export default class RateLimiter {
    lastAPICall: number | null;
    delayBetweenCallsMs: number;
    verboseLogging: boolean;
    /** Optional tokens-per-minute cap enforced alongside the RPM slot. */
    tokensPerMinute?: number;
    private nextSlot;
    /**
     * Sliding 60-second window of recent token consumption. Each entry
     * is [unix-millis-at-consume-time, token-count]. Entries older than
     * 60s are pruned on every acquire() call.
     */
    private tokenWindow;
    constructor(delayBetweenCallsMs: number, verboseLogging: boolean, tokensPerMinute?: number);
    /**
     * Reserve the next RPM slot and, when configured, also wait for
     * enough TPM headroom to fit the estimated token cost of the call
     * about to be made. Concurrent callers reserve distinct slots
     * synchronously; the TPM check runs afterward.
     * @param estimatedTokens - tokens this call is expected to consume; 0 skips the TPM check
     */
    acquire(estimatedTokens?: number): Promise<void>;
    /**
     * TPM enforcement. Prunes stale entries, and if adding this call
     * would exceed the cap, waits until the oldest entry is old enough
     * to fall out of the 60-second window.
     */
    private awaitTokenBudget;
    penalize(penaltyMs: number): void;
    apiCalled(): void;
    wait(): Promise<void>;
}