Laravel 12 AI SDK: Build Real AI Features with Code

Laravel 12 ships with an official AI SDK that lets you build AI features using the same patterns you already know:

services, facades, agents, jobs, queues, tests.

No weird abstractions. No external glue code.

If you can write a Laravel controller, you can ship AI features.

Installation

composer require laravel/ai
php artisan vendor:publish --provider="Laravel\Ai\AiServiceProvider"
php artisan migrate

Add your provider key:

OPENAI_API_KEY=sk-xxxx

Laravel supports multiple providers, but we’ll use OpenAI in examples.

Agents: Your AI Logic Lives Here

Agents are reusable AI classes.
Think of them like Service classes with memory.

Create one:

php artisan make:agent SupportAgent

Basic Agent

use Laravel\Ai\Contracts\Agent;
use Laravel\Ai\Concerns\Promptable;

class SupportAgent implements Agent
{
    use Promptable;

    public function instructions(): string
    {
        return <<<TEXT
You are a customer support assistant for an ecommerce store.
Be polite, short, and helpful.
TEXT;
    }
}

Use it anywhere:

$response = (new SupportAgent)->prompt(
    'My order has not arrived yet'
);

echo $response->content();

Example 1: Customer Support Chatbot (With Memory)

Controller Example

class ChatController extends Controller
{
    public function send(Request $request)
    {
        $reply = SupportAgent::make(user: auth()->user())
            ->prompt($request->message);

        return response()->json([
            'reply' => $reply->content(),
        ]);
    }
}

Why This Is Powerful

• Laravel automatically stores conversation history
• Each user has contextual memory
• No Redis / vector DB setup needed

Example 2: Structured AI Output (JSON You Can Trust)

Let’s analyze product reviews.

php artisan make:agent ReviewAnalyzer --structured

Agent Code

use Laravel\Ai\Schema\JsonSchema;

class ReviewAnalyzer implements Agent
{
    use Promptable;

    public function instructions(): string
    {
        return 'Analyze the customer review.';
    }

    public function schema(JsonSchema $schema): array
    {
        return [
            'sentiment' => $schema->string()->required(),
            'rating' => $schema->integer()->min(1)->max(5)->required(),
            'summary' => $schema->string()->required(),
        ];
    }
}

Usage

$result = (new ReviewAnalyzer)->prompt(
    'The delivery was fast but packaging was damaged.'
);

$data = $result->structured();

$data['sentiment']; // mixed
$data['rating'];    // 3
$data['summary'];   // Short summary

💡 Perfect for dashboards, reports, and automations.

 Example 3: Semantic Search with Embeddings (RAG)

Migration

Schema::create('documents', function (Blueprint $table) {
    $table->id();
    $table->text('content');
    $table->vector('embedding', 1536);
});

Saving Documents

use Illuminate\Support\Str;

Document::create([
    'content' => $text,
    'embedding' => Str::of($text)->toEmbeddings(),
]);

Searching by Meaning

$queryEmbedding = Str::of(
    'How do I reset my password?'
)->toEmbeddings();

$results = Document::query()
    ->whereVectorSimilarTo('embedding', $queryEmbedding)
    ->limit(3)
    ->get();

This works even if:

• No keywords match
• Text is phrased differently

Example 4: AI Knowledge Base (RAG + Agent)

class DocsAgent implements Agent
{
    use Promptable;

    public function instructions(): string
    {
        return 'Answer questions using provided documentation.';
    }
}

$docs = Document::searchByMeaning($question);

$answer = (new DocsAgent)->prompt(
    "Docs:\n{$docs}\n\nQuestion: {$question}"
);

Used for:

• Internal tools
• SaaS help centers
• Company knowledge bots

 Example 5: Image Generation

use Laravel\Ai\Facades\Image;

$image = Image::of(
    'A clean SaaS landing page illustration'
)->generate();

$image->save(storage_path('app/landing.png'));

Use cases:

• Blog banners
• Marketing assets
• Placeholder images

Example 6: Speech → Text

use Laravel\Ai\Facades\Transcription;

$transcript = Transcription::fromStorage(
    'support-call.mp3'
)->generate();

echo $transcript->text();

Great for:

• Support calls
• Meetings
• Voice notes

Example 7: Text → Speech

use Laravel\Ai\Facades\Audio;

$audio = Audio::of(
    'Your order has been shipped!'
)->generate();

$audio->save(storage_path('app/tts.mp3'));

Testing AI (Yes, Really)

use Laravel\Ai\Facades\Ai;

Ai::fake([
    'Hello' => 'Hi! How can I help?',
]);

$response = (new SupportAgent)->prompt('Hello');

$this->assertEquals(
    'Hi! How can I help?',
    $response->content()
);

No API calls. No billing surprises.

When Should Laravel Devs Use AI SDK?

Use it when you need:

• Smart search
• Chatbots
• Content generation
• Summarization
• Voice features
• Data extraction

Avoid it when:

• Rules-based logic is enough
• Deterministic output is required

Final Thoughts

Laravel AI SDK doesn’t try to hide AI.

It makes AI feel like Laravel:

• Artisan commands
• Facades
• Testability
• Clean architecture

If you already build:

• APIs
• SaaS apps
• Admin panels

You can now add real AI features without leaving Laravel.