🪢 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
import { type RowData } from "@tanstack/react-table";
import {
createTableColumn,
type TableColumnOptions,
} from "@/src/components/design-system/table/columns/utils/createTableColumn";
import { Skeleton } from "@/src/components/ui/skeleton";
import { NotRecordedMetric } from "./NotRecordedMetric";
type ExperimentMetric = "cost" | "latency";
/**
* A cost of exactly 0 means the run's calls reported no usage or pricing, which
* reads as free rather than as missing. A latency of 0 is a real measurement.
*/
const isRecorded = (
metric: ExperimentMetric,
value: number | null | undefined,
): value is number =>
value !== null && value !== undefined && (metric !== "cost" || value !== 0);
/**
* A cost or latency column for an experiments table. The metrics load in a
* second query that resolves after the rows, so the cell has three states: the
* skeleton while that query is in flight, `NotRecordedMetric` once it lands
* without a value, and the formatted number otherwise.
*/
export function createExperimentMetricColumn<TData extends RowData>({
formatter,
metric,
metricsLoading,
...options
}: TableColumnOptions<TData, number> & {
formatter: (value: number) => string;
metric: ExperimentMetric;
metricsLoading: boolean;
}) {
const loadingCell = <Skeleton className="h-4 w-1/2" />;
return createTableColumn<TData, number>({
...options,
loadingCell,
renderCell: (value) => {
// Rows render before their metrics, so without this the cell would first
// claim the metric was not recorded and then correct itself.
if (metricsLoading) return loadingCell;
if (!isRecorded(metric, value))
return <NotRecordedMetric metric={metric} />;
return <span>{formatter(value)}</span>;
},
});
}