The BMW manual was off-limits, so I built my friend something better

Developer Alex Georgiev has created a custom, mobile-first repair assistant for a BMW E90, designed to solve the frustration of scattered, unreliable automotive information. Rather than using copyrighted service manuals, Georgiev built a curated knowledge base using general automotive expertise, ensuring every torque spec and part number includes a verification caveat. The project, hosted on GitHub Pages, utilizes a serverless architecture with Google’s open-weight Gemma model via DigitalOcean to provide intelligent, context-aware answers. By avoiding a complex RAG pipeline in favor of a small, high-quality corpus, the system remains transparent and inspectable. The tool prioritizes safety and accuracy, with the AI instructed to refuse tasks beyond its scope, such as transmission rebuilds. This project demonstrates how open-source tools and thoughtful architecture can create a reliable, user-focused alternative to proprietary documentation, emphasizing the importance of provenance and human-in-the-loop verification in technical reference tools.