benbjurstrom/pgvector-scout
Laravel Scout driver for PostgreSQL pgvector. Store embeddings on your models and run fast vector similarity search directly in Postgres. Supports multiple embedding indexes (OpenAI, Gemini, testing) with publishable config and easy setup.
Architecture fit: Excellent fit for Laravel projects using PostgreSQL with pgvector extension. Seamlessly integrates with Laravel Scout's abstraction layer, leveraging existing Scout patterns (traits, observers) without architectural overhaul. The config-driven index system aligns with Laravel's convention-over-configuration philosophy.
Integration feasibility: High. Simple Composer install, config publishing, and migration steps. Demo repo provides clear reference implementation. Requires only minimal model changes (adding traits and searchableAs() method). External API dependencies (OpenAI/Gemini) are configurable via environment variables.
Technical risk: Moderate. Low version (v0.3.3) and zero dependents indicate limited real-world validation. External embedding provider APIs introduce single points of failure. Requires PostgreSQL with pgvector extension (not standard in all environments). No explicit version compatibility matrix in docs – may break with future Laravel releases.
Key questions: How does it handle >1M embeddings in production? What's the performance impact of vector similarity searches at scale? How are API rate limits/retries managed for embedding generation? Are there known issues with Laravel 10+? How does it handle schema changes to embedding tables?
Stack fit: Ideal for PostgreSQL-backed Laravel applications already using Scout. Requires PostgreSQL
How can I help you explore Laravel packages today?