What is Sawdust?

Sawdust is one logging API for every JavaScript runtime you ship.

@cues/sawdust is a runtime-agnostic logging toolkit built on LogLayer. Import a single logger, call .info() / .warn() / .error(), and the right implementation runs whether your code executes in Node.js, the browser, or a background worker.

Core is provider-agnostic — console, pretty, and Consola transports out of the box, and no vendor lock-in. Need Datadog logs, APM trace injection, or RUM? Add @cues/sawdust-datadog and wire it through the extraTransports / plugins seams.

import { logger } from '@cues/sawdust/logger'

logger.info('checkout completed', { orderId, amountCents })
logger.error('payment failed', err, { orderId })

No per-environment logger. No factory to wire up before you can log. No conditional imports.

The problem it solves

Every non-trivial JavaScript codebase eventually grows a pile of logging glue:

• A server logger with Datadog and pretty output.
• A separate browser logger with Datadog browser logs and RUM.
• A "get the request id into the logs" helper that everyone copies.
• A different Jest/Vitest mock in every package.
• Ad-hoc error instanceof Error ? error.message : 'unknown' at hundreds of call sites.

Each of those is written slightly differently in every package. Sawdust is that layer, built once, correctly, and shared everywhere — so consuming code just imports logger and moves on.

What you get

• One API across runtimes — the same import and calls in Node, browser, and workers. The correct .node / .web build resolves automatically through conditional exports.
• A singleton façade that upgrades in place — import logger before it is configured; bootstrap later promotes it to a richer logger without breaking existing imports.
• Pluggable transports — console, pretty terminal, and Consola in core; Datadog server logs (+ APM trace injection), Datadog browser logs, and RUM via the @cues/sawdust-datadog provider. Enable per environment.
• Request-scoped context — withRequestContext uses AsyncLocalStorage on Node to stamp every log with requestId, userId, and route automatically.
• A service locator — shared libraries call getLogger() / getRumClient() and always get the canonical instance. No props, no providers, no DI container.
• Testing built in — resetLoggerLocator, noopLogger, and mock helpers make suites deterministic.
• Structured errors — round-trippable error serialization and context sanitization.

Where to go next

You want to…	Read
Install and log in 5 minutes	Getting Started
Sell it to your team	Why Sawdust?
See real Node / browser flows	Guides
Understand how it works inside	Architecture
Copy a battle-tested pattern	Pattern Catalog