Hand-writing mock SQL data for local testing is slow, and using LLMs breaks down when you need thousands of rows with strict relational constraints.
To solve this, I built SeedSQL—a free tool to generate referentially accurate synthetic data in seconds.
Why I Built
ItMocking local database schemas usually causes three main headaches:
Broken Foreign Keys: Standard generators treat tables independently, creating orphaned relationships (Users $\rightarrow$ Orders).
AI Limits: Prompting LLMs for large datasets gets slow, hits token caps, and yields non-deterministic outputs.
Formatting Hassles: Cleaning up raw text or CSVs into usable SQL takes extra time.
Key Features
Preserves Foreign Keys: Automatically maps parent-child primary keys across connected tables.
Flexible Data Types: Supports UUIDs, timestamps, custom regex patterns, localized names, and ranges.
Instant Export: Outputs ready-to-run INSERT statements (PostgreSQL, MySQL, SQLite) or structured JSON.
Deterministic Output: Re-run identical seeds across team dev setups and test pipelines.
I’d love to get feedback from developers using it:
Try the tool: https://seed-sql.ai.studio/
Feedback: What database engines or features would make your seeding workflow easier? Drop a comment below!
Top comments (0)