ember-auto-import

After years of wrestling with custom Ember-CLI import shims, ember-auto-import genuinely delivers on its "zero-config" promise. You install it, import any npm package with standard ES6 syntax, and it just works. No custom loaders, no manual webpack configuration, no special Ember-specific wrappers needed. The magic happens transparently at build time.

The developer experience is stellar because there's essentially nothing to learn. TypeScript users get full type inference through standard import statements. IDE autocompletion works perfectly since you're using real imports, not Ember's old "app.import()" pattern. Error messages are clear when builds fail, typically pointing directly to the problematic package rather than cryptic webpack errors.

Version 2.x brought significant performance improvements with its lazy-loading strategy. Upgrading from v1 was straightforward - mostly just bumping the version and clearing node_modules. The package handles dynamic imports intelligently, code-splitting them automatically. The only real learning curve is understanding how it interacts with tree-shaking, but the docs cover this adequately.

In production Ember apps, ember-auto-import has become essentially mandatory for consuming modern npm packages. The zero-config promise mostly holds true - you import from npm packages and it just works. The webpack-based bundling happens transparently, and the build performance is acceptable for most projects, though large dependency trees can add 10-30 seconds to cold builds.

The v2 rewrite brought significant memory improvements and better tree-shaking. Resource management during builds is solid - it properly releases webpack instances and doesn't leak memory during development rebuilds. The main operational concern is debugging when something goes wrong: error messages during failed imports can be cryptic, especially with packages that have complex export patterns or peer dependency issues.

Configuration is minimal but available when needed - you can tune webpack settings, adjust chunk splitting, and control which packages get bundled. The build output logging is adequate but not detailed enough for diagnosing performance bottlenecks without digging into webpack stats. Breaking changes between v1 and v2 required migration effort, but the v2 architecture is more stable.

In daily use, ember-auto-import does exactly what it promises: allows importing npm packages without manual webpack configuration. The v2 architecture using Embroider under the hood significantly improved build times compared to v1, especially in larger apps. Cold builds are reasonable, and incremental rebuilds are fast enough for productive development.

The zero-config approach works well for most npm packages, but you'll occasionally hit edge cases with packages that have complex side effects or unusual module setups. The webpack configuration escape hatch exists but requires understanding the underlying build pipeline. Memory usage during builds scales with app size but is generally acceptable - we've run it in CI with 4GB RAM without issues.

Error messages when imports fail can be cryptic, often requiring digging into webpack output. The lack of explicit resource management hooks means you're dependent on the build tool's behavior. Breaking changes between v1 and v2 required careful migration, though the payoff was worth it. Configuration options for chunking and tree-shaking exist but could be more discoverable.