The Right Way to Handle AI API Timeouts in Laravel

When integrating third-party AI or LLM APIs into your application, running them synchronously inside your standard web controllers is an architectural trap. If the external API takes 10+ seconds to respond, your server's worker pool can exhaust its memory constraints rapidly.

To solve this, you must shift external execution tracks into decoupled, asynchronous background processes. Here is a lean, production-ready Laravel Job setup designed to safely handle external API latency:

namespace App\Jobs;

use Illuminate\Bus\Queueable;

use Illuminate\Contracts\Queue\ShouldQueue;

use Illuminate\Foundation\Bus\Dispatchable;

use Illuminate\Queue\InteractsWithQueue;

class ProcessAIPipeline implements ShouldQueue

{

use Dispatchable, InteractsWithQueue, Queueable;

// Retry settings to survive external API timeouts safely

public $tries = 3;

public $backoff = [15, 45, 90];



protected array $data;



public function __construct(array $data)

{

    $this->data = $data;

}



public function handle(AIEngineService $ai): void

{

    // Off-thread processing protects your application core

    $response = $ai->generate($this->data['prompt']);

}

}

Key Architectural Advantages:

Immediate Client Release: Your controller handles rapid validation, dispatches the job, and instantly returns an HTTP 202 Accepted status to the client application.

Exponential Backoff Protection: If the AI provider hits a rate limit or unexpected downtime, your worker handles retries gracefully (15s, 45s, 90s) instead of throwing unhandled 500 errors.

Decoupled Extensibility: By shifting this logic off the HTTP thread, you can easily wrap this implementation into reusable, open-source vendor tools.

By building decoupled pipelines, you protect your infrastructure from connection timeouts and keep your core ecosystem incredibly resilient.

I’m a Software Engineer & Data Scientist (MSc) with 9+ years of experience specializing in high-concurrency Laravel systems and open-source utility design.