, evaluate, and improve LLM applications with one open platform.
TypeScript analyticsautogenevaluationlangchainlarge-language-modelsllama-indexllmllm-evaluationllm-observabilityllmopsmonitoringobservabilityopen-sourceopenaiplaygroundprompt-engineeringprompt-managementself-hostedycombinator
/* eslint-disable no-nested-ternary */
import {
Card,
CardContent,
CardDescription,
CardHeader,
CardTitle,
} from "@/src/components/ui/card";
import { useScoreAnalytics } from "../ScoreAnalyticsProvider";
import { MetricCard } from "../charts/MetricCard";
import { SamplingDetailsHoverCard } from "../SamplingDetailsHoverCard";
import {
calculateCohensKappa,
calculateWeightedF1Score,
calculateOverallAgreement,
interpretPearsonCorrelation,
interpretSpearmanCorrelation,
interpretCohensKappa,
interpretF1Score,
interpretOverallAgreement,
interpretMAE,
interpretRMSE,
} from "@/src/features/score-analytics/lib/statistics-utils";
import { Spinner } from "@/src/components/design-system/Spinner/Spinner";
/**
* StatisticsCard - Smart card component for displaying score statistics
*
* Consumes ScoreAnalyticsProvider context and displays:
* - Score 1 stats (always shown)
* - Score 2 stats (shown in two-score mode)
* - Comparison metrics (shown in two-score mode)
*
* Handles:
* - Loading states
* - Empty states
* - Single vs two-score modes
* - Numeric vs categorical data types
*/
export function StatisticsCard() {
const { data, isLoading, params } = useScoreAnalytics();
// Loading state
if (isLoading) {
return (
<Card>
<CardHeader>
<CardTitle>Statistics</CardTitle>
<CardDescription>Loading statistics...</CardDescription>
</CardHeader>
<CardContent className="flex items-center justify-center py-12">
<Spinner size="xl" variant="muted" />
</CardContent>
</Card>
);
}
// No data state
if (!data) {
return (
<Card>
<CardHeader>
<CardTitle>Statistics</CardTitle>
<CardDescription>No data available</CardDescription>
</CardHeader>
<CardContent className="text-muted-foreground py-12 text-center text-sm">
Select a score to view statistics
</CardContent>
</Card>
);
}
// Extract data from context
const { statistics, metadata } = data;
const { dataType } = metadata;
const { score1, score2 } = params;
// Check if Cartesian product occurred (matched count exceeds both individual counts)
const hasCartesianProduct =
statistics.comparison &&
statistics.comparison.matchedCount > statistics.score1.total &&
statistics.score2 &&
statistics.comparison.matchedCount > statistics.score2.total;
// Determine what to show
const showScore1Data = statistics.score1.total > 0;
const showScore2Data = statistics.score2 !== null;
const showComparisonMetrics = statistics.comparison !== null;
// Always show Score 2 and Comparison sections once score1 is selected
// to set user expectations about what information will be available
const showScore2Section = true; // Always show when on this page
const showComparisonSection = true; // Always show when on this page
// Calculate categorical metrics if available
const cohensKappa =
showComparisonMetrics && statistics.comparison?.confusionMatrix
? calculateCohensKappa(statistics.comparison.confusionMatrix)
: null;
const f1Score =
showComparisonMetrics && statistics.comparison?.confusionMatrix
? calculateWeightedF1Score(statistics.comparison.confusionMatrix)
: null;
const overallAgreement =
showComparisonMetrics && statistics.comparison?.confusionMatrix
? calculateOverallAgreement(statistics.comparison.confusionMatrix)
: null;
return (
<Card>
<CardHeader>
<CardTitle className="flex items-center gap-2">
Statistics
{data.samplingMetadata.isSampled && (
<SamplingDetailsHoverCard
samplingMetadata={data.samplingMetadata}
mode={data.metadata.mode}
showLabel
/>
)}
</CardTitle>
<CardDescription>
{score2
? `${score1.name} vs ${score2.name}`
: `${score1.name} - Select a second score for comparison`}
</CardDescription>
</CardHeader>
<CardContent className="space-y-4">
{/* Section 1: Score 1 Data */}
<div>
<h4 className="mb-2 text-xs font-bold">
{score1.name} ({score1.source})
</h4>
{dataType === "NUMERIC" ? (
<div className="grid grid-cols-1 gap-4 sm:grid-cols-3">
<MetricCard
label="Total"
value={
showScore1Data
? statistics.score1.total.toLocaleString()
: "--"
}
helpText={`Total number of ${score1.name} scores`}
isPlaceholder={!showScore1Data}
isContext
/>
<MetricCard
label="Mean"
value={
showScore1Data && statistics.score1.mean !== null
? statistics.score1.mean.toFixed(2)
: !showScore1Data
? "--"
: "N/A"
}
helpText={`Average value for ${score1.name}`}
isPlaceholder={!showScore1Data}
isContext
/>
<MetricCard
label="Std Dev"
value={
showScore1Data && statistics.score1.std !== null
? statistics.score1.std.toFixed(2)
: !showScore1Data
? "--"
: "N/A"
}
helpText={`Standard deviation for ${score1.name}`}
isPlaceholder={!showScore1Data}
isContext
/>
</div>
) : (
<div className="grid grid-cols-1 gap-4 sm:grid-cols-3">
<MetricCard
label="Total"
value={
showScore1Data
? statistics.score1.total.toLocaleString()
: "--"
}
helpText={`Total number of ${score1.name} scores`}
isPlaceholder={!showScore1Data}
isContext
/>
<MetricCard
label="Mode"
value={
showScore1Data && statistics.score1.mode
? `${statistics.score1.mode.category} (${statistics.score1.mode.count.toLocaleString()})`
: !showScore1Data
? "--"
: "N/A"
}
helpText="Most frequent category and its count"
isPlaceholder={!showScore1Data}
isContext
/>
<MetricCard
label="Mode %"
value={
showScore1Data && statistics.score1.modePercentage !== null
? `${statistics.score1.modePercentage.toFixed(1)}%`
: !showScore1Data
? "--"
: "N/A"
}
helpText="Percentage of observations with the most frequent category"
isPlaceholder={!showScore1Data}
isContext
/>
</div>
)}
</div>
{/* Section 2: Score 2 Data - Always show to set expectations */}
{showScore2Section && (
<div>
<h4 className="mb-2 text-xs font-bold">
{score2?.name ?? "Score 2"}
{score2?.source ? ` (${score2.source})` : ""}
</h4>
{dataType === "NUMERIC" ? (
<div className="grid grid-cols-1 gap-4 sm:grid-cols-3">
<MetricCard
label="Total"
value={
showScore2Data && statistics.score2
? statistics.score2.total.toLocaleString()
: "--"
}
helpText={`Total number of ${score2?.name ?? "Score 2"} scores`}
isPlaceholder={!showScore2Data}
isContext
/>
<MetricCard
label="Mean"
value={
showScore2Data &&
statistics.score2 &&
statistics.score2.mean !== null
? statistics.score2.mean.toFixed(2)
: !showScore2Data
? "--"
: "N/A"
}
helpText={`Average value for ${score2?.name ?? "Score 2"}`}
isPlaceholder={!showScore2Data}
isContext
/>
<MetricCard
label="Std Dev"
value={
showScore2Data &&
statistics.score2 &&
statistics.score2.std !== null
? statistics.score2.std.toFixed(2)
: !showScore2Data
? "--"
: "N/A"
}
helpText={`Standard deviation for ${score2?.name ?? "Score 2"}`}
isPlaceholder={!showScore2Data}
isContext
/>
</div>
) : (
<div className="grid grid-cols-1 gap-4 sm:grid-cols-3">
<MetricCard
label="Total"
value={
showScore2Data && statistics.score2
? statistics.score2.total.toLocaleString()
: "--"
}
helpText={`Total number of ${score2?.name ?? "Score 2"} scores`}
isPlaceholder={!showScore2Data}
isContext
/>
<MetricCard
label="Mode"
value={
showScore2Data && statistics.score2?.mode
? `${statistics.score2.mode.category} (${statistics.score2.mode.count.toLocaleString()})`
: !showScore2Data
? "--"
: "N/A"
}
helpText="Most frequent category and its count"
isPlaceholder={!showScore2Data}
isContext
/>
<MetricCard
label="Mode %"
value={
showScore2Data &&
statistics.score2 &&
statistics.score2.modePercentage !== null
? `${statistics.score2.modePercentage.toFixed(1)}%`
: !showScore2Data
? "--"
: "N/A"
}
helpText="Percentage of observations with the most frequent category"
isPlaceholder={!showScore2Data}
isContext
/>
</div>
)}
</div>
)}
{/* Section 3: Comparison Metrics - Always show to set expectations */}
{showComparisonSection && (
<div>
<h4 className="mb-2 text-xs font-bold">Comparison</h4>
{dataType === "NUMERIC" ? (
<div className="space-y-4">
{/* First row: Matched, Pearson, Spearman */}
<div className="grid grid-cols-1 gap-4 sm:grid-cols-3">
<MetricCard
label="Matched"
value={
showComparisonMetrics && statistics.comparison
? statistics.comparison.matchedCount.toLocaleString()
: "--"
}
helpText="Number of observations with both scores"
warning={
hasCartesianProduct
? {
show: true,
content: (
<div className="space-y-2 text-xs">
<p className="font-bold">
Matched count exceeds individual score counts
due to Cartesian product
</p>
<p>
This occurs when multiple scores of the same
name/source exist on a single attachment point
(trace/observation/session/run). Each
combination creates a match.
</p>
<p className="text-muted-foreground">
<strong>Example:</strong> If one trace has 2
"gpt4" scores and 3
"gemini" scores, this creates 6
matched pairs (2 × 3 = 6).
</p>
</div>
),
}
: undefined
}
isContext
isPlaceholder={!showComparisonMetrics}
/>
<MetricCard
label="Pearson r"
value={
showComparisonMetrics &&
statistics.comparison &&
statistics.comparison.pearsonCorrelation !== null
? statistics.comparison.pearsonCorrelation.toFixed(3)
: showComparisonMetrics
? "N/A"
: "--"
}
interpretation={
showComparisonMetrics &&
statistics.comparison &&
statistics.comparison.pearsonCorrelation !== null
? interpretPearsonCorrelation(
statistics.comparison.pearsonCorrelation,
)
: undefined
}
helpText="Linear correlation (-1 to 1)"
isPlaceholder={!showComparisonMetrics}
/>
<MetricCard
label="Spearman ρ"
value={
showComparisonMetrics &&
statistics.comparison &&
statistics.comparison.spearmanCorrelation !== null
? statistics.comparison.spearmanCorrelation.toFixed(3)
: showComparisonMetrics
? "N/A"
: "--"
}
interpretation={
showComparisonMetrics &&
statistics.comparison &&
statistics.comparison.spearmanCorrelation !== null
? interpretSpearmanCorrelation(
statistics.comparison.spearmanCorrelation,
)
: undefined
}
helpText="Rank correlation (-1 to 1)"
isPlaceholder={!showComparisonMetrics}
/>
</div>
{/* Second row: Empty, MAE, RMSE */}
<div className="grid grid-cols-1 gap-4 sm:grid-cols-3">
<div />
<MetricCard
label="MAE"
value={
showComparisonMetrics &&
statistics.comparison &&
statistics.comparison.mae !== null
? statistics.comparison.mae.toFixed(3)
: showComparisonMetrics
? "N/A"
: "--"
}
interpretation={
showComparisonMetrics &&
statistics.comparison &&
statistics.comparison.mae !== null
? interpretMAE(statistics.comparison.mae)
: undefined
}
helpText="Mean Absolute Error"
isPlaceholder={!showComparisonMetrics}
/>
<MetricCard
label="RMSE"
value={
showComparisonMetrics &&
statistics.comparison &&
statistics.comparison.rmse !== null
? statistics.comparison.rmse.toFixed(3)
: showComparisonMetrics
? "N/A"
: "--"
}
interpretation={
showComparisonMetrics &&
statistics.comparison &&
statistics.comparison.rmse !== null
? interpretRMSE(statistics.comparison.rmse)
: undefined
}
helpText="Root Mean Square Error"
isPlaceholder={!showComparisonMetrics}
/>
</div>
</div>
) : (
<div className="space-y-4">
{/* First row: Matched, Agreement */}
<div className="grid grid-cols-1 gap-4 sm:grid-cols-3">
<MetricCard
label="Matched"
value={
showComparisonMetrics && statistics.comparison
? statistics.comparison.matchedCount.toLocaleString()
: "--"
}
helpText="Number of observations with both scores"
warning={
hasCartesianProduct
? {
show: true,
content: (
<div className="space-y-2 text-xs">
<p className="font-bold">
Matched count exceeds individual score counts
due to Cartesian product
</p>
<p>
This occurs when multiple scores of the same
name/source exist on a single attachment point
(trace/observation/session/run). Each
combination creates a match.
</p>
<p className="text-muted-foreground">
<strong>Example:</strong> If one trace has 2
"gpt4" scores and 3
"gemini" scores, this creates 6
matched pairs (2 × 3 = 6).
</p>
</div>
),
}
: undefined
}
isContext
isPlaceholder={!showComparisonMetrics}
/>
<MetricCard
label="Agreement"
value={
showComparisonMetrics && overallAgreement !== null
? `${(overallAgreement * 100).toFixed(1)}%`
: showComparisonMetrics
? "N/A"
: "--"
}
interpretation={
showComparisonMetrics && overallAgreement !== null
? interpretOverallAgreement(overallAgreement)
: undefined
}
helpText="Overall agreement percentage"
isPlaceholder={!showComparisonMetrics}
/>
</div>
{/* Second row: Empty, Cohen's κ, F1 Score */}
<div className="grid grid-cols-1 gap-4 sm:grid-cols-3">
<div />
<MetricCard
label="Cohen's κ"
value={
showComparisonMetrics && cohensKappa !== null
? cohensKappa.toFixed(3)
: showComparisonMetrics
? "N/A"
: "--"
}
interpretation={
showComparisonMetrics && cohensKappa !== null
? interpretCohensKappa(cohensKappa)
: undefined
}
helpText="Inter-rater reliability (-1 to 1)"
isPlaceholder={!showComparisonMetrics}
/>
<MetricCard
label="F1 Score"
value={
showComparisonMetrics && f1Score !== null
? f1Score.toFixed(3)
: showComparisonMetrics
? "N/A"
: "--"
}
interpretation={
showComparisonMetrics && f1Score !== null
? interpretF1Score(f1Score)
: undefined
}
helpText="Weighted F1 score (0 to 1)"
isPlaceholder={!showComparisonMetrics}
/>
</div>
</div>
)}
</div>
)}
</CardContent>
</Card>
);
}