SQLSage: local review of PostgreSQL SELECTs generated by Drizzle

[Figo5]

If you're investigating a PostgreSQL SELECT generated by Drizzle, I've made a local CLI to review the SQL alongside its schema: SQLSage. I maintain the project; this post was prepared with AI assistance.

SQLSage explains the SELECT and flags supported correctness/performance risks. One example is ORDER BY created_at DESC LIMIT 50: when created_at isn't unique, it can flag ties at the boundary and suggest a unique tiebreaker as optional cleanup if stable results matter. It doesn't assume that every tie is a bug.

I checked the file workflow with a Drizzle 0.45.4 generated SELECT and matching supported CREATE TABLE DDL (a synthetic fixture, with no database execution). To try your own:

1. Use Drizzle's documented .toSQL() to get your SELECT text and save that SQL string as query.sql.
2. Save the actual supported PostgreSQL schema DDL as schema.sql; check it against SQLSage's supported inputs. This step needs SQL DDL, not the TypeScript table definitions.
3. With Node.js 22.18+:

npx --yes sqlsage@0.2.2 analyze --query query.sql --schema schema.sql --format markdown

The MIT CLI is free. This file workflow doesn't execute the query, upload inputs, connect to your database or require credentials. It reads SQL and supported DDL/catalog metadata; it isn't a Drizzle plugin and doesn't read TypeScript schemas directly. Saved PostgreSQL plans are optional. Parameters such as $1 bind, but advice is weaker without values; reports don't prove correct results or a speedup.

For one real SELECT you're already investigating, did the report suggest a useful next check, get something wrong, or fail on your input? A plain-language reply here is enough; the feedback issue is optional. Keep SQL, schemas, plans, full reports, credentials and private data local.