Download the PHP package wireninja/accelerator without Composer

WireNinja Accelerator

Opinionated Laravel 13 + Filament foundation for one solo developer shipping repeated single-VPS monoliths.

Fixed product contract

• Filament, System settings, Shield RBAC, activity logging, auth/MFA, custom sidebar/topbar, and Indonesian defaults are core.
• Root / belongs to userland.
• pnpm is default; npm is fallback.
• OAuth, PWA, Telegram, centralized realtime, Scout, and OpenTelemetry observability are optional integrations.
• PHP-FPM is the only client HTTP runtime.
• Queues always use Laravel's database driver.
• Client applications own zero Supervisor programs.
• Realtime is served by centralized Reverb; telemetry is exported directly to centralized OpenObserve.

Installation

The installer is resumable and intentionally destructive: package-owned recipe files are overwritten and the database may be rebuilt with migrate:fresh --seed. A completed .accelerator/install-state.json receipt triggers an explicit confirmation before reinstalling; automation must opt in with --force. Use the surgical migration workflow only when existing application code or data must be preserved.

Local development:

This runs Laravel, schedule:work, Pail, and Vite. Accelerator excludes Laravel's default queue listener because the package-owned schedule drains the database queue every 10 seconds with a bounded queue:work --stop-when-empty process.

Runtime

There is no client Octane, Horizon, local Reverb server, Nightwatch, NightOwl, OpenTelemetry Collector, telemetry PostgreSQL database, or Supervisor program.

Configuration ownership

Accelerator configures the application only. Easyploy owns all deployment topology and stage environments.

Deployment

All normal operations originate from the local Mac through Easyploy. Manual SSH is break-glass only.

Stable root is /var/www/{domain} with Easyploy-owned releases, shared, and current. Easyploy owns SSH, release switching, Nginx, a dedicated FPM pool, native cron, stage env transport, backups orchestration, and recovery. Accelerator contains no Deployer integration or public remote-operation commands.

The shared runtime .env remains owned by the SSH deploy user, uses the application runtime user's primary group, and has mode 0640. This keeps secrets private while allowing PHP-FPM and native cron to load the same configuration.

Dual stages run identical code and features. They isolate database, uploads, APP_KEY, sessions/cache namespace, Reverb credentials, OTLP credential and identity, backups, domain, and cron entry. Production promotion deploys the exact successful staging revision. A code rollback never reverses database migrations.

Centralized Reverb

Every {deployment_key}-{stage} uses its own app ID, key, secret, and allowed origin on the same centralized Reverb daemon. Accelerator writes the ignored .accelerator/reverb-apps.json registration payload; import it into the centralized server, clear its config cache, and restart Reverb before connecting the client. Clients do not bind a Reverb port and Nginx does not proxy websocket routes locally.

Rotate local credentials with accelerator:configure features --rotate-reverb-app. Stage credentials live in Easyploy's ignored environment files and are changed through easyploy env edit|push.

OpenObserve

Accelerator uses keepsuit/laravel-opentelemetry and OTLP/HTTP protobuf.

Required stage values:

OTEL_INSTRUMENTATION_HTTP_SERVER=false is mandatory: Accelerator starts HTTP telemetry only after session authentication resolves. Guest requests and guest logs are excluded from OTLP. Authenticated requests, CLI/scheduled work, and queue jobs remain observable. OpenObserve retains telemetry for 60 days. Never commit or print OTLP headers.

Queue timeouts

The scheduled database worker raises retry_after above its configured timeout and derives the overlap-lock lifetime from the longest worker bound. The default worker timeout remains configurable through ACCELERATOR_QUEUE_WORKER_TIMEOUT. Individual jobs may use Laravel's #[Timeout(...)], but that value must not exceed the worker timeout unless the worker configuration is raised too.

Backups

Backups remain stage-owned, support local plus optional S3-compatible storage, verify checksums, notify through operator Telegram, and require an exact backup ID for restore.

AI routing

Boost resources are canonical:

Never run Composer update on the VPS, print stage env contents, infer release/tag/push authority, or mutate another project.