README

Use Eloquent natively on Manticore Search. This package registers Manticore as a first-class Laravel database connection over the MySQL protocol (port 9306), so the entire Eloquent / Query Builder API works unchanged — where, find, create, update, delete, paginate, chunk, scopes, casts, relations, migrations — plus Manticore helpers (match, knn, highlight, option, maxMatches, facet, replace).

This is a separate, ground-up library. If you use the older renatomaldonado/manticore-laravel-search (HTTP/JSON fluent builder + consolidation), keep using it for those features — this package does not replace it, it sits beside it.

Why a database driver?

"Make Eloquent work, just pointing at Manticore instead of the DB" is, by definition, a Laravel database driver. Manticore speaks the MySQL wire protocol, and all of Eloquent is built on a Connection + Grammar. So instead of re-implementing Eloquent method by method, we implement a thin Manticore grammar and inherit the whole framework.

Requirements

Installation

composer require renatomaldonado/laravel-manticore-eloquent
php artisan vendor:publish --provider="ManticoreEloquent\ManticoreEloquentServiceProvider" --tag=config

Configure config/manticore-eloquent.php (or the matching env vars):

'connection' => env('MANTICORE_CONNECTION', 'manticore'),
'host'       => env('MANTICORE_HOST', '127.0.0.1'),
'port'       => env('MANTICORE_PORT', 9306),   // MySQL protocol port
'username'   => env('MANTICORE_USERNAME'),
'password'   => env('MANTICORE_PASSWORD'),
'engine'     => env('MANTICORE_ENGINE'),       // 'columnar' | 'rowwise' | null

The package auto-registers a manticore connection in config/database.connections derived from this file, so you don't have to edit config/database.php.

Usage

Option A — a model that lives entirely in Manticore

Extend ManticoreModel; its default connection is Manticore.

use ManticoreEloquent\Eloquent\ManticoreModel;

class Article extends ManticoreModel
{
    protected $table = 'articles_rt';
    protected $guarded = [];
    protected $casts = ['published' => 'boolean', 'created_at' => 'datetime'];
}

Article::where('published', true)->match('laravel')->orderBy('created_at', 'desc')->paginate();
Article::find(10);
Article::create(['title' => 'Hello', 'body' => '…', 'published' => true]);
Article::where('id', 10)->update(['published' => false]);
Article::destroy(10);

Option B — a relational model that can also be queried via Manticore

Add the HasManticore trait; the model keeps its normal DB connection, and ::manticore() opens an Eloquent builder on Manticore.

use Illuminate\Database\Eloquent\Model;
use ManticoreEloquent\Concerns\HasManticore;

class Company extends Model
{
    use HasManticore;

    public function searchableAs(): string
    {
        return 'companies_rt'; // Manticore index name
    }
}

Company::query()->where('id', 1)->first();          // relational DB
Company::manticore()->match('fintech')->paginate(); // Manticore

Manticore-specific helpers

Article::manticore()
    ->match('cloud native', 'title')   // MATCH('@title cloud native')
    ->where('views', '>', 100)
    ->option('ranker', 'sph04')        // OPTION ranker=sph04
    ->maxMatches(5000)                 // OPTION max_matches=5000
    ->facet('category')                // FACET category
    ->get();

// Full-document write (REPLACE INTO) — needed to update full-text fields
Article::manticore()->replace(['id' => 1, 'title' => 'New', 'body' => '…']);

max_matches is raised automatically when you page beyond Manticore's default cap of 1000 rows, so deep paginate()/offset() just works.

KNN / vector search

Requires the column to be declared as a floatVector in the migration (see below).

// knn(`embedding`, 10, (0.12, 0.98, …))  — pair with knn_dist() to read the score
Article::manticore()
    ->knn('embedding', $queryVector, k: 10)   // optional ef: ->knn(..., ef: 2000)
    ->selectRaw('*, knn_dist() as distance')
    ->get();

Highlighting

Article::manticore()
    ->match('cloud native')
    ->highlight('body', ['before_match' => '<mark>', 'after_match' => '</mark>'])
    ->get();   // each row gets a `highlight` column (alias configurable)

Migrations

Manticore real-time indexes are created with standard Laravel migrations. The schema grammar maps Laravel column types to Manticore types and adds helpers for Manticore-only types (mva, mva64, floatVector).

use Illuminate\Database\Migrations\Migration;
use Illuminate\Database\Schema\Blueprint;
use Illuminate\Support\Facades\Schema;

return new class extends Migration {
    public function up(): void
    {
        Schema::connection('manticore')->create('articles_rt', function (Blueprint $table) {
            // `id` is implicit in Manticore — no need to declare it.
            $table->text('body');                       // text  (full-text indexed)
            $table->string('title');                    // string attribute
            $table->integer('views');                   // integer
            $table->bigInteger('author_id');            // bigint
            $table->float('score');                     // float
            $table->boolean('published');               // bool
            $table->json('meta');                       // json
            $table->timestamp('created_at');            // timestamp
            $table->mva('tag_ids');                     // multi  (uint set)
            $table->mva64('big_ids');                   // multi64
            $table->floatVector('embedding', dims: 384); // float_vector for KNN
            $table->text('summary')->indexed()->stored();// spell out text options

            // Table-level Manticore options (trailing key='value' on CREATE TABLE)
            $table->minInfixLen(2);                       // infix search
            $table->morphology('stem_en');                // or ->manticoreOptions([...])
            // $table->engine = 'columnar';
        });
    }

    public function down(): void
    {
        Schema::connection('manticore')->dropIfExists('articles_rt');
    }
};

Type mapping: string/char → string, text → text, integer/tinyInteger/smallInteger/mediumInteger → integer, bigInteger → bigint, float/double/decimal → float, boolean → bool, json/jsonb → json, timestamp/dateTime/date → timestamp, plus mva, mva64, floatVector. Text columns accept ->indexed()/->stored(), and tables accept minInfixLen(), morphology() and the generic manticoreOptions([...]).

Known limits & gotchas

• No transactions / row locks. Manticore has none; lock() / ->lockForUpdate() are no-ops, and DB transactions should not be relied on.
• update() vs replace(). Manticore UPDATE only changes attribute (non-text) columns in place. To rewrite a full document (including text fields) use replace().
• id is special. Manticore assigns the document id; migrations skip any id column.
• JOINs are limited. Cross-index JOINs in Manticore are restricted; prefer denormalized indexes. Eager-loading relations across a relational connection still works as usual.
• Prepared statements. The driver forces emulated prepares (Manticore's server-side prepare support is limited), so bindings — including MATCH(?) — are interpolated by PDO.

Tests

composer install
vendor/bin/pest --testsuite=Unit,Feature   # offline tests (no server needed)
# Integration tests require a running Manticore on the configured MySQL port
# (defaults to 9306; override with MANTICORE_HOST / MANTICORE_PORT):
MANTICORE_HOST=127.0.0.1 MANTICORE_PORT=9306 vendor/bin/pest --testsuite=Integration

A quick local Manticore for testing:

docker run --name manticore -p 9306:9306 -p 9308:9308 manticoresearch/manticore

Coverage (requires PCOV or Xdebug):

composer test:coverage        # full report
composer test:coverage-min    # fails under 95%

See docs/ARCHITECTURE.md for the design and the SQL differences handled.

License

MIT