🪢 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 { createTRPCRouter, publicProcedure } from "@/src/server/api/trpc";
import { env } from "@/src/env.mjs";
import { logger } from "@langfuse/shared/src/server";
import { CloudStatus } from "@/src/features/cloud-status-notification/types";
import { z } from "zod";
// Cache the response for 1 minute
let statusCache: { status: CloudStatus | null; timestamp: number } | null =
null;
const CACHE_TTL = 60 * 1000; // 1 minute in milliseconds
// incident.io widget API response schema
const IncidentIoWidgetResponse = z.object({
ongoing_incidents: z.array(
z.object({
current_worst_impact: z.enum([
"partial_outage",
"degraded_performance",
"full_outage",
]),
}),
),
in_progress_maintenances: z.array(z.object({})),
scheduled_maintenances: z.array(z.object({})),
});
export const cloudStatusRouter = createTRPCRouter({
getStatus: publicProcedure
.output(
z.object({
status: CloudStatus,
}),
)
.query(async () => {
// Skip status check if not running on Langfuse Cloud
if (!env.NEXT_PUBLIC_LANGFUSE_CLOUD_REGION) {
return { status: null };
}
// Return cached result if it exists and is still valid
if (statusCache && Date.now() - statusCache.timestamp < CACHE_TTL) {
return {
status: statusCache.status,
};
}
try {
// incident.io widget API for status.langfuse.com
const response = await fetch(
"https://status.langfuse.com/api/v1/summary",
);
if (!response.ok) {
logger.error(
`Failed to fetch status from incident.io: ${response.statusText}`,
);
statusCache = {
status: null,
timestamp: Date.now(),
};
return { status: null };
}
const data = await response.json();
logger.info(`Data from incident.io: ${JSON.stringify(data)}`);
const parsed = IncidentIoWidgetResponse.parse(data);
// Determine status based on incidents and maintenances
let newStatus: CloudStatus = "operational";
// Check for ongoing incidents
if (parsed.ongoing_incidents.length > 0) {
const worstImpact = parsed.ongoing_incidents.reduce(
(worst, incident) => {
if (incident.current_worst_impact === "full_outage")
return "full_outage";
if (
incident.current_worst_impact === "partial_outage" &&
worst !== "full_outage"
)
return "partial_outage";
if (worst === "degraded_performance") return worst;
return incident.current_worst_impact;
},
"degraded_performance" as
| "degraded_performance"
| "partial_outage"
| "full_outage",
);
if (worstImpact === "full_outage") {
newStatus = "downtime";
} else {
newStatus = "degraded";
}
}
// Check for in-progress or scheduled maintenances
else if (
parsed.in_progress_maintenances.length > 0 ||
parsed.scheduled_maintenances.length > 0
) {
newStatus = "maintenance";
}
statusCache = {
status: newStatus,
timestamp: Date.now(),
};
return {
status: newStatus,
};
} catch (error) {
logger.error(`Error fetching status from incident.io: ${error}`);
// If there's an error, default to no incident to avoid showing false positives
statusCache = {
status: null,
timestamp: Date.now(),
};
return { status: null };
}
}),
});