What database setup is required for native vector search in Laravel 13?

Here is a complete, step-by-step example of setting up native vector search in Laravel 13. The example uses PostgreSQL with the pgvectorextension, which is the most common configuration.

Step 1: Database Setup (PostgreSQL)

Before writing any Laravel code, ensure your PostgreSQL database has the pgvector extension available. You can enable it directly from a Laravel migration using the helper method Schema::ensureVectorExtensionExists().

Step 2: Create the Migration

Create a migration for a documents table. The vector column stores the embedding, and the index() method creates an HNSW index for fast similarity searches.

use IlluminateDatabaseMigrationsMigration;
use IlluminateDatabaseSchemaBlueprint;
use IlluminateSupportFacadesSchema;

return new class extends Migration
{
   public function up(): void
   {
       // Ensures the pgvector extension is enabled (PostgreSQL only)
       Schema::ensureVectorExtensionExists();

       Schema::create( documents , function (Blueprint $table) {
           $table->id();
           $table->string( notitle );
           $table->text( content );
           // 1536 dimensions is standard for OpenAI is embedding models
           $table->vector( embedding , dimensions: 1536)->index();
           $table->timestamps();
       });
   }
};

Step 3: Configure the Model

Add the AsVector cast to your Eloquent model so Laravel automatically converts the database is vector format to a PHP array (and vice versa).

namespace AppModels;

use IlluminateDatabaseEloquentModel;
use IlluminateDatabaseEloquentCastsAsVector;

class Document extends Model
{
   protected $fillable = [ notitle , content , embedding ];

   protected function casts(): array
   {
       return [
            embedding => AsVector::class,
       ];
   }
}

Step 4: Generate and Store Embeddings

Use the Laravel AI SDK to generate an embedding for your text. You can use the toEmbeddings() helper on a string.

use IlluminateSupportStr;
use AppModelsDocument;

$content = "Napa Valley is famous for its wineries and beautiful landscapes.";

// Generate the embedding vector from the text
$embedding = Str::of($content)->toEmbeddings();

// Store the document with its embedding
Document::create([
    notitle => Napa Valley Wine Guide ,
    content => $content,
    embedding => $embedding,
]);

Step 5: Perform a Similarity Search

Use the whereVectorSimilarTo method on the query builder. If you pass a plain string, Laravel will automatically generate the embedding for you using your configured AI provider.

use AppModelsDocument;

// Search for documents semantically similar to the query
$documents = Document::query()
   ->whereVectorSimilarTo( embedding , best wineries in Napa Valley )
   ->limit(5)
   ->get();

foreach ($documents as $document) {
   echo $document->title . PHP_EOL;
}