“MVP or early-stage AI SaaS | Supabase + pgvector + LLM API”
+ single database for normal data and embeddings
“One database for normal data and embeddings.”
+ easier relationships between users, documents, and AI results
“Easier relationships between users, documents, and AI results.”
+ unified authentication and permission system
“One authentication and permission system.”
+ fewer services to deploy and maintain
“Fewer services to deploy and maintain.”
+ lower architectural complexity
“Lower architectural complexity.”
+ AI provider integrations and pgvector vector search
“Supabase supports integrations with AI providers and vector search using pgvector.”
− vector workloads can compete with normal application queries
“vector workloads can eventually compete with normal application queries for database resources.”
“For example, now offers VectorsDB, which supports vector storage, HNSW indexing, and similarity search.”
+ built-in vector capabilities (VectorsDB)
“now offers VectorsDB, which supports vector storage, HNSW indexing, and similarity search.”
pgvectorRecommendedpositive
“pgvector: Embeddings and semantic search.”
+ embeddings and semantic search
“pgvector: Embeddings and semantic search.”
“OpenAI or another LLM provider: Generation and reasoning.”