Launch Week 02 wrapped — explore all five launches

Runtime Controls

Assess production observability metrics against a threshold.

Included on the Enterprise plan. Book a demo, opens in a new tab. Not included on the Team plan. Not included on the Starter plan. Not included on the Free plan.

Runtime controls assess production observability data against a threshold. They use the same metric model as alerts, but turn the result into an auditable governance requirement.

Each runtime control evaluates a trailing 24-hour window.

When should you use runtime controls?

  • When a requirement depends on measurable production behavior rather than project configuration.
  • When you need to continuously enforce reliability, quality, latency, cost, traffic, or human-feedback thresholds.
  • When the requirement should apply to a precisely filtered set of traces, spans, threads, metric results, or annotations.

Configure a runtime control

A runtime control consists of four parts:

  1. Data model — select Trace, Span, Thread, Metric data, or Annotation.
  2. Aggregation — select the value to compute. The available aggregations depend on the data model.
  3. Threshold — select Above or Below, then enter a numeric value.
  4. Filters — optionally narrow the evaluated data by environment, tags, metadata, or other available properties.

Data models and aggregations

Trace

Trace controls aggregate end-to-end LLM application requests. Use them to govern request volume, quality, latency, cost, and user activity.

AggregationWhat it measures
Trace countTotal number of matching traces
Error ratePercentage of matching traces that contain an error
Pass ratePercentage of matching traces that pass evaluation
Unique end usersNumber of distinct end users represented by matching traces
Unique threadsNumber of distinct threads represented by matching traces
Avg latencyMean latency across matching traces
P50 latencyMedian latency across matching traces
P90 latency90th-percentile latency across matching traces
P99 latency99th-percentile latency across matching traces
Total costSum of LLM cost across matching traces
Avg costMean LLM cost per matching trace
Unique metadata valuesNumber of distinct values for selected metadata

Span

Span controls aggregate individual operations within traces. Use them to isolate the behavior of a model call, tool call, retriever, or another instrumented operation.

AggregationWhat it measures
Span countTotal number of matching spans
Error ratePercentage of matching spans that contain an error
Error countTotal number of matching spans that contain an error
Avg latencyMean latency across matching spans
P50 latencyMedian latency across matching spans
P90 latency90th-percentile latency across matching spans
P99 latency99th-percentile latency across matching spans
Total costSum of LLM cost across matching spans
Input costSum of input-token cost across matching spans
Output costSum of output-token cost across matching spans
Avg costMean LLM cost per matching span
Unique metadata valuesNumber of distinct values for selected metadata
Input tokensTotal input-token usage across matching spans
Output tokensTotal output-token usage across matching spans
Total tokensTotal input and output tokens across matching spans

Thread

Thread controls aggregate multi-turn conversations.

AggregationWhat it measures
Thread countTotal number of matching threads
Unique end usersNumber of distinct end users represented by matching threads
Unique metadata valuesNumber of distinct values for selected metadata

Metric data

Metric data controls aggregate evaluation metric results. First select the metric data source—Trace, Span, Thread, or Test run—then choose the aggregation.

AggregationWhat it measures
Metric countTotal number of matching metric results
Avg scoreMean score across matching metric results
Median scoreMedian score across matching metric results
Pass ratePercentage of matching metric results that pass
Failure ratePercentage of matching metric results that fail

Annotation

Annotation controls aggregate human feedback attached to your data.

AggregationWhat it measures
Annotation countTotal number of matching annotations
Avg ratingMean rating across matching annotations

Threshold behavior

The threshold direction describes the condition that causes the control to fail:

DirectionThe control fails when…Example
AboveThe aggregated value is greater than the thresholdFail when error rate is above 5%
BelowThe aggregated value is less than the thresholdFail when successful trace count is below 1,000

If the aggregated value remains within the required bound, the control passes.

Common runtime controls

  • Reliability — require error rate to remain below an agreed limit.
  • Latency — require average or percentile latency to remain below an SLA.
  • Cost — require token cost to remain below a daily budget.
  • Traffic — require trace, span, thread, metric, or annotation volume to remain above a minimum.
  • Adoption — require the number of unique end users to remain above a target.
  • Quality — require evaluation pass rate, metric scores, or human ratings to remain above a target.
Scaling beyond prototype?For teams evaluating Confident AI in productionTalk to us

Last updated on

Built byConfident AI