@autopilotstartup/log-sdk

Cross-runtime TypeScript client for an Autopilot Log Server deployment. The package has no runtime dependencies and provides target-specific entry points without duplicating the logging protocol.

Install

npm install @autopilotstartup/log-sdk

Browser or React Native

Client loggers use the restricted /client-logs endpoint and expose only warn and error. The application must be registered by the Log Server.

import { createClientLogger } from '@autopilotstartup/log-sdk/browser'

const logger = createClientLogger({
  applicationId: 'my-web-app',
  apiKey: process.env.PUBLIC_LOG_INGESTION_KEY!,
  environment: 'production',
})

logger.warn('Search results are stale', { operation: 'search' })
logger.error('Checkout failed', new Error('Gateway unavailable'))
await logger.flush()

Use @autopilotstartup/log-sdk/react-native for React Native. Pass a custom fetch implementation only in runtimes that do not provide the Fetch API.

Client ingestion keys identify an application but cannot be confidential when embedded in distributed software. Log Server must enforce client validation, sanitization, and rate limits.

Node

Backend loggers use the privileged /logs endpoint and additionally expose info and child.

import { createBackendLogger } from '@autopilotstartup/log-sdk/node'

const logger = createBackendLogger({
  applicationId: 'my-api',
  apiKey: process.env.LOG_APP_SECRET!,
  environment: process.env.NODE_ENV,
})

const requestLogger = logger.child({ requestId: 'req_123' })
requestLogger.info('Request started', { route: '/v1/orders' })
requestLogger.error('Request failed', new Error('Database unavailable'))
await logger.close()

The SDK keeps a bounded in-memory queue, batches requests, redacts common sensitive fields, times out requests, and requeues failed batches. It does not persist queued events across process or app restarts.

Durable Node transports

Applications that already own a disk spool can use sendLogBatch from the Node entry point. It sends prebuilt events without changing IDs, timestamps, or metadata, and exposes HTTP status and Retry-After through LogTransportError. The application remains responsible for persistence, ordering, and retry policy.

Configuration

Required options are applicationId and apiKey. The SDK owns the official Log Server endpoint and defaults service to applicationId. Optional endpoint and service overrides support local development, self-hosting, and multi-service applications. Other controls include environment, batchSize, maxQueueSize, flushIntervalMs, requestTimeoutMs, fetch, generateEventId, and now.

Call close() during graceful Node shutdown. Browser and mobile telemetry is best-effort; call flush() at an appropriate lifecycle boundary when possible.