🪢 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 {
MonitorSeverity,
MonitorThresholdOperator,
Monitor,
} from "../types";
import type { MonitorCompletion } from "./applyStateMachine";
/** renderAlertMessage builds the human-readable title/body for a MonitorAlert. The body distinguishes no-data alerts from threshold-crossing alerts. */
export function renderAlertMessage(args: {
monitor: Monitor;
completion: MonitorCompletion;
}): { title: string; body: string } {
const { monitor, completion } = args;
const prevSeverity = monitor.severity;
const severity = completion.severity;
const title = `[${severity}] ${monitor.name}`;
// standard markdown; the Slack builder later runs it through slackify-markdown
const metricRef = `\`${monitor.metric.aggregation}(${monitor.view}.${monitor.metric.measure})\``;
let body: string;
if (severity === "NO_DATA") {
body = `${metricRef} has no data over the last **${monitor.window}**`;
} else if (prevSeverity === "NO_DATA" && severity === "OK") {
body = `${metricRef} recovered and is reporting data again`;
} else if (severity === "OK") {
body = `${metricRef} recovered`;
} else {
const threshold = selectThreshold(
severity,
monitor.alertThreshold,
monitor.warningThreshold ?? null,
);
body = `${metricRef} is **${operatorWord(monitor.thresholdOperator)}** \`${threshold}\` over the last **${monitor.window}**`;
}
return { title, body };
}
/** operatorWord returns the human-readable form of a threshold operator. */
function operatorWord(op: MonitorThresholdOperator): string {
switch (op) {
case "GT":
return "above";
case "GTE":
return "at or above";
case "LT":
return "below";
case "LTE":
return "at or below";
case "EQ":
return "equal to";
case "NEQ":
return "not equal to";
}
}
/** selectThreshold picks the threshold relevant to the current severity (warning band for WARNING, alert for everything else). */
function selectThreshold(
severity: MonitorSeverity,
alertThreshold: number,
warningThreshold: number | null,
): number {
if (severity === "WARNING" && warningThreshold !== null) {
return warningThreshold;
}
return alertThreshold;
}