Show HN: Goldseam – heal broken Cypress selectors with a local LLM (github.com)

Goldseam has introduced an innovative solution for handling broken selectors in Cypress testing through a "bring-your-own-model" approach, allowing developers to repair their tests locally without relying on vendor cloud services. When a selector fails, the system captures a redacted snapshot of the Page Object Model and the error, after which a selected model (like Claude Code CLI, Ollama, or any OpenAI-compatible endpoint) proposes a minimal fix. This proposed change undergoes a six-step verification process before landing as a reviewable git diff. By focusing only on selector strings, Goldseam ensures that failures are transparently reported while maintaining the integrity of the tests. This development is significant for the AI and ML community as it emphasizes the importance of keeping testing processes streamlined and independent from vendor constraints. The open-source nature of Goldseam (licensed under MIT) coupled with its ability to integrate seamlessly into existing workflows (e.g., using cached commands stored in .goldseam-prompts) offers a trustworthy and efficient alternative for developers. Furthermore, the transparent nature of repairs ensures that developers can review and understand the fixes made, aligning with best practices in software development while leveraging machine learning effectively.