Revealing Data Patterns

A demonstration of how agentic visualization components automatically reveal patterns, trends, and anomalies without manual analysis. The components make the insights obvious.

Proof model

Patterns become useful when the component reveals the exception.

The experiment shows service health, trend, and anomaly states without requiring manual inspection first.

Stream

Metrics arrive as raw series.

Compare

Small multiples expose difference.

Anomaly

Degradation becomes visible.

Pattern

The user sees what matters.

Pattern Revealed: Performance Degradation

ComparativeSparklines component reveals which services are improving vs. degrading

Auth
Database
Cache
Storage

What it reveals:

  • Auth is improving (trending down = faster)
  • Database is degrading (trending up = slower)
  • Cache is stable and optimal
  • Storage is flat (no change)

Pattern Revealed: AI-Powered Service Health

MetricCard + TrendIndicator with automatic semantic understanding via Cloudflare Workers AI

auth
45
ms avg
vs. last week
↓ 7
database
120
ms avg
vs. last week
↑ 25
cache
12
ms avg
vs. last week
↓ 3
storage
89
ms avg
vs. last week
→ 0

What it reveals: Database response time increased +26%, indicating a problem. Auth improved -13%, cache improved -20%, storage flat.

AI Enhancement: The component uses Cloudflare Workers AI to understand that "response_time" semantically means "lower is better"—automatically inverting colors without manual configuration.

First load calls AI API. Subsequent loads use cached results. Falls back to heuristics if AI unavailable.

Pattern Revealed: Error Distribution Health

DistributionBar shows proportional breakdown of HTTP responses

99%
2xx Success 9,850 (99%)
4xx Client 120 (1%)
5xx Server 30 (0%)

What it reveals: System is healthy - 98.5% success rate, 1.2% client errors (expected), only 0.3% server errors (excellent). The visual makes this distribution immediately obvious.

The Pattern Recognition Insight

Without visualization: You would need to manually:

  • Calculate percentage changes for each service
  • Compare trends across time periods
  • Compute proportions for error distribution
  • Identify which metrics are improving vs. degrading

With agentic components: Patterns are immediately visible. The database degradation, auth improvement, and healthy error distribution are obvious at a glance. The components do the analytical work.

This is the power of encoding expert knowledge into autonomous software.

All visualizations powered by @create-something/tufte. View full research paper or all experiments.