Weave Code
Ai Maria Db Store Laravel Package
symfony/ai-maria-db-store
MariaDB vector store integration for Symfony AI Store. Requires MariaDB 11.7+ for VECTOR columns, vector indexing, and distance search. Useful for building RAG and similarity search apps backed by MariaDB.
MariaDB vector store bridge for Symfony AI
Frequently asked questions about Ai Maria Db Store
- Can I use symfony/ai-maria-db-store directly in Laravel without Symfony?
- No, this package is Symfony-specific and requires Symfony’s AI Store interface. For Laravel, you’d need to create a bridge (e.g., a custom service provider or facade) to map Symfony’s `AiStoreInterface` to Laravel’s container. Alternatively, use raw PDO or a Laravel-compatible vector store like `laravel-ai/vector` for tighter integration.
- What Laravel versions support symfony/ai-maria-db-store?
- Laravel 10.x or 11.x with PHP 8.2+ can technically use this package, but it’s not natively supported. You’ll need to manually resolve Symfony’s dependencies (e.g., `symfony/dependency-injection`) and configure Laravel’s service container. Test thoroughly, as Symfony’s AI stack may introduce compatibility quirks.
- How do I configure MariaDB 11.7+ for vector search in Laravel?
- First, upgrade MariaDB to 11.7+ and enable the `vector` engine. Then, create a table with a `VECTOR` column (e.g., `embedding VECTOR(1536)`) and a `VECTOR INDEX`. For Laravel, use raw PDO or a repository pattern to avoid Eloquent conflicts. Example: `ALTER TABLE documents ADD COLUMN embedding VECTOR(1536); CREATE VECTOR INDEX idx_embedding ON documents(embedding);`
- What are the performance limits of MariaDB vector search in production?
- MariaDB’s vector search is CPU-bound and lacks GPU acceleration. Expect latency spikes for datasets >1M vectors or QPS >1K. For high throughput, consider caching (e.g., Redis) or sharding. Benchmark against alternatives like `pgvector` or dedicated vector databases (e.g., Milvus) if scalability is critical.
- How do I migrate existing vector data (e.g., from Elasticsearch) to MariaDB?
- Export your embeddings as CSV/JSON, then use Laravel’s DB migrations or raw SQL to insert into the `VECTOR` column. Example: `INSERT INTO documents (embedding) VALUES (VECTOR('...'));`. For large datasets, batch inserts and monitor MariaDB’s vector index rebuilds to avoid locks.
- Will this package work with Laravel Scout or Eloquent models?
- No, this package is designed for Symfony’s AI Store and doesn’t integrate with Laravel Scout or Eloquent’s query builder. Use raw PDO or a repository pattern to interact with MariaDB’s vector tables. For Eloquent, consider abstracting the vector store behind a custom trait or service.
- Are there alternatives to symfony/ai-maria-db-store for Laravel?
- Yes. For PostgreSQL, use `anahkiasen/pgvector` (via `laravel-ai/vector`). For dedicated vector DBs, try `weaviate/weaviate` or `milvus-io/milvus`. If you need a pure Laravel solution, explore `spatie/laravel-ai` or build a custom vector store with `brick/math` for embeddings.
- How do I test vector queries in Laravel’s testing environment?
- Mock MariaDB’s vector functions using Laravel’s `DatabaseMigrations` or `DatabaseTransactions`. For unit tests, use a library like `mockery` to stub PDO calls to `VECTOR` or distance functions. Example: `DB::shouldReceive('select')->andReturn([...]);` in your test cases.
- What distance metrics does MariaDB support for vector search?
- MariaDB 11.7+ supports cosine, Euclidean (L2), and inner product distance metrics natively. Custom metrics require raw SQL workarounds. Example: `SELECT * FROM documents ORDER BY embedding <=> ? LIMIT 10;` for cosine similarity.
- How do I handle schema changes (e.g., adding a new VECTOR column) in Laravel deployments?
- Use Laravel migrations to alter tables, but be cautious—`ALTER TABLE` on large tables with `VECTOR INDEX` can lock the database. Schedule migrations during low-traffic periods. Example: `Schema::table('documents', function (Blueprint $table) { $table->vector('embedding', 1536); });`
Weaver
How can I help you explore Laravel packages today?