One thing building FastDoc CLI taught me:
Users rarely ask for the architecture you want to build. They ask for the features they need.
Supporting multiple AI providers forced me to rethink several design decisions
A simple "Test Cédric" push notification was enough to make millions of banking customers think they were being hacked.
One lesson:
In critical systems, users don't see intentions.
They only see outcomes.
@valerymelou Je travaille actuellement sur FastDoc CLI, un outil open source développé en TypeScript qui utilise l'IA pour générer automatiquement de la documentation technique à partir du code source.
La dernière version prend en charge OpenAI, Anthropic et Gemini.
The hardest part wasn't integrating multiple AI providers. It was redesigning an architecture originally built for one.
Building FastDoc CLI taught me that good feedback often leads to better software.
@alaphati_t Building FastDoc CLI.
This week was mostly about refactoring the architecture to support multiple AI providers without changing the user experience.
Today, users can choose their preferred AI provider during initialization. Building software teaches you an important lesson:
Good feedback often leads to better architecture.
One thing I learned while building FastDoc CLI:
Adding support for multiple AI providers wasn't the hardest part.
The hardest part was redesigning the architecture.
FastDoc CLI was originally built around Anthropic only.
After feedback from developers, I decided to support OpenAI and Gemini as well. That forced me to rethink parts of the project that were tightly coupled to a single provider.
One thing I learned while building FastDoc CLI:
Developers don't hate documentation. They hate spending hours writing and maintaining it.
That's one of the reasons I started building an AI-powered documentation generator.
#buildinpublic#typescript#opensource