Production deployment

A production Enverif installation has five logical components:

1. Nginx/Apache + PHP-FPM web application.
2. MySQL as authoritative state.
3. Redis for queues and cache.
4. One or more queue workers consuming agents and default.
5. Laravel scheduler executing once per minute.

Scaling

Web nodes are stateless except for the shared application key and configuration. Add queue workers horizontally for more agent throughput. The continuation job is unique until processing so duplicate dispatches for a run do not execute the same turn concurrently.

MySQL should use automated backups and point-in-time recovery where possible. Redis persistence is useful for queue durability but does not replace the MySQL run state.

Deploy

Run migrations before replacing workers, then restart queue workers so they load new code:

php artisan migrate --force
php artisan optimize
php artisan queue:restart

Keep APP_KEY stable across releases or encrypted connector/model credentials will become unreadable.