Concurrency - Knowledge of Laravel

Stdio mode

Starting a process per task is not efficient, but this pattern works reliably.

Because Concurrency serializes internal values, return a plain content() string instead of returning SessionEvent.

use Revolution\Copilot\Facades\Copilot;
use Illuminate\Support\Facades\Concurrency;

$prompt = 'Tell me something about Copilot.';

[$gpt5_response, $sonnet_response] = Concurrency::run([
    fn () => Copilot::run($prompt, config: ['model' => 'gpt-5.2'])->content(),
    fn () => Copilot::run($prompt, config: ['model' => 'claude-sonnet-4.5'])->content(),
]);

echo 'GPT-5 Response: '.$gpt5_response;
echo 'Claude Sonnet Response: '.$sonnet_response;

TCP mode

Since there is no process boot per request, TCP mode can look faster, but observed latency may still be similar.

With the process driver, timeouts around 60 seconds are common in either mode. With the CLI-only fork driver, one request may succeed while another fails. In observed tests, switching from gpt-5 to gpt-5.2 allowed both requests to complete.

Concurrency exists as a feature, but queue-based processing is still the safer default for long-running workloads.

For the latest updates, see the GitHub repository.

Documentation Index

​Parallel execution with Laravel Concurrency

​Stdio mode

​TCP mode

Parallel execution with Laravel Concurrency

Stdio mode

TCP mode