Chart Types Reference
Every metric in a chart has an associated chart type that controls how it is visualized. The chart type of the first metric in the list determines the overall rendering mode of the chart.
Composable types
These types can be freely mixed within a single chart. For example, one metric can render as a bar while another renders as a line — both plotted on the same axes.
Bar
A standard vertical bar chart. Each category value on the x-axis produces one bar whose height represents the aggregated metric value.
Best for: Comparing discrete values across categories (e.g., inference count per device, average latency per workflow).
Bar Stacked
Similar to Bar, but multiple metrics or grouped series are stacked vertically within each bar rather than placed side by side. The total height of the bar represents the combined value of all stacked segments.
Best for: Showing the part-to-whole relationship of multiple metrics or groups (e.g., total detections split by skill across devices).
Line
A continuous line connecting data points across the x-axis. Missing data points result in a gap in the line unless interpolation is enabled.
Best for: Trends over time (e.g., CPU usage over the past 24 hours, detection rate over a week).
Area
Similar to Line, but the region between the line and the x-axis is filled with a translucent gradient. Emphasizes volume and magnitude rather than just the trend direction.
Best for: Showing cumulative or volume-oriented metrics over time (e.g., total frames processed, network bytes transferred).
Area Stacked
Multiple area series stacked on top of each other. Each series' area starts where the previous one ends, so the topmost edge represents the combined total.
Best for: Comparing cumulative contributions of multiple groups or metrics over time (e.g., total processing time broken down per skill).
Non-composable types
These types take over the entire chart — they cannot be mixed with other chart types.
Radar
A radial (spider/web) chart. Each metric or category becomes a spoke radiating from the center. Values are plotted along each spoke and connected to form a polygon.
Best for: Comparing multiple metrics on a common normalized scale, or comparing the shape of a profile across different categories.
Large Number
Displays a single large aggregated value, styled as a KPI/stat card. Shows the metric key, calculation method, and granularity as a subtitle.
Best for: Dashboard-style summary tiles where you want a prominent single-number display (e.g., total detections today, average latency this week).
Table
Displays raw data in a sortable tabular format with column headers. Supports infinite scroll pagination and renders rich content inline:
| Value type | Rendering |
|---|---|
| URL (image) | Inline thumbnail (80×80px) |
| URL (video) | Inline video player |
| URL (audio) | Inline audio player |
| URL (other) | Clickable external link |
| Boolean | Color-coded badge (green = true, red = false) |
| Timestamp | Formatted per granularity; full datetime on hover |
| NULL | Displayed as italicized "NULL" |
| Number | Rounded to 3 decimal places |
Best for: Inspecting raw event-level data, viewing output fields that aren't numeric, or auditing exact values (e.g., snapshot images, event payloads).
Choosing the right type
| Goal | Recommended type |
|---|---|
| Track a metric over time | Line or Area |
| Compare values across categories | Bar |
| Show contribution of multiple groups | Bar Stacked or Area Stacked |
| Profile comparison across many attributes | Radar |
| KPI summary / single headline number | Large Number |
| Inspect raw data rows | Table |