jonaspauleta/laravel-ai-moonshot

Moonshot AI (Kimi) provider for the Laravel AI SDK.

v1.0.0 2026-04-24 02:52 UTC

Requires

• php: ^8.5
• illuminate/contracts: ^11.0 || ^12.0 || ^13.0
• illuminate/http: ^11.0 || ^12.0 || ^13.0
• illuminate/support: ^11.0 || ^12.0 || ^13.0
• laravel/ai: ^0.6.3

Requires (Dev)

• driftingly/rector-laravel: ^2.2
• larastan/larastan: ^3.0
• laravel/pint: ^1.25
• orchestra/testbench: ^9.0 || ^10.0 || ^11.0
• pestphp/pest: ^3.0 || ^4.0
• pestphp/pest-plugin-laravel: ^3.0 || ^4.0
• phpstan/phpstan-deprecation-rules: ^2.0
• phpstan/phpstan-phpunit: ^2.0
• rector/rector: ^2.2

MIT

• João Paulo Santos

• laravel
• ai
• llm
• laravel-ai
• kimi
• moonshot
• kimi-k2

README

Moonshot AI (Kimi K2) provider for the Laravel AI SDK.

Moonshot's API is OpenAI-compatible (POST https://api.moonshot.ai/v1/chat/completions), so this driver supports text generation, streaming, tool calling, and thinking-mode reasoning (reasoning_content deltas are surfaced as Laravel AI SDK ReasoningStart / ReasoningDelta / ReasoningEnd stream events).

Requirements

• PHP 8.5+
• laravel/ai ^0.6.3

Install

composer require jonaspauleta/laravel-ai-moonshot

The service provider auto-registers via Laravel package discovery.

Configure

Set the API key:

MOONSHOT_API_KEY=sk-...

In config/ai.php:

'providers' => [
    // ...
    'moonshot' => [
        'driver' => 'moonshot',
        'name' => 'moonshot',
        'key' => env('MOONSHOT_API_KEY'),

        // Optional per-tier model overrides — defaults shown below.
        'models' => [
            'text' => [
                'default' => 'kimi-k2.6',
                'cheapest' => 'kimi-k2-0905-preview',
                'smartest' => 'kimi-k2-thinking',
            ],
        ],
    ],
],

Usage

use App\Ai\Agents\RaceEngineer;
use Illuminate\Broadcasting\PrivateChannel;

$agent = new RaceEngineer($user);

$agent
    ->forUser($user)
    ->broadcastOnQueue(
        'What is the best setup for the Nordschleife?',
        new PrivateChannel("agent.{$user->id}.{$requestId}"),
        provider: 'moonshot',
        model: 'kimi-k2.6',
    );

Thinking mode

Kimi's kimi-k2.6 supports persistent reasoning across turns with thinking.keep = all. Return the payload from your agent's providerOptions() and the gateway merges it into the chat-completions body unchanged:

public function providerOptions(Lab|string $provider): array
{
    return $provider === 'moonshot'
        ? ['thinking' => ['type' => 'enabled', 'keep' => 'all']]
        : [];
}

reasoning_content deltas then surface as ReasoningStart / ReasoningDelta / ReasoningEnd on the broadcast channel, before text streaming begins.

License

MIT © João Paulo Santos