@expo/rudder-sdk-node

A lightweight Node SDK for RudderStack with minimal dependencies. This is designed for client-side Node applications like CLIs. This library is smaller than RudderStack's Node library and doesn't include support for the Redis persistence queue.

It is fully written in TypeScript and exports first-class type declarations to users of this package.

Installation

$ npm install @expo/rudder-sdk-node

Usage

import Analytics from '@expo/rudder-sdk-node';

// Specify the batch endpoint of the Rudder server you are running
const client = new Analytics('write key', '<data-plane-uri>/v1/batch');

client.track({
  event: 'event name',
  userId: 'user id',
});

const flushResponse = await client.flush();

Documentation

Look at the TypeScript type declarations and the source code of this library.

RudderStack's documentation for a different but related library is available here.

v1.1.1

Minor changes

• adds some more verbose logging for future debugging

Bug Fixes

• fixed an issue where returning a nullResponse from a flush resulted in future flushes returning that nullResponse.

v1.1.0

Breaking changes

• The main Analytics class is now a default export instead of being the value of module.exports. If you're using require with this library, you'll need to write require('@expo/rudder-sdk-node').default. If you're using import with this library, no changes are needed
• Removed support for persisting events in Redis
• Changed the logger from winston to bunyan (specifically, @expo/bunyan)
• Changed the default log level from INFO to FATAL
• The timeout option must be a number (milliseconds). Strings are no longer supported. As before, 0 means no timeout but now negative numbers also disable timeouts
• Changed how flush() works
  • flush() now sends all messages present in the message queue at the time of the request
    • previously flush() would only send flushAt number of events
  • flush() now guards against batch sizes too large for the server via the configurable maxFlushSizeInBytes property. Trimmed messages remain in the queue for subsequent HTTP requests
  • flush() now enforces a limit of one concurrent HTTP request at a time
    • simultaneous flush() requests will be handled when the current HTTP request finishes
  • Flushing will now occur
    • the first time an event is enqueued after instantiation (occurs immediately)
    • when the message queue size is divisible by flushAt (occurs immediately)
    • when a new event is enqueued after a flush (a timer is set for flush to occur after flushInterval ms)
    • when a user requests a flush()
  • flush() is now an async function
  • flush() now has a return type of Promise<FlushResponse[]>
  • flush callbacks now accept arguments of type FlushResponse[]
  • flush() no longer throws upon failure but returns an error object instead
  • Changed when fetch requests are retried - All network errors are retried
• No longer causes an exiting process to indefinitely stall due to a looping timer. If a process is naturally exiting (not due to process.exit()), you must call analytics.flush() to send the last events for them to be sent
• The userId and anonymousId message fields must be primitive strings. Other types like numbers are not supported

Minor changes

• Convert from JavaScript to TypeScript. First-class type
• Replaced axios with node-fetch
• The library name in the log context object is now @expo/rudder-sdk-node
• The user agent string is now expo-rudder-sdk-node/<version>
• enqueue now limits the max events in the queue via configurable maxQueueLength property. If enqueue() is called and the message queue is already at or above the maxQueueLength that message is dropped

v1.0.8

Based on 1.0.3+ of @rudderstack/rudder-sdk-node.

• Consistently use winston instead of console.log for logging