Show HN: Pg-aiguide – Write better PostgreSQL code with AI (github.com)

Pg-aiguide is an AI-oriented PostgreSQL knowledge service that plugs into code assistants (public MCP server and a Claude Code plugin) to make generated Postgres schemas and queries substantially better. It offers version-aware semantic search across the official Postgres manual, curated “skills” encoding opinionated best practices (schema design, indexing strategies, data types, constraints, naming, performance tuning and modern PG features), and an extension docs corpus starting with TimescaleDB (pgvector and PostGIS planned). The repo supports easy integration across many tools (Claude, Codex, Cursor, Gemini CLI, VS Code, Windsurf) and is Apache‑2.0 with contribution guides for adding skills and docs. Technically, pg-aiguide feeds AI agents deep, versioned Postgres knowledge and proven patterns, so generated SQL is less likely to be outdated or missing essential constraints and indexes. In a demo comparing schemas produced with and without the service, outputs with pg-aiguide had 4× more constraints, ~55% more indexes (including partial/expression indexes), adopted PG17‑recommended patterns and modern features (GENERATED ALWAYS AS IDENTITY, NULLS NOT DISTINCT), and produced cleaner naming and documentation. That implies more robust, performant and maintainable DB designs from AI assistants — reducing runtime bugs and manual rework — while still requiring human review of opinionated recommendations.