Lexem

Built Lexem because prompt engineering deserved the same discipline we already expect from code. Lexem gives AI prompts:

• Version control with commits, diffs, branching, merges and one-click rollback
• Eval and regression testing with exact match, regex and LLM-as-judge scoring
• AI-generated test cases for happy paths, edge cases and refusals
• AI diff analysis that explains prompt changes and their risks in plain English
• Eval-gated deploys across dev, staging and prod
• Token analytics per version and across prompts
• SDKs for TypeScript and Python to ship prompt updates without redeploying the app
• Fully open source and self-hostable on your own Postgres

The goal was simple: make prompt iteration safer, testable and production-ready.