Skip to main content

​
RetryOptions

Retry-policy options accepted by withRetry. The same policy is used by every namespace client method; the exported helper is useful when you need that behavior on a call that isn’t wrapped yet (e.g., a direct call to adaline.deploymentsApi.* or adaline.logsApi.*).

​
Definition

interface RetryOptions {
  retries?: number;
  minTimeout?: number;
  maxTimeout?: number;
  randomize?: boolean;
  maxRetryTime?: number;
  factor?: number;
}

​
Fields

FieldTypeDefaultDescription
retriesnumber10Maximum number of retry attempts on 5xx responses.
minTimeoutnumber1000Minimum delay between attempts (ms).
maxTimeoutnumber10000Maximum delay between attempts (ms).
factornumber2Exponential backoff factor. Each retry waits factor × the previous delay.
randomizebooleantrueIf true, adds jitter to each delay.
maxRetryTimenumber20000Hard ceiling on total time spent retrying (ms). The retry loop gives up once this is exceeded.

​
Defaults

The exported DEFAULT_RETRY_OPTIONS constant is what every namespace client uses: 
const DEFAULT_RETRY_OPTIONS: RetryOptions = {
  retries: 10,
  minTimeout: 1000,
  maxTimeout: 10000,
  randomize: true,
  maxRetryTime: 20000,
  factor: 2,
};

​
Behavior

  • 5xx responses — retried with exponential backoff and jitter.
  • 4xx responses — aborted immediately via AbortError; no retries.
  • Network errors — retried like 5xx.

​
Usage

import { Adaline, withRetry } from '@adaline/client';

const adaline = new Adaline();

// Short-circuit retries for a latency-sensitive path
const deployment = await withRetry(
  () => adaline.deploymentsApi.getDeployment('prompt_abc123', 'deploy_xyz789'),
  { retries: 2, maxRetryTime: 2000 },
);

​
See Also