Metrics Reference

Charts visualize data from metrics — named values emitted by skills running on your devices. There are two metric types available in the chart editor:

Output metrics

Outputs are key-value pairs written by skills as they run. Every time a skill processes data (e.g., a frame from a camera, a sensor reading), it can emit one or more named output values. These are stored and made available for charting.

Output keys are specific to your organization and the skills deployed on your devices. Common examples include:

• inference_count — number of detections or inferences in a time window
• processing_time_ms — time taken to run a model
• confidence — model confidence score for a detection
• person_count — number of people detected
• temperature — a sensor reading

The available metric list in the chart editor reflects all output keys that have been reported by your devices. As new skills are deployed or new output keys are added, they appear automatically.

You can also pre-register an output key before any device has reported it — useful when setting up a chart ahead of a new skill deployment. Click the + icon in the metric picker, select Output, and enter the key name.

tip

If you don't see an expected metric in the list, it may not have been reported by any device yet. Check that the relevant skill is deployed and running on at least one device.

---

Calculated metrics

Calculated metrics are custom metrics defined by a SQL expression over your output data. They allow you to:

• Combine multiple output fields (e.g., "detections" / "frames" * 100 for a detection rate percentage)
• Normalize or transform values
• Define reusable derived metrics across multiple charts

Calculated metrics are managed at the organization level — any metric you create is available across all charts and workspaces.

Create a calculated metric

1. In the chart editor, open the metric picker by clicking an existing metric or the Add button.
2. Click the + icon at the top of the metrics list.
3. Select Calculated Metric as the type.
4. Enter a key — this is the display name used to identify the metric across the portal.
5. Enter a SQL expression — reference output field names using double-quoted identifiers (e.g., "detections", "frames").
6. Click Create.

Edit or delete a calculated metric

1. Open the metric picker in the chart editor.
2. Find the calculated metric in the list and click Edit next to it.
3. Modify the key or SQL expression and click Save, or use the Delete button to remove it.

caution

Deleting a calculated metric removes it from all charts that use it. Any charts that relied on it will show no data for that metric.

---

Calculations (aggregations)

When adding a metric to a chart, you select a calculation that determines how individual data points are aggregated per time bucket or category:

Calculation	Description	Use when...
Last	Most recent value in the period	You want the latest reading, not a summary
Count	Number of events/records	Tracking how many times something occurred
Sum	Total of all values	Accumulating totals (e.g., total frames processed)
Average	Mean of all values	Smoothing noisy readings (e.g., average latency)
Minimum	Lowest value	Tracking floor values (e.g., minimum confidence)
Maximum	Highest value	Tracking peak values (e.g., max CPU spike)