Never stop coding. Free MIT AI gateway: one endpoint, 375 providers (150+ free), 1,400+ models Kimi, Claude, GPT, Gemini, GLM, DeepSeek, MiniMax. Works with Claude Code, Codex, Cursor, OpenCode, Cline & Copilot. Quota-aware auto-fallback, RTK+Caveman compression saves 15-95% tokens, MCP/A2A, Desktop/PWA. Built by 900+ contributors
TypeScript a2aai-agentsai-gatewayanthropicclaudeclaude-codeclinecodexcopilotcursordeepseekfree-aigeminikimillm-gatewaymcpopenaiopenai-proxyqwentoken-saver
/**
* Lightweight usage-event bus.
*
* Decouples usage recording (usageHistory.ts) from the provider-limits
* subsystem (providerLimits.ts). usageHistory must NOT import providerLimits:
* providerLimits pulls in the executors barrel (and the whole translator graph),
* so a direct or dynamic import from usageHistory expands the type-check surface
* across modules that have nothing to do with usage recording. usageHistory
* emits here; providerLimits subscribes at module load.
*
* @module lib/usage/usageEvents
*/
export type UsageRecordedListener = (provider: string, connectionId: string) => void;
const listeners = new Set<UsageRecordedListener>();
/** Register a listener for usage-recorded events. Returns an unsubscribe fn. */
export function onUsageRecorded(listener: UsageRecordedListener): () => void {
listeners.add(listener);
return () => {
listeners.delete(listener);
};
}
/** Emit a usage-recorded event. No-ops when provider/connectionId is missing. */
export function emitUsageRecorded(
provider: string | null | undefined,
connectionId: string | null | undefined
): void {
if (!provider || !connectionId) return;
for (const listener of listeners) {
try {
listener(provider, connectionId);
} catch (error) {
const message = error instanceof Error ? error.message : String(error);
console.warn(`[usageEvents] usage-recorded listener failed: ${message}`);
}
}
}
// ── Stats-event debounce ─────────────────────────────────────────────────────
//
// Rapid back-to-back inserts (e.g. combo routing that fans out to multiple
// models simultaneously) can fire dozens of "update" events per second.
// scheduleStatsEvent collapses bursts: the first call within a window sets a
// timer; subsequent calls within the same window are no-ops. The timer fires
// once at the end of the window.
type StatsEventKey = "update" | "pending";
const statsEmitTimers: Record<StatsEventKey, ReturnType<typeof setTimeout> | null> = {
update: null,
pending: null,
};
const statsListeners: Record<StatsEventKey, Set<() => void>> = {
update: new Set(),
pending: new Set(),
};
/** Register a debounced stats listener. Returns an unsubscribe fn. */
export function onStatsEvent(event: StatsEventKey, listener: () => void): () => void {
statsListeners[event].add(listener);
return () => {
statsListeners[event].delete(listener);
};
}
/**
* Schedule a stats event emission after `delayMs`, collapsing rapid bursts
* into a single notification. Safe to call from hot paths — subsequent calls
* within the same window are no-ops.
*/
export function scheduleStatsEvent(event: StatsEventKey, delayMs = 200): void {
if (statsEmitTimers[event] != null) return;
statsEmitTimers[event] = setTimeout(() => {
statsEmitTimers[event] = null;
for (const listener of statsListeners[event]) {
try {
listener();
} catch (error) {
const message = error instanceof Error ? error.message : String(error);
console.warn(`[usageEvents] stats listener (${event}) failed: ${message}`);
}
}
}, delayMs);
// Allow Node.js to exit naturally even if the timer is still pending
// (avoids keeping the event loop alive for a stray stats notification).
(statsEmitTimers[event] as any)?.unref?.();
}