The Problem: When Binary Becomes Text
In the modern web development world, efficiency is everything. Formats like WebP have become the de facto standard to guarantee lightweight images without sacrificing quality. However, when working in cloud development environments like AI Studio, integrated with GitHub, unexpected technical challenges can arise.
The symptom is frustrating: you upload a perfect WebP image to the development environment, see it correctly in the preview, but after clicking on "Publish" and pulling to the production server, the image is unreadable.
Why does this happen?
The problem lies in how some synchronization tools handle files during transfer to GitHub. In rare cases, the system can mistakenly interpret a binary file as if it were a text file, altering its bytes and making it unreadable to the browser.
Explored Solutions
1. Forcing file type with .gitattributes
We tried adding explicit rules in the .gitattributes file to force binary treatment, but in some sandboxed environments this may not be enough if the commit occurs via intermediate APIs.
2. The "Safe Folder" Workaround
A temporary solution was to keep a copy of the images in a protected folder on the server, restoring them manually after each pull. Effective, but not very practical.
The Definitive Solution
The winning strategy was to upload valid images directly to the public/ folder of the production server. Since these images are not touched by the development environment, Git does not overwrite them during the pull, and the build tool (Vite) copies them correctly into the final folder.
Lesson learned: When automatic tools fail, maintaining granular and direct control over binary assets is the key to project stability.