Louisa
Creator
Open-Source Developer Tools · 2026
Built Louisa, an open-source AI workflow that automatically generates polished release notes for GitHub and GitLab repositories, eliminating manual changelog writing by using Claude to enrich PRs at merge time and synthesize user-facing release notes on tag push.
Technology Partners
Challenge
Engineering teams waste hours every release cycle on changelog maintenance, pulling from vague PR titles, inconsistent commit messages, and half-written descriptions to produce something readable for users. The gap between raw commits and user-facing release notes was a consistent pain point across Arthur's own repositories.
Solution
Built Louisa as a Node.js workflow running on Vercel and GitHub Actions. Claude Haiku enriches each PR at merge time, rewriting descriptions with structured context, problem, solution, and user impact. On tag push, Claude Opus synthesizes those pre-computed summaries into polished, grouped release notes organized by product area. Instrumented end-to-end with OpenTelemetry and Arthur Engine for full observability into AI execution.
Impact
- •Eliminated changelog maintenance entirely, push a tag, production-quality release notes appear automatically
- •Designed a two-model AI strategy that balanced speed and quality at every stage of the pipeline
- •Built observability-first: every AI call instrumented with OpenTelemetry and Arthur Engine from day one
- •Shipped GitHub and GitLab support out of the box, usable by any engineering team, immediately
- •Extended to auto-generated blog posts and changelog publishing with zero additional engineering effort
Results
- •Published as open-source under the arthur-ai org on GitHub
- •Became a reference implementation for AI workflow design vs. autonomous agent design
- •Featured in Arthur AI blog on building production-ready AI systems
- •Demonstrated observability-first approach to shipping AI in the real world
Design
