Self-hosted, plugin-based server management.
TypeScript dockerhomelabpluginsrdpremote-desktopself-hostedserver-managementsftpsshterminaltermixvnc
import { isElectron } from "@/lib/electron";
export type EmbeddedServerFailureReason =
"port-in-use" | "old-database" | "crashed";
export interface EmbeddedServerFailure {
reason: EmbeddedServerFailureReason;
port: number | null;
}
type EmbeddedServerStatusWindow = Window &
typeof globalThis & {
electronAPI?: {
getEmbeddedServerStatus?: () => Promise<{
running?: boolean;
failure?: { reason?: string; port?: number | null } | null;
}>;
};
};
/**
* Asks the Electron main process whether the embedded backend has failed for
* good.
*
* Returns null in every case where waiting is still the right thing to do:
* outside Electron, while the backend is only slow to boot, and on a desktop
* build old enough not to expose the channel at all. A non-null result means
* the backend process is gone and no amount of retrying will bring it back.
*/
export async function getEmbeddedServerFailure(): Promise<EmbeddedServerFailure | null> {
if (!isElectron()) return null;
const getStatus = (window as EmbeddedServerStatusWindow).electronAPI
?.getEmbeddedServerStatus;
if (typeof getStatus !== "function") return null;
try {
const status = await getStatus();
const failure = status?.failure;
if (!failure) return null;
// An unrecognized reason from a newer/older main process still means the
// backend is down, which is the part the caller acts on -- so it is
// reported as the generic crash rather than dropped.
if (failure.reason === "port-in-use") {
return { reason: "port-in-use", port: failure.port ?? null };
}
if (failure.reason === "old-database") {
return { reason: "old-database", port: null };
}
return { reason: "crashed", port: null };
} catch {
// No status channel, or the main process is tearing down. Neither is
// evidence that the backend failed, so keep waiting.
return null;
}
}