Configuration reference

Askdepth.init(options) returns the Askdepth API object. Named imports such as init, track, and setConsent expose the same singleton functions.

Initialization options

import { init, type AskdepthInitOptions } from '@askdepth/core';

const options = {
  writeKey: 'YOUR_BROWSER_WRITE_KEY',
  endpoint: 'https://YOUR_COLLECTOR/v1/telemetry',
  consent: 'unknown',
  environment: 'staging',
  sampleRate: 0.25,
  allowRapidClickSelectors: ['[data-stepper]'],
} satisfies AskdepthInitOptions;

init(options);

Sampling and reconfiguration

Sampling is decided with Math.random() < sampleRate when a runtime is created. It is an in-memory decision, with no persisted assignment across page loads. The SDK does not validate or clamp the sampling value, so validate configuration in your application.

An identical init call leaves the runtime in place. A changed option tears down the previous collectors and queue, then creates a new runtime and sampling decision. Keep initialization stable; avoid calling it on every render.

For consent changes, use setConsent. Consent is also part of the initialization comparison key, so mixing prop changes and direct consent calls requires care.

Delivery behavior

The queue flushes every two seconds, when it reaches ten events, or immediately for high-priority anomalies. It retains up to 50 queued events and approximately 64 KiB of serialized event data. The envelope adds overhead, so this bound is not a promise that every request fits a browser's keepalive budget. Low-priority events are dropped first under pressure.

Requests use fetch with keepalive: true. Page hide and visibility changes attempt to send pending data. There is no durable offline queue. Retried batches retain their IDs; retries can produce duplicate network requests.

The kill signal prevents reinitialization for the lifetime of the loaded module. Reloading the application creates a new module runtime. Inspect collector responses when initialization succeeds but data never appears.