README

A production-safe API performance profiler for Laravel — in the spirit of Telescope/Pulse but focused exclusively on API performance. Low-overhead capture, an admin-gated dashboard, and a REST API for every metric.

• Package: ismailelbery/api-performance-analyzer
• Namespace: ApiPerformanceAnalyzer\
• Requires: PHP 8.2+, Laravel 11/12, MySQL or PostgreSQL
• Vendor prefix: apa

Install

composer require ismailelbery/api-performance-analyzer

# Publish config (optional) and run migrations
php artisan vendor:publish --tag=apa-config
php artisan migrate

The middleware auto-attaches to your api route group — no code changes needed. The dashboard lives at /apa and the REST API at /apa/api.

How capture works

Request → ApaMiddleware → [Collectors] → ProfileContext (in-memory)
                                              │ terminate()
                                              ▼
                                  ProfileStore (sync|queue|batch)
                                              ▼
                              apa_request_profiles (+ child rows)

• Lightweight always-on capture (timing/status) runs for every request.
• Retention is decided at the end of the request: a profile is kept when it is sampled or slow or an error. Child rows (queries, HTTP calls) are only stored for retained profiles.
• All writes happen in terminate() — never in the response path.

Configuration highlights (config/apa.php)

Storage drivers

• sync — inline write. Dev / low traffic only.
• queue — one StoreProfileJob per retained request. Default.
• batch — buffer in Redis and bulk-insert on size or via apa:flush (schedule it). Recommended for high traffic.

REST API

GET /apa/api/requests              list + filters (method,status,slow,n_plus_one,uri,from,to,per_page)
GET /apa/api/requests/{uuid}       single profile + queries + http calls
GET /apa/api/endpoints             per-endpoint stats (count, avg, p95, error_rate, n+1)  ?sort=p95|avg|count|error_rate|queries
GET /apa/api/stats/overview        window totals
GET /apa/api/stats/slow-queries    cross-request slow query analysis (by sql_hash)
GET /apa/api/stats/n-plus-one      aggregate N+1 suspects per endpoint

Commands

Suggested scheduler

$schedule->command('apa:prune')->daily();
$schedule->command('apa:rollup --recommend')->dailyAt('00:30');
$schedule->command('apa:flush')->everyMinute();      // only with the batch driver
$schedule->command('apa:alerts')->everyFiveMinutes(); // only if alerts enabled

Phases

• Phase 1 — timing, query count/time, memory, status; slow-endpoint ranking; request list/detail; REST API.
• Phase 2 — N+1 detection (per-request and aggregate), external HTTP timing (HttpCollector), daily rollups, slow-query analysis.
• Phase 3 — per-endpoint health scores (with insufficient-data guardrails), advisory index recommendations, threshold alert engine with cooldown, Slack/email notifications, PDF digest.

Production notes

• Run profiler storage on a separate connection; use the batch driver under heavy load.
• Sampling decides what to capture; the batch driver decides how cheaply to write it.
• The per-request cache kill switch disables capture during an incident without a deploy.
• Octane-safe: ProfileContext is a scoped binding reset on every request; collectors hold no cross-request state.

Tests

composer install
vendor/bin/phpunit