An interactive platform for learning SQL by writing SQL. Instructors seed a database and author problems; students solve them in the browser against a real Postgres engine and get instant, automated grading.
- Instructors create a dataset (raw seed SQL that builds tables and inserts sample rows) and write problems with a hidden solution query. The solution is run when the problem is saved, and its result is cached as the answer key.
- Students open a problem, write a query in a Monaco editor, and run it. The query executes against a fresh, isolated database and the result is compared to the answer key. Feedback is immediate: correct, or where the output diverged.
Untrusted student SQL runs inside PGlite (Postgres compiled
to WebAssembly), spun up fresh in memory for every run and discarded afterwards.
Each execution happens in a worker thread that is hard-terminated if it exceeds a
time limit, which is the only reliable guard against runaway queries since PGlite
runs single-threaded (a SQL statement_timeout cannot interrupt it).
The grader compares the student's result to the cached answer key: column count, row count, then the row data, comparing cells by column position (so different aliases still pass) and respecting row order only when the problem requires it.
- Next.js (App Router) with server actions
- Prisma 7 with the
pgdriver adapter, against Neon Postgres - Better Auth (email and password)
- PGlite for the sandboxed query engine
- Tailwind CSS with shadcn/ui (Base UI)
Requires Node 20+ and pnpm.
pnpm installCreate a .env file (see .env.example):
DATABASE_URL= # Neon pooled connection string
DIRECT_URL= # Neon direct connection string (for migrations)
BETTER_AUTH_SECRET= # openssl rand -base64 32
BETTER_AUTH_URL=http://localhost:3000
Apply the schema and start the app:
pnpm db:deploy # apply migrations
pnpm devpnpm dev- start the dev serverpnpm build- production buildpnpm db:migrate- create and apply a migration in developmentpnpm db:deploy- apply migrations (production)pnpm db:studio- open Prisma Studio
Schema Lab is open source and contributions are welcome. The backend (SQL sandbox, grading, auth, data model) is in a working state; the frontend is intentionally plain, so UI and UX improvements are an especially good place to start. Some ideas:
- Redesign the problem-solving and instructor screens, or theme the app.
- Improve the query editor experience (schema autocomplete, result formatting).
- Add features on top of the existing model (leaderboards, hints, more problem types).
To contribute, fork the repo, create a branch, and open a pull request. For anything larger, opening an issue first to discuss the direction is appreciated. You are also free to fork it and build your own app on top of it.
Released under the MIT License - use it, modify it, and build your own apps with it.