Skip to content

HOSTING.md

Latest commit

 

History

History
121 lines (88 loc) · 4.46 KB

HOSTING.md

File metadata and controls

121 lines (88 loc) · 4.46 KB

Scolta Hosting Guide

How Scolta Stores Data

Scolta produces a static search index (Pagefind format) stored in your web-accessible directory. The index is a set of small binary fragment files plus a JavaScript loader. Total size scales with content: roughly 1 MB per 1,000 pages.

Build state (progress tracking, chunk manifests) is stored in a non-public state directory and is safe to delete after a successful build.

Indexer Options

Scolta supports two indexers:

  • binary (default): Runs the Pagefind binary. Fast, native performance. Requires exec() — unavailable on some managed hosts.
  • php: Pure PHP indexer. No binary dependencies. Works everywhere PHP runs. Set indexer: php in config (or SCOLTA_INDEXER=php in .env).

Auto-detection (indexer: auto) tries the binary first, falls back to PHP.

For faster indexing and 33+ language support, use the binary indexer when exec() is available. The PHP indexer supports 14 languages (Snowball stemming) and has no external dependencies.

Platform-Specific Notes

WordPress — Managed Hosts

Host exec() Recommended Indexer Notes
WP Engine No php Ephemeral filesystem resets on deploy; rebuild via Action Scheduler
Kinsta No php
Flywheel No php
Pressable No php
WordPress.com Business No php Requires Business plan for plugin installation
Self-hosted / VPS Yes binary Best performance

Drupal — Managed Hosts

Host exec() Recommended Indexer Notes
Pantheon Limited php Use PHP indexer; filesystem is ephemeral outside /files
Acquia Yes binary Configure output_dir under /files for persistence
Platform.sh Yes binary Mount output directory as a persistent disk

Laravel — Cloud Hosts

Host exec() Recommended Indexer Notes
Vapor (serverless) No php Use S3 StorageDriver for state + index persistence
Forge Yes binary Standard VPS, no restrictions
Ploi Yes binary

Ephemeral Filesystems

Hosts like WP Engine, Pantheon, and Vapor reset the filesystem on deploy. The search index must be rebuilt after each deploy. Options:

  1. Auto-rebuild on deploy — trigger a build via deploy hook or post-deploy cron. WordPress: Action Scheduler. Drupal: drush scolta:build. Laravel: artisan scolta:build in a deploy step.

  2. Persistent storage — configure output_dir to a persistent mount or cloud bucket (see StorageDriver below).

  3. CI build — generate the index during CI and include it in the deployment artifact. The index is static files; committing them is fine for small sites.

StorageDriver for Cloud Persistence

For Vapor or other serverless environments, implement a custom StorageDriver that writes to S3 or GCS instead of the local filesystem. Scolta's StorageDriverInterface supports this pattern:

  • FilesystemDriver (default) — local disk, works everywhere.
  • Custom: implement Tag1\Scolta\Storage\StorageDriverInterface with your cloud SDK and bind it in the service container.
// Example: bind a custom driver in AppServiceProvider::register()
$this->app->bind(StorageDriverInterface::class, S3StorageDriver::class);

Cron / Scheduled Rebuilds

All platforms support scheduled rebuilds. If auto-rebuild on content change is insufficient (for example, content imported via external ETL), schedule a periodic full rebuild:

  • WordPress: Action Scheduler (bundled with WooCommerce, or standalone). Register a recurring action: as_schedule_recurring_action(time(), 86400, 'scolta_rebuild').

  • Drupal: Queue Worker via cron or Drush.

    drush scolta:build

    Add to your crontab or platform cron configuration.

  • Laravel: Laravel Scheduler.

    // In App\Console\Kernel::schedule()
$schedule->command('scolta:build')->daily();

Index Size Reference

Content size Index size (approx) Build time (PHP indexer)
100 pages ~1 MB < 1 s
1,000 pages ~10 MB ~10 s
10,000 pages ~80 MB ~90 s
50,000 pages ~350 MB ~7 min

The binary (Pagefind CLI) indexer is approximately 10× faster.

Fragment files are individually small (< 50 KB each) and served on demand by the browser — visitors only download the fragments needed for their query.