Amp Laravel Package

Technical Evaluation

Architecture fit

• A pure Composer library (lullabot/amp), framework-agnostic: HTML in, AMP HTML plus a validation report out. It has no opinion about routing, caching, or persistence, which is exactly what you want.
• Its correct Laravel shape is a thin injected service (AmpConverter) behind a dedicated route group, not middleware. Any design that wraps the whole response pipeline couples the app to a lossy rewriter and is the wrong shape.
• The library assumes ownership of the entire document — it rewrites <html>/<head> and strips disallowed scripts. That is fine for server-rendered full pages, awkward for SPA/JSON delivery or already-cached fragments.

Integration feasibility

• Composer install is trivial; conversion is roughly loadHtml() → convertToAmpHtml() → getWarnings().
• The real friction is validation: it shells out to Node.js (amphtml-validator) via a bundled JS entry point. That is a new runtime dependency, not just a package. Slim PHP-FPM images and shared hosting will fail here.
• Transitive dependency on QueryPath pulls a legacy DOM stack that has historically lagged PHP 8.x. Needs a compatibility spike before commitment.

Technical risk

• License is NOASSERTION — GitHub detects no recognized license file. Legal review is a hard gate; this alone can block enterprise adoption.
• Project health is thin: ~375 stars, zero flagged dependents, effectively one maintainer, maintenance-cadence releases. The 2024-12-23 release is recent but the trend is stability, not investment.
• Strategic risk sits with AMP itself. Google removed AMP as a Top Stories requirement in 2021 and Core Web Vitals replaced it as a ranking signal. Remaining drivers are narrow: Google News/Discover eligibility for publishers, partner feeds, inherited contracts. Confirm the driver before accepting the operational cost.
• Conversion is lossy by design. Custom elements, arbitrary JS, and oversized inline CSS get stripped or rejected. Passing the validator is an iterative loop against a strict spec, not a one-time integration.

Key questions

1. What is the concrete business driver — Discover/News eligibility, a named partner, or legacy mandate? How is success measured?
2. Is Node.js available in the target runtime, and is out-of-band validation (CI/batch only, never per request) acceptable to stakeholders?
3. Which license actually governs the code, and who signs off?
4. Full-document conversion, or article-fragment conversion only?
5. Who owns the AMP template long-term, given upstream is no longer actively evolving the spec surface?

Integration Approach

Stack fit

• Strong fit for Laravel 10/11 as a service: bind a singleton in a package service provider, inject into an AmpController.
• Author AMP-first: a dedicated layouts/amp.blade.php plus Blade components (<x-amp-img>, <x-amp-youtube>) that emit compliant markup. Use the library as a safety net and warning source, not the primary authoring mechanism — fighting the converter is more expensive than writing correct markup.
• Cache aggressively: converted output is deterministic per content revision. Use response caching or a keyed amp_cache table on (model_id, updated_at).

Migration path

1. Spike (≈2 days). Install, convert one representative article, run the validator locally. Go/no-go on PHP version and QueryPath.
2. If green: add route + controller, convert the article body fragment only, route warnings to a dedicated log channel.
3. Add a queued ValidateAmpJob fired on publish — never validate on the request path.
4. Backfill the cache via an Artisan command.
5. Publish <link rel="amphtml"> on canonical pages last, once error counts are zero.

Compatibility

• Verify against the app's PHP minor version; QueryPath's PHP 8.2+ deprecation surface is the likely failure point.
• Blade output must be well-formed for the rewriter; unclosed tags and bare void elements are a common source of silent mis-conversion. Run a snapshot test on rendered output.

Sequencing

• Clear license and business driver first, in parallel — either can kill the initiative, and both are cheap to check.
• Then compatibility spike → template work (the bulk of the effort, not the library) → caching → observability → rollout.

Operational Impact

Maintenance

• Low bus factor. Plan to fork-and-vendor into an internal namespace within year one; budget 1–2 engineer-days per quarter tracking upstream. QueryPath patching for future PHP releases becomes your work, not the maintainer's.

Support

• Community-only, best-effort, no SLA. Documentation is Drupal-centric — expect to read the Drupal AMP module's source as the de facto usage guide. The AMP validator's errors are opaque and community familiarity has thinned.

Scaling

• Conversion is CPU-bound HTML parsing: low-milliseconds per article typically, unbounded on pathological input — cap input size.
• Validation spawns a Node process: tens to hundreds of milliseconds plus startup. Must be asynchronous. At volume, move it to a worker pool or a CI gate rather than production.
• Output is static, compresses well, and is CDN-friendly; there is no reason for it to touch PHP on the hot path.

Failure modes

• Missing node binary → validation throws while conversion still succeeds. Default to log-and-alert, not request failure.
• Malformed HTML → converter exceptions. Wrap in try/catch and fall back to the canonical URL.
• Silent stripping of unsupported components → pages render wrong but pass validation. Snapshot tests are the only defense.
• Inline CSS over the spec byte limit → styles vanish. Enforce a CSS byte budget in CI.

Ramp-up

• A mid-level engineer wires the library in a day. Reaching zero validator errors on a real template is the long tail — budget 1–2 weeks for a typical editorial site, more for rich designs.

Bottom line: technically feasible and cheap to start, but the value is narrowing as AMP's strategic relevance declines, the project is thinly maintained, and the license is unresolved. Treat it as a narrow, cached, asynchronous add-on — not a platform capability — and clear the business driver and legal questions before any engineering investment.