timescale/pg-aiguide

Use the design-postgres-tables skill to draft, but not execute, a PostgreSQL schema for [describe the small application]. Start with only [list the two or three core entities] and the access patterns [list common lookups or joins]. Include appropriate data types, primary and foreign keys, semantic NOT NULL constraints, checks, and only the indexes justified by those access patterns. Target PostgreSQL [version]. Briefly explain each design choice.

Use the schema-exploration skill to answer this focused question about my PostgreSQL database: [question]. Work only within the authorized database and schemas [scope]. Begin with read-only catalog metadata, identify the most relevant schema-qualified objects and relationships, distinguish observations from inferences, and stop once you can answer. Do not inspect rows or data-derived statistics unless you first request my approval. If no database connection is available, work from the schema files I provide.

Use the pgvector-semantic-search skill to draft, but not execute, the default SQL setup for semantic search over [content type]. My embedding dimension is [N], and both stored and query embeddings come from [embedding model]. Include the extension, a minimal table, a default HNSW cosine index, and a parameterized top-[k] nearest-neighbor query with an explicit halfvec([N]) cast. Briefly identify the placeholders I must replace.