MySQL 8.4 Vector support

[corgab]

Hi! Now that MySQL 8.4 has native vector support, are there any plans to add official support (like VECTOR_DISTANCE) to the core, similar to what's available for Postgres?
If you're open to it, I'd love to take a crack at it and submit a PR

[angel-cm-dev]

Hi @corgab! This is an excellent and very timely proposal. With the rise of AI-driven applications, having native support for the VECTOR data type and VECTOR_DISTANCE() functions in the Laravel core would be a massive win for the community, especially for those of us building RAG systems on the MySQL stack.

To make this a "first-class citizen" in Laravel, I believe a robust implementation should cover these three fundamental areas:

1. Schema & Blueprint Support

We should be able to define vector columns directly in our migrations. A clean implementation would look like this:

$table->vector('embedding', 1536); // Support for specific dimensions

2. Database Grammar (The Core Task)
   The heavy lifting would happen in Illuminate\Database\Schema\Grammars\MySqlGrammar and Illuminate\Database\Query\Grammars\MySqlGrammar.

Schema: Adding the logic to compile the VECTOR type.

Query: Implementing a wrapper for distance functions (Cosine, Euclidean, or Dot Product).

3. Eloquent & Query Builder DX
   We could aim for a syntax similar to what some community packages do for PostgreSQL, allowing for clean distance-based ordering:

$users = User::query()
    ->orderByDistance('embedding', $queryVector, 'cosine')
    ->limit(5)
    ->get();

Why this belongs in the Core:
Developer Experience (DX): It removes the need for raw SQL or third-party wrappers for basic AI operations.

Consistency: It aligns MySQL's capabilities with what is already becoming standard in the pgvector/PostgreSQL ecosystem within Laravel.

Future-proofing: As LLMs become more integrated into web apps, vector search will be as common as full-text search.

If you are willing to start the PR, I’d suggest looking at how geometry types or fulltext indexes are handled in the current MySqlGrammar for inspiration. I’m sure many of us would be happy to help with testing or code review! 🚀