Laravel package: full Eloquent ORM over Manticore's MySQL protocol

Contributions
1

Hi everyone,

I’ve been running Manticore in production for about two years, mostly from a Laravel API. I ended up building something I haven’t seen in the ecosystem yet and wanted to share it here.

Since Manticore speaks the MySQL wire protocol, I registered it as a native Laravel database connection — a custom connection + a thin query/schema grammar. The result is that the entire Eloquent ORM works on Manticore directly: models, query builder, pagination, casts, and even migrations that create RT tables.

Article::where('published', true)
    ->match('laravel', '(title,body)')
    ->option('ranker', 'proximity_bm25')
    ->paginate();

Migrations map Laravel column types to Manticore types, plus Manticore-specific ones:

Schema::connection('manticore')->create('articles_rt', function (Blueprint $table) {
    $table->text('body');
    $table->mva('tag_ids');
    $table->floatVector('embedding', dims: 384); // KNN
    $table->minInfixLen(2);
    $table->morphology('stem_en');
});

KNN, FACET, OPTION, HIGHLIGHT and REPLACE are exposed as builder methods. Things that don’t map to MySQL semantics are handled explicitly — e.g. update() only touches attribute columns, and full-document rewrites go through an explicit replace(); transactions/locks are documented as unsupported.

Repo (MIT, 95% test coverage, Laravel 10/11/12):

It’s running in production on my side, but I’d really appreciate eyes from people who know Manticore internals better than I do — especially:

  1. Edge cases in SQL dialect differences I might have missed in the grammar (builder combinations that compile to SQL manticore rejects);
  2. Whether forcing PDO emulated prepares is still the right call for the MySQL endpoint, or if there’s a better approach for bound MATCH() parameters.

Happy to answer anything, and feedback/issues are very welcome.

2

Hi Renato,

This is very impressive. Thank you for building it and sharing it with the community.

1 Like
3

@Renato_Maldonado here’s a link to the Slack thread where Dario Rigolin is going to integrate your module into Hyperf.

1 Like