tzurot-db-vector

pgvector is a PostgreSQL extension that stores and queries vector embeddings for similarity search. This Skill provides Prisma patterns to define typed schemas, create ivfflat indexes for fast retrieval, and write migration-safe queries that retrieve the most similar embeddings to a given input vector.

Yes. This Skill enables type-safe database access over pgvector by defining Prisma models for embeddings, applying migrations to create vector columns and indexes, and writing typed queries that leverage PostgreSQL's similarity operators without raw SQL.

Use Prisma's migration-first workflow with idempotent SQL scripts. This Skill covers checksum-safe migration patterns, pre-commit safeguards to review generated SQL, and deployment considerations for Railway environments to ensure schema consistency across environments.

Implement connection pooling to prevent hitting connection limits in containerized or Railway environments. This Skill provides patterns for configuring pooled connections in Prisma, managing cold starts, and maintaining stable database access under load.

ivfflat indexing accelerates similarity search on large vector datasets by partitioning embeddings into clusters. This Skill demonstrates when to apply ivfflat indexes, how to create them safely via migrations, and trade-offs between index creation time and query speed.

Pgvector index modifications require careful planning to avoid blocking queries. This Skill provides safe patterns for altering indexes through migrations, includes pre-commit review steps for generated SQL, and outlines idempotent migration practices to ensure no duplicate index creation.